Change log

1.1.0 (2024-06-06)


  • Deprecated boto is replaced with boto3 (used to handle AWS S3 downloads). Fixes #5597 via PR #7340 (by @mslw, @effigies, and @yarikoptic). Remaining issues:

    • no download progress indication,

    • no “Range” support (for partial downloads).


  • Retry logic for S3 connections is now handed over to Boto3 and its standard mode, removing our custom method. PR #7340

1.0.3 (2024-06-06)

Bug Fixes

  • Raise exception if an annex remote process without console tries to interact with the user, e.g. prompt for a password. PR #7578 (by @christian-monch)

  • Fix add-archive-content for patool>=2.0. PR #7603 (by @dguibert)



  • Stop testing on Python 3.7. Switch MacOS tests to 3.11, include 3.11 in Appveyor, and use 3.8 for other tests. Fixes #7584 via PR #7585 (by @mslw)

  • Convert .travis.yml to GitHub Actions workflow. Fixes #7574 via PR #7600 (by @jwodder)

  • Cancel lengthy running workflows if a new commit is pushed. PR #7601 (by @jwodder)

1.0.2 (2024-04-19)


1.0.1 (2024-04-17)


  • The main entrypoint for annex remotes now also runs the standard extension load hook. This enables extensions to alter annex remote implementation behavior in the same way than other DataLad components. (by @mih)

1.0.0 (2024-04-06)

Breaking Changes

Enhancements and New Features

0.19.6 (2024-02-02)

Enhancements and New Features

  • Add the “http_token” authentication mechanism which provides ‘Authentication: Token {TOKEN}’ header. PR #7551 (by @yarikoptic)


0.19.5 (2023-12-28)


  • Fix text to account for a recent change in git-annex dropping sub-second clock precision. As a result we might not report push of git-annex branch since there would be none. PR #7544 (by @yarikoptic)

0.19.4 (2023-12-13)

Bug Fixes

  • Update target detection for adjusted mode datasets has been improved. Fixes #7507 via PR #7522 (by @mih)

  • Fix typos found by new codespell 2.2.6 and also add checking/fixing “hidden files”. PR #7530 (by @yarikoptic)




  • Cache value of the has_symlink_capability to spare some cycles. PR #7471 (by @yarikoptic)

  • RF(TST): use setup_method and teardown_method in TestAddArchiveOptions. PR #7488 (by @yarikoptic)

  • Announce test_clone_datasets_root xfail on github osx. PR #7489 (by @yarikoptic)

  • Inform asv that there should be no warmup runs for time_remove benchmark. PR #7505 (by @yarikoptic)

  • BF(TST): Relax matching of git-annex error message about unsafe drop, which was changed in 10.20231129-18-gfd0b510573. PR #7541 (by @yarikoptic)

0.19.3 (2023-08-10)

Bug Fixes

  • Type annotate get_status_dict and note that we can pass Exception or CapturedException which is not subclass. PR #7403 (by @yarikoptic)

  • BF: create-sibling-gitlab used to raise a TypeError when attempting a recursive operation in a dataset with uninstalled subdatasets. It now raises an impossible result instead. PR #7430 (by @adswa)

  • Pass branch option into recursive call within Install - for the cases whenever install is invoked with URL(s). Fixes #7461 via PR #7463 (by @yarikoptic)

  • Allow for reckless=ephemeral clone using relative path for the original location. Fixes #7469 via PR #7472 (by @yarikoptic)


  • Fix a property name and default costs described in “getting subdatasets” section of get documentation. Fixes #7458 via PR #7460 (by @mslw)



  • Disable some S3 tests of their VCR taping where they fail for known issues. PR #7467 (by @yarikoptic)

0.19.2 (2023-07-03)

Bug Fixes

  • Remove surrounding quotes in output filenames even for newer version of annex. Fixes #7440 via PR #7443 (by @yarikoptic)


  • DOC: clarify description of the “install” interface to reflect its convoluted behavior. PR #7445 (by @yarikoptic)

0.19.1 (2023-06-26)


  • Make compatible with upcoming release of git-annex (next after 10.20230407) and pass explicit core.quotepath=false to all git calls. Also added tools/find-hanged-tests helper. PR #7372 (by @yarikoptic)


  • Adjust tests for upcoming release of git-annex (next after 10.20230407) and ignore DeprecationWarning for pkg_resources for now. PR #7372 (by @yarikoptic)

0.19.0 (2023-06-14)

Enhancements and New Features

  • Address gitlab API special character restrictions. PR #7407 (by @jsheunis)

  • BF: The default layout of create-sibling-gitlab is now collection. The previous default, hierarchy has been removed as it failed in –recursive mode in different edgecases. For single-level datasets, the outcome of collection and hierarchy is identical. PR #7410 (by @jsheunis and @adswa)

Bug Fixes

  • WTF - bring back and extend information on metadata extractors etc, and allow for sections to have subsections and be selected at both levels PR #7309 (by @yarikoptic)

  • BF: Run an actual git invocation with interactive commit config. PR #7398 (by @adswa)


  • Raise minimal version of tqdm (progress bars) to v.4.32.0 PR #7330 (by @mslw)



  • Remove nose-based testing utils and possibility to test extensions using nose. PR #7261 (by @yarikoptic)

0.18.5 (2023-06-13)

Bug Fixes

  • More correct summary reporting for relaxed (no size) –annex. PR #7050 (by @yarikoptic)

  • ENH: minor tune up of addurls to be more tolerant and “informative”. PR #7388 (by @yarikoptic)

  • Ensure that data generated by timeout handlers in the asynchronous runner are accessible via the result generator, even if no other other events occur. PR #7390 (by @christian-monch)

  • Do not map (leave as is) trailing / or  in github URLs. PR #7418 (by @yarikoptic)




0.18.4 (2023-05-16)

Bug Fixes

  • Provider config files were ignored, when CWD changed between different datasets during runtime. Fixes #7347 via PR #7357 (by @bpoldrack)


  • Added a workaround for an issue with documentation theme (search function not working on Read the Docs). Fixes #7374 via PR #7385 (by @mslw)



  • Fix failing testing on CI PR #7379 (by @yarikoptic)

    • use sample S3 url DANDI archive,

    • use our copy of old .deb from instead of snapshots.d.o

    • use specific miniconda installer for py 3.7.

0.18.3 (2023-03-25)

Bug Fixes

  • Fixed that the get command would fail, when subdataset source-candidate-templates where using the path property from .gitmodules. Also enhance the respective documentation for the get command. Fixes #7274 via PR #7280 (by @bpoldrack)

  • Improve up-to-dateness of config reports across manager instances. Fixes #7299 via PR #7301 (by @mih)

  • BF: GitRepo.merge do not allow merging unrelated unconditionally. PR #7312 (by @yarikoptic)

  • Do not render (empty) WTF report on other records. PR #7322 (by @yarikoptic)

  • Fixed a bug where changing DataLad’s log level could lead to failing git-annex calls. Fixes #7328 via PR #7329 (by @bpoldrack)

  • Fix an issue with uninformative error reporting by the datalad special remote. Fixes #7332 via PR #7333 (by @bpoldrack)

  • Fix save to not force committing into git if reference dataset is pure git (not git-annex). Fixes #7351 via PR #7355 (by @yarikoptic)


  • Include a few previously missing commands in html API docs. Fixes #7288 via PR #7289 (by @mslw)



0.18.2 (2023-02-27)

Bug Fixes



  • Codespell more (CHANGELOGs etc) and remove custom CLI options from tox.ini. PR #7271 (by @yarikoptic)


0.18.1 (2023-01-16)

Bug Fixes

  • Fixes crashes on windows where DataLad was mistaking git-annex 10.20221212 for a not yet released git-annex version and trying to use a new feature. Fixes #7248 via PR #7249 (by @bpoldrack)



  • Integrate buffer size optimization from datalad-next, leading to significant performance improvement for status and diff. Fixes #7190 via PR #7250 (by @bpoldrack)

0.18.0 (2022-12-31)

Breaking Changes

  • Move all old-style metadata commands aggregate_metadata, search, metadata and extract-metadata, as well as the cfg_metadatatypes procedure and the old metadata extractors into the datalad-deprecated extension. Now recommended way of handling metadata is to install the datalad-metalad extension instead. Fixes #7012 via PR #7014

  • Automatic reconfiguration of the ORA special remote when cloning from RIA stores now only applies locally rather than being committed. PR #7235 (by @bpoldrack)

Enhancements and New Features

  • A repository description can be specified with a new --description option when creating siblings using create-sibling-[gin|gitea|github|gogs]. Fixes #6816 via PR #7109 (by @mslw)

  • Make validation failure of alternative constraints more informative. Fixes #7092 via PR #7132 (by @bpoldrack)

  • Saving removed dataset content was sped-up, and reporting of types of removed content now accurately states dataset for added and removed subdatasets, instead of file. Moreover, saving previously staged deletions is now also reported. PR #6784 (by @mih)

  • foreach-dataset command got a new possible value for the –output-streamns|–o-s option ‘relpath’ to capture and pass-through prefixing with path to subds. Very handy for e.g. running git grep command across subdatasets. PR #7071 (by @yarikoptic)

  • New config datalad.create-sibling-ghlike.extra-remote-settings.NETLOC.KEY=VALUE allows to add and/or overwrite local configuration for the created sibling by the commands create-sibling-<gin|gitea|github|gitlab|gogs>. PR #7213 (by @matrss)

  • The siblings command does not concern the user with messages about inconsequential failure to annex-enable a remote anymore. PR #7217 (by @bpoldrack)

  • ORA special remote now allows to override its configuration locally. PR #7235 (by @bpoldrack)

  • Added a ‘ria’ special remote to provide backwards compatibility with datasets that were set up with the deprecated ria-remote. PR #7235 (by @bpoldrack)

Bug Fixes

  • When create-sibling-ria was invoked with a sibling name of a pre-existing sibling, a duplicate key in the result record caused a crashed. Fixes #6950 via PR #6952 (by @adswa)


  • create-sibling-ria’s docstring now defines the schema of RIA URLs and clarifies internal layout of a RIA store. PR #6861 (by @adswa)

  • Move maintenance team info from issue to CONTRIBUTING. PR #6904 (by @adswa)

  • Describe specifications for a DataLad GitHub Action. PR #6931 (by @thewtex)

  • Fix capitalization of some service names. PR #6936 (by @aqw)

  • Command categories in help text are more consistently named. PR #7027 (by @aqw)

  • DOC: Add design document on Tests and CI. PR #7195 (by @adswa)

  • was extended with up-to-date information on CI logging, changelog and release procedures. PR #7204 (by @yarikoptic)


  • Allow EnsureDataset constraint to handle Path instances. Fixes #7069 via PR #7133 (by @bpoldrack)

  • Use looseversion.LooseVersion as drop-in replacement for distutils.version.LooseVersion Fixes #6307 via PR #6839 (by @effigies)

  • Use –pathspec-from-file where possible instead of passing long lists of paths to git/git-annex calls. Fixes #6922 via PR #6932 (by @yarikoptic)

  • Make clone_dataset() better patchable ny extensions and less monolithic. PR #7017 (by @mih)

  • Remove simplejson in favor of using json. Fixes #7034 via PR #7035 (by @christian-monch)

  • Fix an error in the command group names-test. PR #7044 (by @christian-monch)

  • Move eval_results() into interface.base to simplify imports for command implementations. Deprecate use from interface.utils accordingly. Fixes #6694 via PR #7170 (by @adswa)


  • Use regular dicts instead of OrderedDicts for speedier operations. Fixes #6566 via PR #7174 (by @adswa)

  • Reimplement get_submodules_() without get_content_info() for substantial performance boosts especially for large datasets with few subdatasets. Originally proposed in PR #6942 by @mih, fixing #6940. PR #7189 (by @adswa). Complemented with PR #7220 (by @yarikoptic) to avoid O(N^2) (instead of O(N*log(N)) performance in some cases.

  • Use –include=* or –anything instead of –copies 0 to speed up get_content_annexinfo. PR #7230 (by @yarikoptic)


  • Re-enable two now-passing core test on Windows CI. PR #7152 (by @adswa)

  • Remove the with_testrepos decorator and associated tests for it Fixes #6752 via PR #7176 (by @adswa)

0.17.10 (2022-12-14)

Enhancements and New Features

  • Enhance concurrent invocation behavior of If possible invocations are serialized instead of raising re-enter runtime errors. Deadlock situations are detected and runtime errors are raised instead of deadlocking. Fixes #7138 via PR #7201 (by @christian-monch)

  • Exceptions bubbling up through CLI are now reported on including their chain of cause. Fixes #7163 via PR #7210 (by @bpoldrack)

Bug Fixes

  • BF: read RIA config from stdin instead of temporary file. Fixes #6514 via PR #7147 (by @adswa)

  • Prevent doomed annex calls on files we already know are untracked. Fixes #7032 via PR #7166 (by @adswa)

  • Comply to Posix-like clone URL formats on Windows. Fixes #7180 via PR #7181 (by @adswa)

  • Ensure that paths used in the datalad-url field of .gitmodules are posix. Fixes #7182 via PR #7183 (by @adswa)

  • Bandaids for export-to-figshare to restore functionality. PR #7188 (by @adswa)

  • Fixes hanging threads when close() or del where called in BatchedCommand instances. That could lead to hanging tests if the tests used the @serve_path_via_http()-decorator Fixes #6804 via PR #7201 (by @christian-monch)

  • Interpret file-URL path components according to the local operating system as described in RFC 8089. With this fix,'file:...').localpath returns a correct local path on Windows if the RI is constructed with a file-URL. Fixes #7186 via PR #7206 (by @christian-monch)

  • Fix a bug when retrieving several files from a RIA store via SSH, when the annex key does not contain size information. Fixes #7214 via PR #7215 (by @mslw)

  • Interface-specific (python vs CLI) doc generation for commands and their parameters was broken when brackets were used within the interface markups. Fixes #7225 via PR #7226 (by @bpoldrack)


  • Fix documentation of to not accept strings. Instead, encoding must be ensured by the caller. Fixes #7145 via PR #7155 (by @bpoldrack)



0.17.9 (2022-11-07)

Bug Fixes

  • Various small fixups ran after looking post-release and trying to build Debian package. PR #7112 (by @yarikoptic)

  • BF: Fix add-archive-contents try-finally statement by defining variable earlier. PR #7117 (by @adswa)

  • Fix RIA file URL reporting in exception handling. PR #7123 (by @adswa)

  • HTTP download treated ‘429 - too many requests’ as an authentication issue and was consequently trying to obtain credentials. Fixes #7129 via PR #7129 (by @bpoldrack)


  • Unrestrict pytest and pytest-cov versions. PR #7125 (by @jwodder)

  • Remove remaining references to nose and the implied requirement for building the documentation Fixes #7100 via PR #7136 (by @bpoldrack)


  • Use datalad/release-action. Fixes #7110. PR #7111 (by @jwodder)

  • Fix all logging to use %-interpolation and not .format, sort imports in touched files, add pylint-ing for % formatting in log messages to tox -e lint. PR #7118 (by @yarikoptic)


  • Increase the upper time limit after which we assume that a process is stalling. That should reduce false positives from, without impacting the runtime of passing tests. PR #7119 (by @christian-monch)

  • XFAIL a check on length of results in test_gracefull_death. PR #7126 (by @yarikoptic)

  • Configure Git to allow for “file” protocol in tests. PR #7130 (by @yarikoptic)

0.17.8 (2022-10-24)

Bug Fixes

  • Prevent adding duplicate entries to .gitmodules. PR #7088 (by @yarikoptic)

  • [BF] Prevent double yielding of impossible get result Fixes #5537. PR #7093 (by @jsheunis)

  • Stop rendering the output of internal subdatset() call in the results of run_procedure(). Fixes #7091 via PR #7094 (by @mslw & @mih)

  • Improve handling of --existing reconfigure in create-sibling-ria: previously, the command would not make the underlying git init call for existing local repositories, leading to some configuration updates not being applied. Partially addresses via (by @mslw)

  • Ensure subprocess environments have a valid path in os.environ['PWD'], even if a Path-like object was given to the runner on subprocess creation or invocation. Fixes #7040 via PR #7107 (by @christian-monch)

  • Improved reporting when using dry-run with github-like create-sibling* commands (-gin, -gitea, -github, -gogs). The result messages will now display names of the repositories which would be created (useful for recursive operations). PR #7103 (by @mslw)

0.17.7 (2022-10-14)

Bug Fixes


  • Do not use gen4-metadata methods in datalad metadata-command. PR #7001 (by @christian-monch)

  • Revert “Remove chardet version upper limit” (introduced in 0.17.6~11^2) to bring back upper limit <= 5.0.0 on chardet. Otherwise we can get some deprecation warnings from requests PR #7057 (by @yarikoptic)

  • Ensure that BatchedCommandError is raised if the subprocesses of BatchedCommand fails or raises a CommandError. PR #7068 (by @christian-monch)

  • RF: remove unused code str-ing PurePath. PR #7073 (by @yarikoptic)

  • Update GitHub Actions action versions. PR #7082 (by @jwodder)


  • Fix broken test helpers for result record testing that would falsely pass. PR #7002 (by @bpoldrack)

0.17.6 (2022-09-21)

Bug Fixes

  • UX: push - provide specific error with details if push failed due to permission issue. PR #7011 (by @yarikoptic)

  • Fix datalad –help to not have Global options empty with python 3.10 and list options in “options:” section. PR #7028 (by @yarikoptic)

  • Let create touch the dataset root, if not saving in parent dataset. PR #7036 (by @mih)

  • Let get_status_dict() use exception message if none is passed. PR #7037 (by @mih)

  • Make choices for status|diff --annex and status|diff --untracked visible. PR #7039 (by @mih)

  • push: Assume 0 bytes pushed if git-annex does not provide bytesize. PR #7049 (by @yarikoptic)



  • Allow for any 2 from first 3 to be consumed in test_gracefull_death. PR #7041 (by @yarikoptic)

0.17.5 (Fri Sep 02 2022)

Bug Fix

Authors: 3

0.17.4 (Tue Aug 30 2022)

Bug Fix

  • BF: make logic more consistent for files=[] argument (which is False but not None) #6976 (@yarikoptic)

  • Run pytests in parallel (-n 2) on appveyor #6987 (@yarikoptic)

  • Add workflow for autogenerating changelog snippets #6981 (@jwodder)

  • Provide /dev/null (b:\nul on Windows) instead of empty string as a git-repo to avoid reading local repo configuration #6986 (@yarikoptic)

  • RF: call_from_parser - move code into “else” to simplify reading etc #6982 (@yarikoptic)

  • BF: if early attempt to parse resulted in error, setup subparsers #6980 (@yarikoptic)

  • Run pytests in parallel (-n 2) on Travis #6915 (@yarikoptic)

  • Send one character (no newline) to stdout in protocol test to guarantee a single “message” and thus a single custom value #6978 (@christian-monch)


Authors: 3

0.17.3 (Tue Aug 23 2022)

Bug Fix

  • BF: git_ignore_check do not overload possible value of stdout/err if present #6937 (@yarikoptic)

  • DOCfix: fix docstring GeneratorStdOutErrCapture to say that treats both stdout and stderr identically #6930 (@yarikoptic)

  • Explain purpose of create-sibling-ria’s –post-update-hook #6958 (@mih)

  • ENH+BF: get_parent_paths - make / into sep option and consistently use “/” as path separator #6963 (@yarikoptic)

  • BF(TEMP): use git-annex from neurodebian -devel to gain fix for bug detected with datalad-crawler #6965 (@yarikoptic)

  • BF(TST): make tests use path helper for Windows “friendliness” of the tests #6955 (@yarikoptic)

  • BF(TST): prevent auto-upgrade of “remote” test sibling, do not use local path for URL #6957 (@yarikoptic)

  • Forbid drop operation from symlink’ed annex (e.g. due to being cloned with –reckless=ephemeral) to prevent data-loss #6959 (@mih)

  • Acknowledge git-config comment chars #6944 (@mih @yarikoptic)

  • Minor tuneups to please updated codespell #6956 (@yarikoptic)

  • TST: Add a testcase for #6950 #6957 (@adswa)

  • BF+ENH(TST): fix typo in code of wtf filesystems reports #6920 (@yarikoptic)

  • DOC: Datalad -> DataLad #6937 (@aqw)

  • BF: fix typo which prevented silently to not show details of filesystems #6930 (@yarikoptic)

  • BF(TST): allow for a annex repo version to upgrade if running in adjusted branches #6927 (@yarikoptic)

  • RF extensions github action to centralize configuration for extensions etc, use pytest for crawler #6914 (@yarikoptic)

  • BF: travis - mark our directory as safe to interact with as root #6919 (@yarikoptic)

  • BF: do not pretend we know what repo version git-annex would upgrade to #6902 (@yarikoptic)

  • BF(TST): do not expect log message for guessing Path to be possibly a URL on windows #6911 (@yarikoptic)

  • ENH(TST): Disable coverage reporting on travis while running pytest #6898 (@yarikoptic)

  • RF: just rename internal variable from unclear “op” to “io” #6907 (@yarikoptic)

  • DX: Demote loglevel of message on url parameters to DEBUG while guessing RI #6891 (@adswa @yarikoptic)

  • Fix and expand datalad.runner type annotations #6893 (@christian-monch @yarikoptic)

  • Use pytest to test datalad-metalad in test_extensions-workflow #6892 (@christian-monch)

  • Let push honor multiple publication dependencies declared via siblings #6869 (@mih @yarikoptic)

  • ENH: upgrade versioneer from versioneer-0.20.dev0 to versioneer-0.23.dev0 #6888 (@yarikoptic)

  • ENH: introduce typing checking and GitHub workflow #6885 (@yarikoptic)

  • RF,ENH(TST): future proof testing of git annex version upgrade + test annex init on all supported versions #6880 (@yarikoptic)

  • ENH(TST): test against supported git annex repo version 10 + make it a full sweep over tests #6881 (@yarikoptic)

  • BF: RF f-string uses in logger to %-interpolations #6886 (@yarikoptic)

  • Merge branch ‘bf-sphinx-5.1.0’ into maint #6883 (@yarikoptic)

  • BF(DOC): workaround for #10701 of sphinx in 5.1.0 #6883 (@yarikoptic)

  • Clarify confusing INFO log message from get() on dataset installation #6871 (@mih)

  • Protect again failing to load a command interface from an extension #6879 (@mih)

  • Support unsetting config via datalad -c :<name> #6864 (@mih)

  • Fix DOC string typo in the path within AnnexRepo.annexstatus, and replace with proper sphinx reference #6858 (@christian-monch)

  • Improved support for saving typechanges #6793 (@mih)

Pushed to maint

  • BF: Remove duplicate ds key from result record (@adswa)

  • DOC: fix capitalization of service names (@aqw)


  • BF(TST,workaround): just xfail failing archives test on NFS #6912 (@yarikoptic)

Authors: 5

0.17.2 (Sat Jul 16 2022)

Bug Fix

  • BF(TST): do proceed to proper test for error being caught for recent git-annex on windows with symlinks #6850 (@yarikoptic)

  • Addressing problem testing against python 3.10 on Travis (skip more annex versions) #6842 (@yarikoptic)

  • XFAIL test_runner_parametrized_protocol on python3.8 when getting duplicate output #6837 (@yarikoptic)

  • BF: Make create’s check for procedures work with several again #6841 (@adswa)

  • Support older pytests #6836 (@jwodder)

Authors: 3

0.17.1 (Mon Jul 11 2022)

Bug Fix

  • DOC: minor fix - consistent DataLad (not Datalad) in docs and CHANGELOG #6830 (@yarikoptic)

  • DOC: fixup/harmonize Changelog for 0.17.0 a little #6828 (@yarikoptic)

  • BF: use –python-match minor option in new datalad-installer release to match outside version of Python #6827 (@christian-monch @yarikoptic)

  • Do not quote paths for ssh >= 9 #6826 (@christian-monch @yarikoptic)

  • Suppress DeprecationWarning to allow for distutils to be used #6819 (@yarikoptic)

  • RM(TST): remove testing of datalad.test which was removed from 0.17.0 #6822 (@yarikoptic)

  • Avoid import of nose-based tests.utils, make skip_if_no_module() and skip_if_no_network() allowed at module level #6817 (@jwodder)

  • BF(TST): use higher level instead of asyncio.get_event_loop in test_inside_async #6808 (@yarikoptic)

Authors: 3

0.17.0 (Thu Jul 7 2022) – pytest migration

Enhancements and new features

  • “log” progress bar now reports about starting a specific action as well. #6756 (by @yarikoptic)

  • Documentation and behavior of traceback reporting for log messages via DATALAD_LOG_TRACEBACK was improved to yield a more compact report. The documentation for this feature has been clarified. #6746 (by @mih)

  • datalad unlock gained a progress bar. #6704 (by @adswa)

  • When create-sibling-gitlab is called on non-existing subdatasets or paths it now returns an impossible result instead of no feedback at all. #6701 (by @adswa)

  • datalad wtf includes a report on file system types of commonly used paths. #6664 (by @adswa)

  • Use next generation metadata code in search, if it is available. #6518 (by @christian-monch)

Deprecations and removals

  • Remove unused and untested log helpers NoProgressLog and OnlyProgressLog. #6747 (by @mih)

  • Remove unused sorted_files() helper. #6722 (by @adswa)

  • Discontinued the value stdout for use with the config variable as its use would inevitably break special remote implementations. #6675 (by @bpoldrack)

  • AnnexRepo.add_urls() is deprecated in favor of AnnexRepo.add_url_to_file() or a direct call to AnnexRepo.call_annex(). #6667 (by @mih)

  • datalad test command and supporting functionality (e.g., datalad.test) were removed. #6273 (by @jwodder)

Bug Fixes

  • export-archive does not rely on normalize_path() methods anymore and became more robust when called from subdirectories. #6745 (by @adswa)

  • Sanitize keys before checking content availability to ensure that the content availability of files with URL- or custom backend keys is correctly determined and marked. #6663 (by @adswa)

  • Ensure saving a new subdataset to a superdataset yields a valid .gitmodules record regardless of whether and how a path constraint is given to the save() call. Fixes #6547 #6790 (by @mih)

  • save now repairs annex symlinks broken by a git-mv operation prior recording a new dataset state. Fixes #4967 #6795 (by @mih)


  • API documentation for log helpers, like log_progress() is now included in the renderer documentation. #6746 (by @mih)

  • New design document on progress reporting. #6734 (by @mih)

  • Explain downstream consequences of using --fast option in addurls. #6684 (by @jdkent)


  • Inline code of create-sibling-ria has been refactored to an internal helper to check for siblings with particular names across dataset hierarchies in datalad-next, and is reintroduced into core to modularize the code base further. #6706 (by @adswa)

  • get_initialized_logger now lets a given logtarget take precedence over #6675 (by @bpoldrack)

  • Many uses of deprecated call options were replaced with the recommended ones. #6273 (by @jwodder)

  • Get rid of asyncio import by defining few noops methods from asyncio.protocols.SubprocessProtocol directly in WitlessProtocol. #6648 (by @yarikoptic)

  • Consolidate GitRepo.remove() and AnnexRepo.remove() into a single implementation. #6783 (by @mih) ## Tests

  • Discontinue use of with_testrepos decorator other than for the deprecation cycle for nose. #6690 (by @mih @bpoldrack) See #6144 for full list of changes.

  • Remove usage of deprecated AnnexRepo.add_urls in tests. #6683 (by @bpoldrack)

  • Minimalistic (adapters, no assert changes, etc) migration from nose to pytest. Support functionality possibly used by extensions and relying on nose helpers is left in place to avoid affecting their run time and defer migration of their test setups.. #6273 (by @jwodder)

Authors: 7

  • Yaroslav Halchenko (@yarikoptic)

  • Michael Hanke (@mih)

  • Benjamin Poldrack (@bpoldrack)

  • Adina Wagner (@adswa)

  • John T. Wodder (@jwodder)

  • Christian Mnch (@christian-monch)

  • James Kent (@jdkent)

0.16.7 (Wed Jul 06 2022)

Bug Fix

Pushed to maint

  • Make sure a subdataset is saved with a complete .gitmodules record (@mih)

Authors: 5

0.16.6 (Tue Jun 14 2022)

Bug Fix

Authors: 2

0.16.5 (Wed Jun 08 2022)

Bug Fix

  • BF: push to github - remove datalad-push-default-first config only in non-dry run to ensure we push default branch separately in next step #6750 (@yarikoptic)

  • In addition to default (system) ssh version, report configured ssh; fix ssh version parsing on Windows #6729 (@yarikoptic)

Authors: 1

0.16.4 (Thu Jun 02 2022)

Bug Fix

  • BF(TST): RO operations - add test directory into git #6726 (@yarikoptic)

  • DOC: fixup of docstring for skip_ssh #6727 (@yarikoptic)

  • DOC: Set language in Sphinx config to en #6727 (@adswa)

  • BF: Catch KeyErrors from unavailable WTF infos #6712 (@adswa)

  • Add annex.private to ephemeral clones. That would make git-annex not assign shared (in git-annex branch) annex uuid. #6702 (@bpoldrack @adswa)

  • BF: require argcomplete version at least 1.12.3 to test/operate correctly #6693 (@yarikoptic)

  • Replace Zenodo DOI with JOSS for due credit #6725 (@adswa)

Authors: 3

0.16.3 (Thu May 12 2022)

Bug Fix

  • No change for a PR to trigger release #6692 (@yarikoptic)

  • Sanitize keys before checking content availability to ensure correct value for keys with URL or custom backend #6665 (@adswa @yarikoptic)

  • Change a key-value pair in drop result record #6625 (@mslw)

  • Link docs of datalad-next #6677 (@mih)

  • Fix GitRepo.get_branch_commits_() to handle branch names conflicts with paths #6661 (@mih)

  • OPT: AnnexJsonProtocol - avoid dragging possibly long data around #6660 (@yarikoptic)

  • Remove two too prominent create() INFO log message that duplicate DEBUG log and harmonize some other log messages #6638 (@mih @yarikoptic)

  • Remove unsupported parameter create_sibling_ria(existing=None) #6637 (@mih)

  • Add released plugin to .autorc to annotate PRs on when released #6639 (@yarikoptic)

Authors: 4

0.16.2 (Thu Apr 21 2022)

Bug Fix

  • Demote (to level 1 from DEBUG) and speed-up API doc logging (parseParameters) #6635 (@mih)

  • Factor out actual data transfer in push #6618 (@christian-monch)

  • ENH: include version of datalad in tests teardown Versions: report #6628 (@yarikoptic)

  • MNT: Require importlib-metadata >=3.6 for Python < 3.10 for entry_points taking kwargs #6631 (@effigies)

  • Factor out credential handling of create-sibling-ghlike #6627 (@mih)

  • BF: Fix wrong key name of annex’ JSON records #6624 (@bpoldrack)

Pushed to maint

Authors: 5

0.16.1 (Fr Apr 8 2022) – April Fools’ Release

  • Fixes forgotten changelog in docs

0.16.0 (Fr Apr 8 2022) – Spring cleaning!

Enhancements and new features

  • A new set of create-sibling-* commands reimplements the GitHub-platform support of create-sibling-github and adds support to interface three new platforms in a unified fashion: GIN (create-sibling-gin), GOGS (create-sibling-gogs), and Gitea (create-sibling-gitea). All commands rely on personal access tokens only for authentication, allow for specifying one of several stored credentials via a uniform --credential parameter, and support a uniform --dry-run mode for testing without network. #5949 (by @mih)

  • create-sibling-github now has supports direct specification of organization repositories via a [<org>/]reposyntax #5949 (by @mih)

  • create-sibling-gitlab gained a --dry-run parameter to match the corresponding parameters in create-sibling-{github,gin,gogs,gitea} #6013 (by @adswa)

  • The --new-store-ok parameter of create-sibling-ria only creates new RIA stores when explicitly provided #6045 (by @adswa)

  • The default performance of status() and diff() commands is improved by up to 700% removing file-type evaluation as a default operation, and simplifying the type reporting rule #6097 (by @mih)

  • drop() and remove() were reimplemented in full, conceptualized as the antagonist commands to get() and clone(). A new, harmonized set of parameters (--what ['filecontent', 'allkeys', 'datasets', 'all'], --reckless ['modification', 'availability', 'undead', 'kill']) simplifies their API. Both commands include additional safeguards. uninstall is replaced with a thin shim command around drop() #6111 (by @mih)

  • add_archive_content() was refactored into a dataset method and gained progress bars #6105 (by @adswa)

  • The datalad and datalad-archives special remotes have been reimplemented based on AnnexRemote #6165 (by @mih)

  • The result_renderer() semantics were decomplexified and harmonized. The previous default result renderer was renamed to generic. #6174 (by @mih)

  • get_status_dict learned to include exit codes in the case of CommandErrors #5642 (by @yarikoptic)

  • datalad clone can now pass options to git-clone, adding support for cloning specific tags or branches, naming siblings other names than origin, and exposing git clone’s optimization arguments #6218 (by @kyleam and @mih)

  • Inactive BatchedCommands are cleaned up #6206 (by @jwodder)

  • export-archive-ora learned to filter files exported to 7z archives #6234 (by @mih and @bpinsard)

  • datalad run learned to glob recursively #6262 (by @AKSoo)

  • The ORA remote learned to recover from interrupted uploads #6267 (by @mih)

  • A new threaded runner with support for timeouts and generator-based subprocess communication is introduced and used in BatchedCommand and AnnexRepo #6244 (by @christian-monch)

  • A new switch allows to enable librarymode and queries for the effective API in use #6213 (by @mih)

  • run and rerun now support parallel jobs via --jobs #6279 (by @AKSoo)

  • A new foreach-dataset plumbing command allows to run commands on each (sub)dataset, similar to git submodule foreach #5517 (by @yarikoptic)

  • The dataset parameter is not restricted to only locally resolvable file-URLs anymore #6276 (by @christian-monch)

  • DataLad’s credential system is now able to query git-credential by specifying credential type git in the respective provider configuration #5796 (by @bpoldrack)

  • DataLad now comes with a git credential helper git-credential-datalad allowing Git to query DataLad’s credential system #5796 (by @bpoldrack and @mih)

  • The new runner now allows for multiple threads #6371 (by @christian-monch)

  • A new configurationcommand provides an interface to manipulate and query the DataLad configuration. #6306 (by @mih)

    • Unlike the global Python-only datalad.cfg or dataset-specific Dataset.config configuration managers, this command offers a uniform API across the Python and the command line interfaces.

    • This command was previously available in the mihextras extension as x-configuration, and has been merged into the core package in an improved version. #5489 (by @mih)

    • In its default dump mode, the command provides an annotated list of the effective configuration after considering all configuration sources, including hints on additional configuration settings and their supported values.

  • The command line interface help-reporting has been sped up by ~20% #6370 #6378 (by @mih)

  • ConfigManager now supports reading committed dataset configuration in bare repositories. Analog to reading .datalad/config from a worktree, blob:HEAD:.datalad/config is read (e.g., the config committed in the default branch). The support includes `reload() change detection using the gitsha of this file. The behavior for non-bare repositories is unchanged. #6332 (by @mih)

  • The CLI help generation has been sped up, and now also supports the completion of parameter values for a fixed set of choices #6415 (by @mih)

  • Individual command implementations can now declare a specific “on-failure” behavior by defining Interface.on_failure to be one of the supported modes (stop, continue, ignore). Previously, such a modification was only possible on a per-call basis. #6430 (by @mih)

  • The run command changed its default “on-failure” behavior from continue to stop. This change prevents the execution of a command in case a declared input can not be obtained. Previously, only an error result was yielded (and run eventually yielded a non-zero exit code or an IncompleteResultsException), but the execution proceeded and potentially saved a dataset modification despite incomplete inputs, in case the command succeeded. This previous default behavior can still be achieved by calling run with the equivalent of --on-failure continue #6430 (by @mih)

  • The `run command now provides readily executable, API-specific instructions how to save the results of a command execution that failed expectedly #6434 (by @mih)

  • create-sibling --since=^ mode will now be as fast as push --since=^ to figure out for which subdatasets to create siblings #6436 (by @yarikoptic)

  • When file names contain illegal characters or reserved file names that are incompatible with Windows systems a configurable check for save ( will either do nothing (none), emit an incompatibility warning (warning, default), or cause save to error (error) #6291 (by @adswa)

  • Improve responsiveness of datalad drop in datasets with a large annex. #6580 (by @christian-monch)

  • save code might operate faster on heavy file trees #6581 (by @yarikoptic)

  • Removed a per-file overhead cost for ORA when downloading over HTTP #6609 (by @bpoldrack)

  • A new module offers the utility functions register_config() and has_config() that allow extension developers to announce additional configuration items to the central configuration management. #6601 (by @mih)

  • When operating in a dirty dataset, export-to-figshare now yields and impossible result instead of raising a RunTimeError #6543 (by @adswa)

  • Loading DataLad extension packages has been sped-up leading to between 2x and 4x faster run times for loading individual extensions and reporting help output across all installed extensions. #6591 (by @mih)

  • Introduces the configuration key datalad.ssh.executable. This key allows specifying an ssh-client executable that should be used by datalad to establish ssh-connections. The default value is ssh unless on a Windows system where $WINDIR\System32\OpenSSH\ssh.exe exists. In this case, the value defaults to $WINDIR\System32\OpenSSH\ssh.exe. #6553 (by @christian-monch)

  • create-sibling should perform much faster in case of --since specification since would consider only submodules related to the changes since that point. #6528 (by @yarikoptic)

  • A new configuration setting datalad.ssh.try-use-annex-bundled-git=yes|no can be used to influence the default remote git-annex bundle sensing for SSH connections. This was previously done unconditionally for any call to datalad sshrun (which is also used for any SSH-related Git or git-annex functionality triggered by DataLad-internal processing) and could incur a substantial per-call runtime cost. The new default is to not perform this sensing, because for, e.g., use as GIT_SSH_COMMAND there is no expectation to have a remote git-annex installation, and even with an existing git-annex/Git bundle on the remote, it is not certain that the bundled Git version is to be preferred over any other Git installation in a user’s PATH. #6533 (by @mih)

  • run now yields a result record immediately after executing a command. This allows callers to use the standard --on-failure switch to control whether dataset modifications will be saved for a command that exited with an error. #6447 (by @mih)

Deprecations and removals

  • The --pbs-runner commandline option (deprecated in 0.15.0) was removed #5981 (by @mih)

  • The dependency to PyGithub was dropped #5949 (by @mih)

  • create-sibling-github’s credential handling was trimmed down to only allow personal access tokens, because GitHub discontinued user/password based authentication #5949 (by @mih)

  • create-sibling-gitlab’s --dryrun parameter is deprecated in favor or --dry-run #6013 (by @adswa)

  • Internal obsolete Gitrepo.*_submodule methods were moved to datalad-deprecated #6010 (by @mih)

  • datalad/support/ is unused in DataLad core and removed #6115 (by @yarikoptic)

  • Support for the undocumented datalad.api.result-renderer config setting has been dropped #6174 (by @mih)

  • Undocumented use of result_renderer=None is replaced with result_renderer='disabled' #6174 (by @mih)

  • remove’s --recursive argument has been deprecated #6257 (by @mih)

  • The use of the internal helper get_repo_instance() is discontinued and deprecated #6268 (by @mih)

  • Support for Python 3.6 has been dropped (#6286 (by @christian-monch) and #6364 (by @yarikoptic))

  • All but one Singularity recipe flavor have been removed due to their limited value with the end of life of Singularity Hub #6303 (by @mih)

  • All code in module datalad.cmdline was (re)moved, only datalad.cmdline.helpers.get_repo_instanceis kept for a deprecation period (by @mih)

  • datalad.interface.common_opts.eval_default has been deprecated. All (command-specific) defaults for common interface parameters can be read from Interface class attributes (#6391 (by @mih)

  • Remove unused and untested datalad.interface.utils helpers cls2cmdlinename and path_is_under #6392 (by @mih)

  • An unused code path for result rendering was removed from the CLI main() #6394 (by @mih)

  • create-sibling will require now "^" instead of an empty string for since option #6436 (by @yarikoptic)

  • run no longer raises a CommandError exception for failed commands, but yields an error result that includes a superset of the information provided by the exception. This change impacts command line usage insofar as the exit code of the underlying command is no longer relayed as the exit code of the run command call – although run continues to exit with a non-zero exit code in case of an error. For Python API users, the nature of the raised exception changes from CommandError to IncompleteResultsError, and the exception handling is now configurable using the standard on_failure command argument. The original CommandError exception remains available via the exception property of the newly introduced result record for the command execution, and this result record is available via IncompleteResultsError.failed, if such an exception is raised. #6447 (by @mih)

  • Custom cast helpers were removed from datalad core and migrated to a standalone repository #6516 (by @adswa)

  • The bundled parameter of get_connection_hash() is now ignored and will be removed with a future release. #6532 (by @mih)

  • BaseDownloader.fetch() is logging download attempts on DEBUG (previously INFO) level to avoid polluting output of higher-level commands. #6564 (by @mih)

Bug Fixes

  • create-sibling-gitlab erroneously overwrote existing sibling configurations. A safeguard will now prevent overwriting and exit with an error result #6015 (by @adswa)

  • create-sibling-gogs now relays HTTP500 errors, such as “no space left on device” #6019 (by @mih)

  • annotate_paths() is removed from the last parts of code base that still contained it #6128 (by @mih)

  • add_archive_content() doesn’t crash with --key and --use-current-dir anymore #6105 (by @adswa)

  • run-procedure now returns an error result when a non-existent procedure name is specified #6143 (by @mslw)

  • A fix for a silent failure of download-url --archive when extracting the archive #6172 (by @adswa)

  • Uninitialized AnnexRepos can now be dropped #6183 (by @mih)

  • Instead of raising an error, the formatters tests are skipped when the formatters module is not found #6212 (by @adswa)

  • create-sibling-gin does not disable git-annex availability on Gin remotes anymore #6230 (by @mih)

  • The ORA special remote messaging is fixed to not break the special remote protocol anymore and to better relay messages from exceptions to communicate underlying causes #6242 (by @mih)

  • A keyring.delete() call was fixed to not call an uninitialized private attribute anymore #6253 (by @bpoldrack)

  • An erroneous placement of result keyword arguments into a format() method instead of get_status_dict() of create-sibling-ria has been fixed #6256 (by @adswa)

  • status, run-procedure, and metadata are no longer swallowing result-related messages in renderers #6280 (by @mih)

  • uninstall now recommends the new --reckless parameter instead of the deprecated --nocheck parameter when reporting hints #6277 (by @adswa)

  • download-url learned to handle Pathobjects #6317 (by @adswa)

  • Restore default result rendering behavior broken by Key interface documentation #6394 (by @mih)

  • Fix a broken check for file presence in the ConfigManager that could have caused a crash in rare cases when a config file is removed during the process runtime #6332 (by @mih) `- ConfigManager.get_from_source() now accesses the correct information when using the documented source='local', avoiding a crash #6332 (by @mih)

  • run no longer let’s the internal call to save render its results unconditionally, but the parameterization f run determines the effective rendering format. #6421 (by @mih)

  • Remove an unnecessary and misleading warning from the runner #6425 (by @christian-monch)

  • A number of commands stopped to double-report results #6446 (by @adswa)

  • create-sibling-ria no longer creates an annex/objects directory in-store, when called with --no-storage-sibling. #6495 (by @bpoldrack )

  • Improve error message when an invalid URL is given to clone. #6500 (by @mih)

  • DataLad declares a minimum version dependency to keyring >= 20.0 to ensure that token-based authentication can be used. #6515 (by @adswa)

  • ORA special remote tries to obtain permissions when dropping a key from a RIA store rather than just failing. Thus having the same permissions in the store’s object trees as one directly managed by git-annex would have, works just fine now. #6493 (by @bpoldrack )

  • require_dataset() now uniformly raises NoDatasetFound when no dataset was found. Implementations that catch the previously documented InsufficientArgumentsError or the actually raised ValueError will continue to work, because NoDatasetFound is derived from both types. #6521 (by @mih)

  • Keyboard-interactive authentication is now possibly with non-multiplexed SSH connections (i.e., when no connection sharing is possible, due to lack of socket support, for example on Windows). Previously, it was disabled forcefully by DataLad for no valid reason. #6537 (by @mih)

  • Remove duplicate exception type in reporting of top-level CLI exception handler. #6563 (by @mih)

  • Fixes DataLad’s parsing of git-annex’ reporting on unknown paths depending on its version and the value of the annex.skipunknown config. #6550 (by @bpoldrack)

  • Fix ORA special remote not properly reporting on HTTP failures. #6535 (by @bpoldrack)

  • ORA special remote didn’t show per-file progress bars when downloading over HTTP #6609 (by @bpoldrack)

  • save now can commit the change where file becomes a directory with a staged for commit file. #6581 (by @yarikoptic)

  • create-sibling will no longer create siblings for not yet saved new subdatasets, and will now create sub-datasets nested in the subdatasets which did not yet have those siblings. #6603 (by @yarikoptic)


  • A new design document sheds light on result records #6167 (by @mih)

  • The disabled result renderer mode is documented #6174 (by @mih)

  • A new design document sheds light on the datalad and datalad-archives special remotes #6181 (by @mih)

  • A new design document sheds light on BatchedCommand and BatchedAnnex #6203 (by @christian-monch)

  • A new design document sheds light on standard parameters #6214 (by @adswa)

  • The DataLad project adopted the Contributor Covenant COC v2.1 #6236 (by @adswa)

  • Docstrings learned to include Sphinx’ “version added” and “deprecated” directives #6249 (by @mih)

  • A design document sheds light on basic docstring handling and formatting #6249 (by @mih)

  • A new design document sheds light on position versus keyword parameter usage #6261 (by @yarikoptic)

  • create-sibling-gin’s examples have been improved to suggest push as an additional step to ensure proper configuration #6289 (by @mslw)

  • A new document describes the credential system from a user’s perspective #5796 (by @bpoldrack)

  • Enhance the design document on DataLad’s credential system #5796 (by @bpoldrack)

  • The documentation of the configuration command now details all locations DataLad is reading configuration items from, and their respective rules of precedence #6306 (by @mih)

  • API docs for datalad.interface.base are now included in the documentation #6378 (by @mih)

  • A new design document is provided that describes the basics of the command line interface implementation #6382 (by @mih)

  • The `datalad.interface.base.Interface class, the basis of all DataLad command implementations, has been extensively documented to provide an overview of basic principles and customization possibilities #6391 (by @mih)

  • --since=^ mode of operation of create-sibling is documented now #6436 (by @yarikoptic)


  • The internal status() helper was equipped with docstrings and promotes “breadth-first” reporting with a new parameter reporting_order #6006 (by @mih)

  • AnnexRepo.get_file_annexinfo() is introduced for more convenient queries for single files and replaces a now deprecated AnnexRepo.get_file_key() to receive information with fewer calls to Git #6104 (by @mih)

  • A new get_paths_by_ds() helper exposes status’ path normalization and sorting #6110 (by @mih)

  • status is optimized with a cache for dataset roots #6137 (by @yarikoptic)

  • The internal get_func_args_doc() helper with Python 2 is removed from DataLad core #6175 (by @yarikoptic)

  • Further restructuring of the source tree to better reflect the internal dependency structure of the code: AddArchiveContent is moved from datalad/interface to datalad/local (#6188 (by @mih)), Clean is moved from datalad/interface to datalad/local (#6191 (by @mih)), Unlock is moved from datalad/interface to datalad/local (#6192 (by @mih)), DownloadURL is moved from datalad/interface to datalad/local (#6217 (by @mih)), Rerun is moved from datalad/interface to datalad/local (#6220 (by @mih)), RunProcedure is moved from datalad/interface to datalad/local (#6222 (by @mih)). The interface command list is restructured and resorted #6223 (by @mih)

  • wrapt is replaced with functools’ wraps #6190 (by @yariktopic)

  • The unmaintained appdirs library has been replaced with platformdirs #6198 (by @adswa)

  • Modelines mismatching the code style in source files were fixed #6263 (by @AKSoo)

  • datalad/ has been cleaned up #6271 (by @mih)

  • GitRepo.call_git_items is implemented with a generator-based runner #6278 (by @christian-monch)

  • Separate positional from keyword arguments in the Python API to match CLI with * #6176 (by @yarikoptic), #6304 (by @christian-monch)

  • GitRepo.bare does not require the ConfigManager anymore #6323 (by @mih)

  • _get_dot_git() was reimplemented to be more efficient and consistent, by testing for common scenarios first and introducing a consistently applied resolved flag for result path reporting #6325 (by @mih)

  • All data files under datalad are now included when installing DataLad #6336 (by @jwodder)

  • Add internal method for non-interactive provider/credential storing #5796 (by @bpoldrack)

  • Allow credential classes to have a context set, consisting of a URL they are to be used with and a dataset DataLad is operating on, allowing to consider “local” and “dataset” config locations #5796 (by @bpoldrack)

  • The Interface method get_refds_path() was deprecated #6387 (by @adswa)

  • datalad.interface.base.Interface is now an abstract class #6391 (by @mih)

  • Simplified the decision making for result rendering, and reduced code complexity #6394 (by @mih)

  • Reduce code duplication in #6398 (by @mih)

  • Use public ArgumentParser.parse_known_args instead of protected _parse_known_args #6414 (by @yarikoptic)

  • add-archive-content does not rely on the deprecated tempfile.mktemp anymore, but uses the more secure tempfile.mkdtemp #6428 (by @adswa)

  • AnnexRepo’s internal annexstatus is deprecated. In its place, a new test helper assists the few tests that rely on it #6413 (by @adswa)

  • config has been refactored from where[="dataset"] to scope[="branch"] #5969 (by @yarikoptic)

  • Common command arguments are now uniformly and exhaustively passed to result renderers and filters for decision making. Previously, the presence of a particular argument depended on the respective API and circumstances of a command call. #6440 (by @mih)

  • Entrypoint processing for extensions and metadata extractors has been consolidated on a uniform helper that is about twice as fast as the previous implementations. #6591 (by @mih)


  • A range of Windows tests pass and were enabled #6136 (by @adswa)

  • Invalid escape sequences in some tests were fixed #6147 (by @mih)

  • A cross-platform compatible HTTP-serving test environment is introduced #6153 (by @mih)

  • A new helper exposes serve_path_via_http to the command line to deploy an ad-hoc instance of the HTTP server used for internal testing, with SSL and auth, if desired. #6169 (by @mih)

  • Windows tests were redistributed across worker runs to harmonize runtime #6200 (by @adswa)

  • Batchedcommand gained a basic test #6203 (by @christian-monch)

  • The use of with_testrepo is discontinued in all core tests #6224 (by @mih)

  • The new git-annex.filter.annex.process configuration is enabled by default on Windows to speed up the test suite #6245 (by @mih)

  • If the available Git version supports it, the test suite now uses GIT_CONFIG_GLOBAL to configure a fake home directory instead of overwriting HOME on OSX (#6251 (by @bpoldrack)) and HOME and USERPROFILE on Windows #6260 (by @adswa)

  • Windows test timeouts of runners were addressed #6311 (by @christian-monch)

  • A handful of Windows tests were fixed (#6352 (by @yarikoptic)) or disabled (#6353 (by @yarikoptic))

  • download-url’s test under http_proxy are skipped when a session can’t be established #6361 (by @yarikoptic)

  • A test for datalad clean was fixed to be invoked within a dataset #6359 (by @yarikoptic)

  • The new datalad.cli.tests have an improved module coverage of 80% #6378 (by @mih)

  • The test_source_candidate_subdataset has been marked as @slow #6429 (by @yarikoptic)

  • Dedicated CLI benchmarks exist now #6381 (by @mih)

  • Enable code coverage report for subprocesses #6546 (by @adswa)

  • Skip a test on annex>=10.20220127 due to a bug in annex. See


  • A new issue template using GitHub forms prestructures bug reports #6048 (by @Remi-Gau)

  • DataLad and its dependency stack were packaged for Gentoo Linux #6088 (by @TheChymera)

  • The readthedocs configuration is modernized to version 2 #6207 (by @adswa)

  • The Windows CI setup now runs on Appveyor’s Visual Studio 2022 configuration #6228 (by @adswa)

  • The readthedocs-theme and Sphinx versions were pinned to re-enable rendering of bullet points in the documentation #6346 (by @adswa)

  • The PR template was updated with a CHANGELOG template. Future PRs should use it to include a summary for the CHANGELOG #6396 (by @mih)

Authors: 11

  • Michael Hanke (@mih)

  • Yaroslav Halchenko (@yarikoptic)

  • Adina Wagner (@adswa)

  • Remi Gau (@Remi-Gau)

  • Horea Christian (@TheChymera)

  • Micha Szczepanik (@mslw)

  • Christian Mnch (@christian-monch)

  • John T. Wodder (@jwodder)

  • Benjamin Poldrack (@bpoldrack)

  • Sin Kim (@AKSoo)

  • Basile Pinsard (@bpinsard)

0.15.6 (Sun Feb 27 2022)

Bug Fix

  • BF: do not use BaseDownloader instance wide InterProcessLock - resolves stalling or errors during parallel installs #6507 (@yarikoptic)

  • release workflow: add -vv to auto invocation (@yarikoptic)

  • Fix version incorrectly incremented by release process in CHANGELOGs #6459 (@yarikoptic)

  • BF(TST): add another condition to skip under http_proxy set #6459 (@yarikoptic)

Authors: 1

0.15.5 (Wed Feb 09 2022)


  • BF: When download-url gets Pathobject as path convert it to a string #6364 (@adswa)

Bug Fix

Authors: 5

0.15.4 (Thu Dec 16 2021)

Bug Fix


  • RF+BF: use skip_if_no_module helper instead of try/except for libxmp and boto #6148 (@yarikoptic)

  • git:// -> #6134 (@mih)

Authors: 6

0.15.3 (Sat Oct 30 2021)

Bug Fix

  • BF: Don’t make create-sibling recursive by default #6116 (@adswa)

  • BF: Add dashes to ‘force’ option in non-empty directory error message #6078 (@DisasterMo)

  • DOC: Add supported URL types to download-url’s docstring #6098 (@adswa)

  • BF: Retain git-annex error messages & don’t show them if operation successful #6070 (@DisasterMo)

  • Remove uses of __full_version__ and datalad.version #6073 (@jwodder)

  • BF: ORA shouldn’t crash while handling a failure #6063 (@bpoldrack)

  • DOC: Refine –reckless docstring on usage and wording #6043 (@adswa)

  • BF: archives upon strip - use rmtree which retries etc instead of rmdir #6064 (@yarikoptic)

  • BF: do not leave test in a tmp dir destined for removal #6059 (@yarikoptic)

  • Next wave of exc_str() removals #6022 (@mih)

Pushed to maint

  • CI: Enable new codecov uploader in Appveyor CI (@adswa)


  • UX: Log clone-candidate number and URLs #6092 (@adswa)

  • UX/ENH: Disable reporting, and don’t do superfluous internal subdatasets calls #6094 (@adswa)

  • Update codecov action to v2 #6072 (@jwodder)


  • Design document on URL substitution feature #6065 (@mih)


Authors: 7

0.15.2 (Wed Oct 06 2021)

Bug Fix

  • BF: Don’t suppress datalad subdatasets output #6035 (@DisasterMo @mih)

  • Honor datalad.runtime.use-patool if set regardless of OS (was Windows only) #6033 (@mih)

  • Discontinue usage of deprecated (public) helper #6032 (@mih)

  • BF: ProgressHandler - close the other handler if was specified #6020 (@yarikoptic)

  • UX: Report GitLab weburl of freshly created projects in the result #6017 (@adswa)

  • Ensure there’s a blank line between the class __doc__ and “Parameters” in build_doc docstrings #6004 (@jwodder)

  • Large code-reorganization of everything runner-related #6008 (@mih)

  • Discontinue exc_str() in all modern parts of the code base #6007 (@mih)


  • TST: Add test to ensure functionality with subdatasets starting with a hyphen (-) #6042 (@DisasterMo)

  • BF(TST): filter away warning from coverage from analysis of stderr of –help #6028 (@yarikoptic)

  • BF: disable outdated SSL root certificate breaking chain on older/buggy clients #6027 (@yarikoptic)

  • BF: start global test_http_server only if not running already #6023 (@yarikoptic)

Authors: 5

0.15.1 (Fri Sep 24 2021)

Bug Fix

  • BF: downloader - fail to download even on non-crippled FS if symlink exists #5991 (@yarikoptic)

  • ENH: import datalad.api to bind extensions methods for discovery of dataset methods #5999 (@yarikoptic)

  • Restructure cmdline API presentation #5988 (@mih)

  • Close file descriptors after process exit #5983 (@mih)

Pushed to maint

  • Discontinue testing of hirni extension (@mih)



  • Coarse description of the credential subsystem’s functionality #5998 (@mih)


  • BF(TST): use sys.executable, mark test_ria_basics.test_url_keys as requiring network #5986 (@yarikoptic)

Authors: 3

0.15.0 (Tue Sep 14 2021) – We miss you Kyle!

Enhancements and new features

  • Command execution is now performed by a new Runner implementation that is no longer based on the asyncio framework, which was found to exhibit fragile performance in interaction with other asyncio-using code, such as Jupyter notebooks. The new implementation is based on threads. It also supports the specification of “protocols” that were introduced with the switch to the asyncio implementation in 0.14.0. (#5667)

  • clone now supports arbitrary URL transformations based on regular expressions. One or more transformation steps can be defined via datalad.clone.url-substitute.<label> configuration settings. The feature can be (and is now) used to support convenience mappings, such as (displayed in a browser window) to osf://q8xnk (clonable via the datalad-osf extension. (#5749)

  • Homogenize SSH use and configurability between DataLad and git-annex, by instructing git-annex to use DataLad’s sshrun for SSH calls (instead of SSH directly). (#5389)

  • The ORA special remote has received several new features:

    • It now support a push-url setting as an alternative to url for write access. An analog parameter was also added to create-sibling-ria. (#5420, #5428)

    • Access of RIA stores now performs homogeneous availability checks, regardless of access protocol. Before, broken HTTP-based access due to misspecified URLs could have gone unnoticed. (#5459, #5672)

    • Error reporting was introduce to inform about undesirable conditions in remote RIA stores. (#5683)

  • create-sibling-ria now supports --alias for the specification of a convenience dataset alias name in a RIA store. (#5592)

  • Analog to git commit, save now features an --amend mode to support incremental updates of a dataset state. (#5430)

  • run now supports a dry-run mode that can be used to inspect the result of parameter expansion on the effective command to ease the composition of more complicated command lines. (#5539)

  • run now supports a --assume-ready switch to avoid the (possibly expensive) preparation of inputs and outputs with large datasets that have already been readied through other means. (#5431)

  • update now features --how and --how-subds parameters to configure how an update shall be performed. Supported modes are fetch (unchanged default), and merge (previously also possible via --merge), but also new strategies like reset or checkout. (#5534)

  • update has a new --follow=parentds-lazy mode that only performs a fetch operation in subdatasets when the desired commit is not yet present. During recursive updates involving many subdatasets this can substantially speed up performance. (#5474)

  • DataLad’s command line API can now report the version for individual commands via datalad <cmd> --version. The output has been homogenized to <providing package> <version>. (#5543)

  • create-sibling now logs information on an auto-generated sibling name, in the case that no --name/-s was provided. (#5550)

  • create-sibling-github has been updated to emit result records like any standard DataLad command. Previously it was implemented as a “plugin”, which did not support all standard API parameters. (#5551)

  • copy-file now also works with content-less files in datasets on crippled filesystems (adjusted mode), when a recent enough git-annex (8.20210428 or later) is available. (#5630)

  • addurls can now be instructed how to behave in the event of file name collision via a new parameter --on-collision. (#5675)

  • addurls reporting now informs which particular subdatasets were created. (#5689)

  • Credentials can now be provided or overwritten via all means supported by ConfigManager. Importantly, datalad.credential.<name>.<field> configuration settings and analog specification via environment variables are now supported (rather than custom environment variables only). Previous specification methods are still supported too. (#5680)

  • A new datalad.credentials.force-ask configuration flag can now be used to force re-entry of already known credentials. This simplifies credential updates without having to use an approach native to individual credential stores. (#5777)

  • Suppression of rendering repeated similar results is now configurable via the configuration switches datalad.ui.suppress-similar-results (bool), and datalad.ui.suppress-similar-results-threshold (int). (#5681)

  • The performance of status and similar functionality when determining local file availability has been improved. (#5692)

  • push now renders a result summary on completion. (#5696)

  • A dedicated info log message indicates when dataset repositories are subjected to an annex version upgrade. (#5698)

  • Error reporting improvements:

    • The NoDatasetFound exception now provides information for which purpose a dataset is required. (#5708)

    • Wording of the MissingExternalDependeny error was rephrased to account for cases of non-functional installations. (#5803)

    • push reports when a --to parameter specification was (likely) forgotten. (#5726)

    • Detailed information is now given when DataLad fails to obtain a lock for credential entry in a timely fashion. Previously only a generic debug log message was emitted. (#5884)

    • Clarified error message when create-sibling-gitlab was called without --project. (#5907)

  • add-readme now provides a README template with more information on the nature and use of DataLad datasets. A README file is no longer annex’ed by default, but can be using the new --annex switch. ([#5723][], [#5725][])

  • clean now supports a --dry-run mode to inform about cleanable content. (#5738)

  • A new configuration setting datalad.locations.locks can be used to control the placement of lock files. (#5740)

  • wtf now also reports branch names and states. (#5804)

  • AnnexRepo.whereis() now supports batch mode. (#5533)

Deprecations and removals

  • The minimum supported git-annex version is now 8.20200309. (#5512)

  • ORA special remote configuration items ssh-host, and base-path are deprecated. They are completely replaced by ria+<protocol>:// URL specifications. (#5425)

  • The deprecated no_annex parameter of create() was removed from the Python API. (#5441)

  • The unused GitRepo.pull() method has been removed. (#5558)

  • Residual support for “plugins” (a mechanism used before DataLad supported extensions) was removed. This includes the configuration switches datalad.locations.{system,user}-plugins. (#5554, #5564)

  • Several features and comments have been moved to the datalad-deprecated package. This package must now be installed to be able to use keep using this functionality.

    • The publish command. Use push instead. (#5837)

    • The ls command. (#5569)

    • The web UI that is deployable via datalad create-sibling --ui. (#5555)

    • The “automagic IO” feature. (#5577)

  • AnnexRepo.copy_to() has been deprecated. The push command should be used instead. (#5560)

  • AnnexRepo.sync() has been deprecated. AnnexRepo.call_annex(['sync', ...]) should be used instead. (#5461)

  • All GitRepo.*_submodule() methods have been deprecated and will be removed in a future release. (#5559)

  • create-sibling-github’s --dryrun switch was deprecated, use --dry-run instead. (#5551)

  • The datalad --pbs-runner option has been deprecated, use condor_run (or similar) instead. (#5956)


  • Prevent invalid declaration of a publication dependencies for ‘origin’ on any auto-detected ORA special remotes, when cloing from a RIA store. An ORA remote is now checked whether it actually points to the RIA store the clone was made from. (#5415)

  • The ORA special remote implementation has received several fixes:

    • It can now handle HTTP redirects. (#5792)

    • Prevents failure when URL-type annex keys contain the ‘/’ character. (#5823)

    • Properly support the specification of usernames, passwords and ports in ria+<protocol>:// URLs. (#5902)

  • It is now possible to specifically select the default (or generic) result renderer via datalad -f default and with that override a tailored result renderer that may be preconfigured for a particular command. (#5476)

  • Starting with 0.14.0, original URLs given to clone were recorded in a subdataset record. This was initially done in a second commit, leading to inflation of commits and slowdown in superdatasets with many subdatasets. Such subdataset record annotation is now collapsed into a single commits. (#5480)

  • run now longer removes leading empty directories as part of the output preparation. This was surprising behavior for commands that do not ensure on their own that output directories exist. (#5492)

  • A potentially existing message property is no longer removed when using the json or json_pp result renderer to avoid undesired withholding of relevant information. (#5536)

  • subdatasets now reports state=present, rather than state=clean, for installed subdatasets to complement state=absent reports for uninstalled dataset. (#5655)

  • create-sibling-ria now executes commands with a consistent environment setup that matches all other command execution in other DataLad commands. (#5682)

  • save no longer saves unspecified subdatasets when called with an explicit path (list). The fix required a behavior change of GitRepo.get_content_info() in its interpretation of None vs. [] path argument values that now aligns the behavior of GitRepo.diff|status() with their respective documentation. (#5693)

  • get now prefers the location of a subdatasets that is recorded in a superdataset’s .gitmodules record. Previously, DataLad tried to obtain a subdataset from an assumed checkout of the superdataset’s origin. This new default order is (re-)configurable via the datalad.get.subdataset-source-candidate-<priority-label> configuration mechanism. (#5760)

  • create-sibling-gitlab no longer skips the root dataset when . is given as a path. (#5789)

  • siblings now rejects a value given to --as-common-datasrc that clashes with the respective Git remote. (#5805)

  • The usage synopsis reported by siblings now lists all supported actions. (#5913)

  • siblings now renders non-ok results to avoid silent failure. (#5915)

  • .gitattribute file manipulations no longer leave the file without a trailing newline. (#5847)

  • Prevent crash when trying to delete a non-existing keyring credential field. (#5892)

  • git-annex is no longer called with an unconditional annex.retry=3 configuration. Instead, this parameterization is now limited to annex get and annex copy calls. (#5904)


  • file:// URLs are no longer the predominant test case for AnnexRepo functionality. A built-in HTTP server now used in most cases. (#5332)

0.14.8 (Sun Sep 12 2021)

Bug Fix

  • BF: add-archive-content on .xz and other non-.gz stream compressed files #5930 (@yarikoptic)

  • BF(UX): do not keep logging ERROR possibly present in progress records #5936 (@yarikoptic)

  • Annotate datalad_core as not needing actual data – just uses annex whereis #5971 (@yarikoptic)

  • BF: limit CMD_MAX_ARG if obnoxious value is encountered. #5945 (@yarikoptic)

  • Download session/credentials locking – inform user if locking is “failing” to be obtained, fail upon ~5min timeout #5884 (@yarikoptic)

  • Render siblings()’s non-ok results with the default renderer #5915 (@mih)

  • BF: do not crash, just skip whenever trying to delete non existing field in the underlying keyring #5892 (@yarikoptic)

  • Fix argument-spec for siblings and improve usage synopsis #5913 (@mih)

  • Clarify error message re unspecified gitlab project #5907 (@mih)

  • Support username, password and port specification in RIA URLs #5902 (@mih)

  • BF: take path from SSHRI, test URLs not only on Windows #5881 (@yarikoptic)

  • ENH(UX): warn user if keyring returned a “null” keyring #5875 (@yarikoptic)

  • ENH(UX): state original purpose in NoDatasetFound exception + detail it for get #5708 (@yarikoptic)

Pushed to maint

  • Merge branch ‘bf-http-headers-agent’ into maint (@yarikoptic)

  • RF(BF?)+DOC: provide User-Agent to entire session headers + use those if provided (@yarikoptic)


  • Pass --no-changelog to auto shipit if changelog already has entry #5952 (@jwodder)

  • Add isort config to match current convention + run isort via pre-commit (if configured) #5923 (@jwodder)

  • .travis.yml: use python -m {nose,coverage} invocations, and always show combined report #5888 (@yarikoptic)

  • Add project URLs into the package metadata for convenience links on Pypi #5866 (@adswa @yarikoptic)


  • BF: do use OBSCURE_FILENAME instead of hardcoded unicode #5944 (@yarikoptic)

  • BF(TST): Skip testing for having PID listed if no psutil #5920 (@yarikoptic)

  • BF(TST): Boost version of git-annex to 8.20201129 to test an error message #5894 (@yarikoptic)

Authors: 4

0.14.7 (Tue Aug 03 2021)

Bug Fix

  • UX: When two or more clone URL templates are found, error out more gracefully #5839 (@adswa)

  • BF: http_auth - follow redirect (just 1) to re-authenticate after initial attempt #5852 (@yarikoptic)

  • addurls Formatter - provide value repr in exception #5850 (@yarikoptic)

  • ENH: allow for “patch” level semver for “master” branch #5839 (@yarikoptic)

  • BF: Report info from annex JSON error message in CommandError #5809 (@mih)

  • RF(TST): do not test for no EASY and pkg_resources in shims #5817 (@yarikoptic)

  • http downloaders: Provide custom informative User-Agent, do not claim to be “Authenticated access” #5802 (@yarikoptic)

  • ENH(UX,DX): inform user with a warning if version is 0+unknown #5787 (@yarikoptic)

  • shell-completion: add argcomplete to ‘misc’ extra_depends, log an ERROR if argcomplete fails to import #5781 (@yarikoptic)

  • ENH (UX): add python-gitlab dependency #5776 (


  • BF: Fix reported paths in ORA remote #5821 (@adswa)

  • BF: import importlib.metadata not importlib_metadata whenever available #5818 (@yarikoptic)


  • TST: set –allow-unrelated-histories in the mk_push_target setup for Windows #5855 (@adswa)

  • Tests: Allow for version to contain + as a separator and provide more information for version related comparisons #5786 (@yarikoptic)

Authors: 4

0.14.6 (Sun Jun 27 2021)


Authors: 2

0.14.5 (Mon Jun 21 2021)

Bug Fix

  • BF(TST): parallel - take longer for producer to produce #5747 (@yarikoptic)

  • add –on-failure default value and document it #5690 (@christian-monch @yarikoptic)

  • ENH: harmonize “purpose” statements to imperative form #5733 (@yarikoptic)

  • ENH(TST): populate heavy tree with 100 unique keys (not just 1) among 10,000 #5734 (@yarikoptic)

  • BF: do not use .acquired - just get state from acquire() #5718 (@yarikoptic)

  • BF: account for annex now “scanning for annexed” instead of “unlocked” files #5705 (@yarikoptic)

  • interface: Don’t repeat custom summary for non-generator results #5688 (@kyleam)

  • RF: just pip install datalad-installer #5676 (@yarikoptic)

  • DOC: addurls.extract: Drop mention of removed ‘stream’ parameter #5690 (@kyleam)

  • Merge pull request #5674 from kyleam/test-addurls-copy-fix #5674 (@kyleam)

  • Merge pull request #5663 from kyleam/status-ds-equal-path #5663 (@kyleam)

  • Merge pull request #5671 from kyleam/update-fetch-fail #5671 (@kyleam)

  • BF: update: Honor –on-failure if fetch fails #5671 (@kyleam)

  • RF: update: Avoid fetch’s deprecated kwargs #5671 (@kyleam)

  • CLN: update: Drop an unused import #5671 (@kyleam)

  • Merge pull request #5664 from kyleam/addurls-better-url-parts-error #5664 (@kyleam)

  • Merge pull request #5661 from kyleam/sphinx-fix-plugin-refs #5661 (@kyleam)

  • BF: status: Provide special treatment of “this dataset” path #5663 (@kyleam)

  • BF: addurls: Provide better placeholder error for special keys #5664 (@kyleam)

  • RF: addurls: Simply construction of placeholder exception message #5664 (@kyleam)

  • RF: addurls._get_placeholder_exception: Rename a parameter #5664 (@kyleam)

  • RF: status: Avoid repeated Dataset.path access #5663 (@kyleam)

  • DOC: Reference plugins via datalad.api #5661 (@kyleam)

  • download-url: Set up datalad special remote if needed #5648 (@kyleam @yarikoptic)

Pushed to maint

  • MNT: Post-release dance (@kyleam)



  • BF(TST): skip testing for showing “Scanning for …” since not shown if too quick #5727 (@yarikoptic)

  • Revert “TST: test_partial_unlocked: Document and avoid recent git-annex failure” #5651 (@kyleam)

Authors: 4

0.14.4 (May 10, 2021) – .


  • Following an internal call to git-clone, clone assumed that the remote name was “origin”, but this may not be the case if clone.defaultRemoteName is configured (available as of Git 2.30). (#5572)

  • Several test fixes, including updates for changes in git-annex. (#5612) (#5632) (#5639)

0.14.3 (April 28, 2021) – .


  • For outputs that include a glob, run didn’t re-glob after executing the command, which is necessary to catch changes if --explicit or --expand={outputs,both} is specified. (#5594)

  • run now gives an error result rather than a warning when an input glob doesn’t match. (#5594)

  • The procedure for creating a RIA store checks for an existing ria-layout-version file and makes sure its version matches the desired version. This check wasn’t done correctly for SSH hosts. (#5607)

  • A helper for transforming git-annex JSON records into DataLad results didn’t account for the unusual case where the git-annex record doesn’t have a “file” key. (#5580)

  • The test suite required updates for recent changes in PyGithub and git-annex. (#5603) (#5609)

Enhancements and new features

  • The DataLad source repository has long had a tools/cmdline-completion helper. This functionality is now exposed as a command, datalad shell-completion. (#5544)

0.14.2 (April 14, 2021) – .


  • push now works bottom-up, pushing submodules first so that hooks on the remote can aggregate updated subdataset information. (#5416)

  • run-procedure didn’t ensure that the configuration of subdatasets was reloaded. (#5552)

0.14.1 (April 01, 2021) – .


  • The recent default branch changes on GitHub’s side can lead to “git-annex” being selected over “master” as the default branch on GitHub when setting up a sibling with create-sibling-github. To work around this, the current branch is now pushed first. (#5010)

  • The logic for reading in a JSON line from git-annex failed if the response exceeded the buffer size (256 KB on *nix systems).

  • Calling unlock with a path of “.” from within an untracked subdataset incorrectly aborted, complaining that the “dataset containing given paths is not underneath the reference dataset”. (#5458)

  • clone didn’t account for the possibility of multiple accessible ORA remotes or the fact that none of them may be associated with the RIA store being cloned. (#5488)

  • create-sibling-ria didn’t call git update-server-info after setting up the remote repository and, as a result, the repository couldn’t be fetched until something else (e.g., a push) triggered a call to git update-server-info. (#5531)

  • The parser for git-config output didn’t properly handle multi-line values and got thrown off by unexpected and unrelated lines. (#5509)

  • The 0.14 release introduced regressions in the handling of progress bars for git-annex actions, including collapsing progress bars for concurrent operations. (#5421) (#5438)

  • save failed if the user configured Git’s diff.ignoreSubmodules to a non-default value. (#5453)

  • A interprocess lock is now used to prevent a race between checking for an SSH socket’s existence and creating it. (#5466)

  • If a Python procedure script is executable, run-procedure invokes it directly rather than passing it to sys.executable. The non-executable Python procedures that ship with DataLad now include shebangs so that invoking them has a chance of working on file systems that present all files as executable. (#5436)

  • DataLad’s wrapper around argparse failed if an underscore was used in a positional argument. (#5525)

Enhancements and new features

  • DataLad’s method for mapping environment variables to configuration options (e.g., DATALAD_FOO_X__Y to doesn’t work if the subsection name (“FOO”) has an underscore. This limitation can be sidestepped with the new DATALAD_CONFIG_OVERRIDES_JSON environment variable, which can be set to a JSON record of configuration values. (#5505)

0.14.0 (February 02, 2021) – .

Major refactoring and deprecations

  • Git versions below v2.19.1 are no longer supported. (#4650)

  • The minimum git-annex version is still 7.20190503, but, if you’re on Windows (or use adjusted branches in general), please upgrade to at least 8.20200330 but ideally 8.20210127 to get subdataset-related fixes. (#4292) (#5290)

  • The minimum supported version of Python is now 3.6. (#4879)

  • publish is now deprecated in favor of push. It will be removed in the 0.15.0 release at the earliest.

  • A new command runner was added in v0.13. Functionality related to the old runner has now been removed: Runner, GitRunner, and run_gitcommand_on_file_list_chunks from the datalad.cmd module along with the datalad.tests.protocolremote, datalad.cmd.protocol, and datalad.cmd.protocol.prefix configuration options. (#5229)

  • The --no-storage-sibling switch of create-sibling-ria is deprecated in favor of --storage-sibling=off and will be removed in a later release. (#5090)

  • The get_git_dir static method of GitRepo is deprecated and will be removed in a later release. Use the dot_git attribute of an instance instead. (#4597)

  • The ProcessAnnexProgressIndicators helper from has been removed. (#5259)

  • The save argument of install, a noop since v0.6.0, has been dropped. (#5278)

  • The get_URLS method of AnnexCustomRemote is deprecated and will be removed in a later release. (#4955)

  • ConfigManager.get now returns a single value rather than a tuple when there are multiple values for the same key, as very few callers correctly accounted for the possibility of a tuple return value. Callers can restore the old behavior by passing get_all=True. (#4924)

  • In 0.12.0, all of the assure_* functions in datalad.utils were renamed as ensure_*, keeping the old names around as compatibility aliases. The assure_* variants are now marked as deprecated and will be removed in a later release. (#4908)

  • The module, which was deprecated in 0.12.0 and kept as a compatibility shim for, has been removed. (#4583)

  • The saver argument of, marked as obsolete in 0.12.0, has been removed. (#4583)

  • The dataset_only argument of the ConfigManager class was deprecated in 0.12 and has now been removed. (#4828)

  • The linux_distribution_name, linux_distribution_release, and on_debian_wheezy attributes in datalad.utils are no longer set at import time and will be removed in a later release. Use datalad.utils.get_linux_distribution instead. (#4696)

  • datalad.distribution.clone, which was marked as obsolete in v0.12 in favor of datalad.core.distributed.clone, has been removed. (#4904)

  •, announced as deprecated in v0.12.6, has been removed. (#4904)

  • The compat parameter of GitRepo.get_submodules, added in v0.12 as a temporary compatibility layer, has been removed. (#4904)

  • The long-deprecated (and non-functional) url parameter of GitRepo.__init__ has been removed. (#5342)


  • Cloning onto a system that enters adjusted branches by default (as Windows does) did not properly record the clone URL. (#5128)

  • The RIA-specific handling after calling clone was correctly triggered by ria+http URLs but not ria+https URLs. (#4977)

  • If the registered commit wasn’t found when cloning a subdataset, the failed attempt was left around. (#5391)

  • The remote calls to cp and chmod in create-sibling were not portable and failed on macOS. (#5108)

  • A more reliable check is now done to decide if configuration files need to be reloaded. (#5276)

  • The internal command runner’s handling of the event loop has been improved to play nicer with outside applications and scripts that use asyncio. (#5350) (#5367)

Enhancements and new features

  • The subdataset handling for adjusted branches, which is particularly important on Windows where git-annex enters an adjusted branch by default, has been improved. A core piece of the new approach is registering the commit of the primary branch, not its checked out adjusted branch, in the superdataset. Note: This means that git   status will always consider a subdataset on an adjusted branch as dirty while datalad status will look more closely and see if the tip of the primary branch matches the registered commit. (#5241)

  • The performance of the subdatasets command has been improved, with substantial speedups for recursive processing of many subdatasets. (#4868) (#5076)

  • Adding new subdatasets via save has been sped up. (#4793)

  • get, save, and addurls gained support for parallel operations that can be enabled via the --jobs command-line option or the new datalad.runtime.max-jobs configuration option. (#5022)

  • addurls

    • learned how to read data from standard input. (#4669)

    • now supports tab-separated input. (#4845)

    • now lets Python callers pass in a list of records rather than a file name. (#5285)

    • gained a --drop-after switch that signals to drop a file’s content after downloading and adding it to the annex. (#5081)

    • is now able to construct a tree of files from known checksums without downloading content via its new --key option. (#5184)

    • records the URL file in the commit message as provided by the caller rather than using the resolved absolute path. (#5091)

    • is now speedier. (#4867) (#5022)

  • create-sibling-github learned how to create private repositories (thanks to Nolan Nichols). (#4769)

  • create-sibling-ria gained a --storage-sibling option. When --storage-sibling=only is specified, the storage sibling is created without an accompanying Git sibling. This enables using hosts without Git installed for storage. (#5090)

  • The download machinery (and thus the datalad special remote) gained support for a new scheme, shub://, which follows the same format used by singularity run and friends. In contrast to the short-lived URLs obtained by querying Singularity Hub directly, shub:// URLs are suitable for registering with git-annex. (#4816)

  • A provider is now included for URLs. This is useful for storing an image’s blobs in a dataset and registering the URLs with git-annex. (#5129)

  • The add-readme command now links to the DataLad handbook rather than (#4991)

  • New option datalad.locations.extra-procedures specifies an additional location that should be searched for procedures. (#5156)

  • The class for handling configuration values, ConfigManager, now takes a lock before writes to allow for multiple processes to modify the configuration of a dataset. (#4829)

  • clone now records the original, unresolved URL for a subdataset under submodule.<name>.datalad-url in the parent’s .gitmodules, enabling later get calls to use the original URL. This is particularly useful for ria+ URLs. (#5346)

  • Installing a subdataset now uses custom handling rather than calling git submodule update --init. This avoids some locking issues when running get in parallel and enables more accurate source URLs to be recorded. (#4853)

  • GitRepo.get_content_info, a helper that gets triggered by many commands, got faster by tweaking its git ls-files call. (#5067)

  • wtf now includes credentials-related information (e.g. active backends) in the its output. (#4982)

  • The call_git* methods of GitRepo now have a read_only parameter. Callers can set this to True to promise that the provided command does not write to the repository, bypassing the cost of some checks and locking. (#5070)

  • New call_annex* methods in the AnnexRepo class provide an interface for running git-annex commands similar to that of the GitRepo.call_git* methods. (#5163)

  • It’s now possible to register a custom metadata indexer that is discovered by search and used to generate an index. (#4963)

  • The ConfigManager methods get, getbool, getfloat, and getint now return a single value (with same precedence as git   config --get) when there are multiple values for the same key (in the non-committed git configuration, if the key is present there, or in the dataset configuration). For get, the old behavior can be restored by specifying get_all=True. (#4924)

  • Command-line scripts are now defined via the entry_points argument of setuptools.setup instead of the scripts argument. (#4695)

  • Interactive use of --help on the command-line now invokes a pager on more systems and installation setups. (#5344)

  • The datalad special remote now tries to eliminate some unnecessary interactions with git-annex by being smarter about how it queries for URLs associated with a key. (#4955)

  • The GitRepo class now does a better job of handling bare repositories, a step towards bare repositories support in DataLad. (#4911)

  • More internal work to move the code base over to the new command runner. (#4699) (#4855) (#4900) (#4996) (#5002) (#5141) (#5142) (#5229)

0.13.7 (January 04, 2021) – .


  • Cloning from a RIA store on the local file system initialized annex in the Git sibling of the RIA source, which is problematic because all annex-related functionality should go through the storage sibling. clone now sets remote.origin.annex-ignore to true after cloning from RIA stores to prevent this. (#5255)

  • create-sibling invoked cp in a way that was not compatible with macOS. (#5269)

  • Due to a bug in older Git versions (before 2.25), calling status with a file under .git/ (e.g., datalad status .git/config) incorrectly reported the file as untracked. A workaround has been added. (#5258)

  • Update tests for compatibility with latest git-annex. (#5254)

Enhancements and new features

  • copy-file now aborts if .git/ is in the target directory, adding to its existing .git/ safety checks. (#5258)

0.13.6 (December 14, 2020) – .


  • An assortment of fixes for Windows compatibility. (#5113) (#5119) (#5125) (#5127) (#5136) (#5201) (#5200) (#5214)

  • Adding a subdataset on a system that defaults to using an adjusted branch (i.e. doesn’t support symlinks) didn’t properly set up the submodule URL if the source dataset was not in an adjusted state. (#5127)

  • push failed to push to a remote that did not have an annex-uuid value in the local .git/config. (#5148)

  • The default renderer has been improved to avoid a spurious leading space, which led to the displayed path being incorrect in some cases. (#5121)

  • siblings showed an uninformative error message when asked to configure an unknown remote. (#5146)

  • drop confusingly relayed a suggestion from git annex drop to use --force, an option that does not exist in datalad drop. (#5194)

  • create-sibling-github no longer offers user/password authentication because it is no longer supported by GitHub. (#5218)

  • The internal command runner’s handling of the event loop has been tweaked to hopefully fix issues with running DataLad from IPython. (#5106)

  • SSH cleanup wasn’t reliably triggered by the ORA special remote on failure, leading to a stall with a particular version of git-annex, 8.20201103. (This is also resolved on git-annex’s end as of 8.20201127.) (#5151)

Enhancements and new features

  • The credential helper no longer asks the user to repeat tokens or AWS keys. (#5219)

  • The new option datalad.locations.sockets controls where DataLad stores SSH sockets, allowing users to more easily work around file system and path length restrictions. (#5238)

0.13.5 (October 30, 2020) – .


  • SSH connection handling has been reworked to fix cloning on Windows. A new configuration option, datalad.ssh.multiplex-connections, defaults to false on Windows. (#5042)

  • The ORA special remote and post-clone RIA configuration now provide authentication via DataLad’s credential mechanism and better handling of HTTP status codes. (#5025) (#5026)

  • By default, if a git executable is present in the same location as git-annex, DataLad modifies PATH when running git and git-annex so that the bundled git is used. This logic has been tightened to avoid unnecessarily adjusting the path, reducing the cases where the adjustment interferes with the local environment, such as special remotes in a virtual environment being masked by the system-wide variants. (#5035)

  • git-annex is now consistently invoked as “git annex” rather than “git-annex” to work around failures on Windows. (#5001)

  • push called git annex sync ... on plain git repositories. (#5051)

  • save in genernal doesn’t support registering multiple levels of untracked subdatasets, but it can now properly register nested subdatasets when all of the subdataset paths are passed explicitly (e.g., datalad save -d. sub-a sub-a/sub-b). (#5049)

  • When called with --sidecar and --explicit, run didn’t save the sidecar. (#5017)

  • A couple of spots didn’t properly quote format fields when combining substrings into a format string. (#4957)

  • The default credentials configured for indi-s3 prevented anonymous access. (#5045)

Enhancements and new features

  • Messages about suppressed similar results are now rate limited to improve performance when there are many similar results coming through quickly. (#5060)

  • create-sibling-github can now be told to replace an existing sibling by passing --existing=replace. (#5008)

  • Progress bars now react to changes in the terminal’s width (requires tqdm 2.1 or later). (#5057)

0.13.4 (October 6, 2020) – .


  • Ephemeral clones mishandled bare repositories. (#4899)

  • The post-clone logic for configuring RIA stores didn’t consider https:// URLs. (#4977)

  • DataLad custom remotes didn’t escape newlines in messages sent to git-annex. (#4926)

  • The datalad-archives special remote incorrectly treated file names as percent-encoded. (#4953)

  • The result handler didn’t properly escape “%” when constructing its message template. (#4953)

  • In v0.13.0, the tailored rendering for specific subtypes of external command failures (e.g., “out of space” or “remote not available”) was unintentionally switched to the default rendering. (#4966)

  • Various fixes and updates for the NDA authenticator. (#4824)

  • The helper for getting a versioned S3 URL did not support anonymous access or buckets with “.” in their name. (#4985)

  • Several issues with the handling of S3 credentials and token expiration have been addressed. (#4927) (#4931) (#4952)

Enhancements and new features

  • A warning is now given if the detected Git is below v2.13.0 to let users that run into problems know that their Git version is likely the culprit. (#4866)

  • A fix to push in v0.13.2 introduced a regression that surfaces when push.default is configured to “matching” and prevents the git-annex branch from being pushed. Note that, as part of the fix, the current branch is now always pushed even when it wouldn’t be based on the configured refspec or push.default value. (#4896)

  • publish

    • now allows spelling the empty string value of --since= as ^ for consistency with push. (#4683)

    • compares a revision given to --since= with HEAD rather than the working tree to speed up the operation. (#4448)

  • rerun

    • emits more INFO-level log messages. (#4764)

    • provides better handling of adjusted branches and aborts with a clear error for cases that are not supported. (#5328)

  • The archives are handled with p7zip, if available, since DataLad v0.12.0. This implementation now supports .tgz and .tbz2 archives. (#4877)

0.13.3 (August 28, 2020) – .


  • Work around a Python bug that led to our asyncio-based command runner intermittently failing to capture the output of commands that exit very quickly. (#4835)

  • push displayed an overestimate of the transfer size when multiple files pointed to the same key. (#4821)

  • When download-url calls git annex addurl, it catches and reports any failures rather than crashing. A change in v0.12.0 broke this handling in a particular case. (#4817)

Enhancements and new features

  • The wrapper functions returned by decorators are now given more meaningful names to hopefully make tracebacks easier to digest. (#4834)

0.13.2 (August 10, 2020) – .


  • The allow_quick parameter of AnnexRepo.file_has_content and AnnexRepo.is_under_annex is now ignored and will be removed in a later release. This parameter was only relevant for git-annex versions before 7.20190912. (#4736)


  • Updates for compatibility with recent git and git-annex releases. (#4746) (#4760) (#4684)

  • push didn’t sync the git-annex branch when --data=nothing was specified. (#4786)

  • The datalad.clone.reckless configuration wasn’t stored in non-annex datasets, preventing the values from being inherited by annex subdatasets. (#4749)

  • Running the post-update hook installed by create-sibling --ui could overwrite web log files from previous runs in the unlikely event that the hook was executed multiple times in the same second. (#4745)

  • clone inspected git’s standard error in a way that could cause an attribute error. (#4775)

  • When cloning a repository whose HEAD points to a branch without commits, clone tries to find a more useful branch to check out. It unwisely considered adjusted branches. (#4792)

  • Since v0.12.0, SSHManager.close hasn’t closed connections when the ctrl_path argument was explicitly given. (#4757)

  • When working in a dataset in which git annex init had not yet been called, the file_has_content and is_under_annex methods of AnnexRepo incorrectly took the “allow quick” code path on file systems that did not support it (#4736)


  • create now assigns version 4 (random) UUIDs instead of version 1 UUIDs that encode the time and hardware address. (#4790)

  • The documentation for create now does a better job of describing the interaction between --dataset and PATH. (#4763)

  • The format_commit and get_hexsha methods of GitRepo have been sped up. (#4807) (#4806)

  • A better error message is now shown when the ^ or ^. shortcuts for --dataset do not resolve to a dataset. (#4759)

  • A more helpful error message is now shown if a caller tries to download an ftp:// link but does not have request_ftp installed. (#4788)

  • clone now tries harder to get up-to-date availability information after auto-enabling type=git special remotes. (#2897)

0.13.1 (July 17, 2020) – .


  • Cloning a subdataset should inherit the parent’s datalad.clone.reckless value, but that did not happen when cloning via datalad get rather than datalad install or datalad clone. (#4657)

  • The default result renderer crashed when the result did not have a path key. (#4666) (#4673)

  • datalad push didn’t show information about git push errors when the output was not in the format that it expected. (#4674)

  • datalad push silently accepted an empty string for --since even though it is an invalid value. (#4682)

  • Our JavaScript testing setup on Travis grew stale and has now been updated. (Thanks to Xiao Gui.) (#4687)

  • The new class for running Git commands (added in v0.13.0) ignored any changes to the process environment that occurred after instantiation. (#4703)

Enhancements and new features

  • datalad push now avoids unnecessary git push dry runs and pushes all refspecs with a single git push call rather than invoking git   push for each one. (#4692) (#4675)

  • The readability of SSH error messages has been improved. (#4729)

  • avoids calling datalad.utils.get_linux_distribution at import time and caches the result once it is called because, as of Python 3.8, the function uses distro underneath, adding noticeable overhead. (#4696)

    Third-party code should be updated to use get_linux_distribution directly in the unlikely event that the code relied on the import-time call to get_linux_distribution setting the linux_distribution_name, linux_distribution_release, or on_debian_wheezy attributes in `datalad.utils.

0.13.0 (June 23, 2020) – .

A handful of new commands, including copy-file, push, and create-sibling-ria, along with various fixes and enhancements

Major refactoring and deprecations

  • The no_annex parameter of create, which is exposed in the Python API but not the command line, is deprecated and will be removed in a later release. Use the new annex argument instead, flipping the value. Command-line callers that use --no-annex are unaffected. (#4321)

  • datalad add, which was deprecated in 0.12.0, has been removed. (#4158) (#4319)

  • The following GitRepo and AnnexRepo methods have been removed: get_changed_files, get_missing_files, and get_deleted_files. (#4169) (#4158)

  • The get_branch_commits method of GitRepo and AnnexRepo has been renamed to get_branch_commits_. (#3834)

  • The custom commit method of AnnexRepo has been removed, and AnnexRepo.commit now resolves to the parent method, GitRepo.commit. (#4168)

  • GitPython’s git.repo.base.Repo class is no longer available via the .repo attribute of GitRepo and AnnexRepo. (#4172)

  • AnnexRepo.get_corresponding_branch now returns None rather than the current branch name when a managed branch is not checked out. (#4274)

  • The special UUID for git-annex web remotes is now available as datalad.consts.WEB_SPECIAL_REMOTE_UUID. It remains accessible as AnnexRepo.WEB_UUID for compatibility, but new code should use consts.WEB_SPECIAL_REMOTE_UUID (#4460).


  • Widespread improvements in functionality and test coverage on Windows and crippled file systems in general. (#4057) (#4245) (#4268) (#4276) (#4291) (#4296) (#4301) (#4303) (#4304) (#4305) (#4306)

  • AnnexRepo.get_size_from_key incorrectly handled file chunks. (#4081)

  • create-sibling would too readily clobber existing paths when called with --existing=replace. It now gets confirmation from the user before doing so if running interactively and unconditionally aborts when running non-interactively. (#4147)

  • update (#4159)

    • queried the incorrect branch configuration when updating non-annex repositories.

    • didn’t account for the fact that the local repository can be configured as the upstream “remote” for a branch.

  • When the caller included --bare as a git init option, create crashed creating the bare repository, which is currently unsupported, rather than aborting with an informative error message. (#4065)

  • The logic for automatically propagating the ‘origin’ remote when cloning a local source could unintentionally trigger a fetch of a non-local remote. (#4196)

  • All remaining get_submodules() call sites that relied on the temporary compatibility layer added in v0.12.0 have been updated. (#4348)

  • The custom result summary renderer for get, which was visible with --output-format=tailored, displayed incorrect and confusing information in some cases. The custom renderer has been removed entirely. (#4471)

  • The documentation for the Python interface of a command listed an incorrect default when the command overrode the value of command parameters such as result_renderer. (#4480)

Enhancements and new features

  • The default result renderer learned to elide a chain of results after seeing ten consecutive results that it considers similar, which improves the display of actions that have many results (e.g., saving hundreds of files). (#4337)

  • The default result renderer, in addition to “tailored” result renderer, now triggers the custom summary renderer, if any. (#4338)

  • The new command create-sibling-ria provides support for creating a sibling in a RIA store. (#4124)

  • DataLad ships with a new special remote, git-annex-remote-ora, for interacting with RIA stores and a new command export-archive-ora for exporting an archive from a local annex object store. (#4260) (#4203)

  • The new command push provides an alternative interface to publish for pushing a dataset hierarchy to a sibling. (#4206) (#4581) (#4617) (#4620)

  • The new command copy-file copies files and associated availability information from one dataset to another. (#4430)

  • The command examples have been expanded and improved. (#4091) (#4314) (#4464)

  • The tooling for linking to the DataLad Handbook from DataLad’s documentation has been improved. (#4046)

  • The --reckless parameter of clone and install learned two new modes:

    • “ephemeral”, where the .git/annex/ of the cloned repository is symlinked to the local source repository’s. (#4099)

    • “shared-{group|all|…}” that can be used to set up datasets for collaborative write access. (#4324)

  • clone

    • learned to handle dataset aliases in RIA stores when given a URL of the form ria+<protocol>://<storelocation>#~<aliasname>. (#4459)

    • now checks datalad.get.subdataset-source-candidate-NAME to see if NAME starts with three digits, which is taken as a “cost”. Sources with lower costs will be tried first. (#4619)

  • update (#4167)

    • learned to disallow non-fast-forward updates when ff-only is given to the --merge option.

    • gained a --follow option that controls how --merge behaves, adding support for merging in the revision that is registered in the parent dataset rather than merging in the configured branch from the sibling.

    • now provides a result record for merge events.

  • create-sibling now supports local paths as targets in addition to SSH URLs. (#4187)

  • siblings now

    • shows a warning if the caller requests to delete a sibling that does not exist. (#4257)

    • phrases its warning about non-annex repositories in a less alarming way. (#4323)

  • The rendering of command errors has been improved. (#4157)

  • save now

    • displays a message to signal that the working tree is clean, making it more obvious that no results being rendered corresponds to a clean state. (#4106)

    • provides a stronger warning against using --to-git. (#4290)

  • diff and save learned about scenarios where they could avoid unnecessary and expensive work. (#4526) (#4544) (#4549)

  • Calling diff without --recursive but with a path constraint within a subdataset (“/”) now traverses into the subdataset, as “/” would, restricting its report to “/”. (#4235)

  • New option datalad.annex.retry controls how many times git-annex will retry on a failed transfer. It defaults to 3 and can be set to 0 to restore the previous behavior. (#4382)

  • wtf now warns when the specified dataset does not exist. (#4331)

  • The repr and str output of the dataset and repo classes got a facelift. (#4420) (#4435) (#4439)

  • The DataLad Singularity container now comes with p7zip-full.

  • DataLad emits a log message when the current working directory is resolved to a different location due to a symlink. This is now logged at the DEBUG rather than WARNING level, as it typically does not indicate a problem. (#4426)

  • DataLad now lets the caller know that git annex init is scanning for unlocked files, as this operation can be slow in some repositories. (#4316)

  • The log_progress helper learned how to set the starting point to a non-zero value and how to update the total of an existing progress bar, two features needed for planned improvements to how some commands display their progress. (#4438)

  • The ExternalVersions object, which is used to check versions of Python modules and external tools (e.g., git-annex), gained an add method that enables DataLad extensions and other third-party code to include other programs of interest. (#4441)

  • All of the remaining spots that use GitPython have been rewritten without it. Most notably, this includes rewrites of the clone, fetch, and push methods of GitRepo. (#4080) (#4087) (#4170) (#4171) (#4175) (#4172)

  • When GitRepo.commit splits its operation across multiple calls to avoid exceeding the maximum command line length, it now amends to initial commit rather than creating multiple commits. (#4156)

  • GitRepo gained a get_corresponding_branch method (which always returns None), allowing a caller to invoke the method without needing to check if the underlying repo class is GitRepo or AnnexRepo. (#4274)

  • A new helper function datalad.core.local.repo.repo_from_path returns a repo class for a specified path. (#4273)

  • New AnnexRepo method localsync performs a git annex sync that disables external interaction and is particularly useful for propagating changes on an adjusted branch back to the main branch. (#4243)

0.12.7 (May 22, 2020) – .


  • Requesting tailored output (--output=tailored) from a command with a custom result summary renderer produced repeated output. (#4463)

  • A longstanding regression in argcomplete-based command-line completion for Bash has been fixed. You can enable completion by configuring a Bash startup file to run eval   "$(register-python-argcomplete datalad)" or source DataLad’s tools/cmdline-completion. The latter should work for Zsh as well. (#4477)

  • publish didn’t prevent git-fetch from recursing into submodules, leading to a failure when the registered submodule was not present locally and the submodule did not have a remote named ‘origin’. (#4560)

  • addurls botched path handling when the file name format started with “./” and the call was made from a subdirectory of the dataset. (#4504)

  • Double dash options in manpages were unintentionally escaped. (#4332)

  • The check for HTTP authentication failures crashed in situations where content came in as bytes rather than unicode. (#4543)

  • A check in AnnexRepo.whereis could lead to a type error. (#4552)

  • When installing a dataset to obtain a subdataset, get confusingly displayed a message that described the containing dataset as “underneath” the subdataset. (#4456)

  • A couple of Makefile rules didn’t properly quote paths. (#4481)

  • With DueCredit support enabled (DUECREDIT_ENABLE=1), the query for metadata information could flood the output with warnings if datasets didn’t have aggregated metadata. The warnings are now silenced, with the overall failure of a metadata call logged at the debug level. (#4568)

Enhancements and new features

  • The resource identifier helper learned to recognize URLs with embedded Git transport information, such as gcrypt:: (#4529)

  • When running non-interactively, a more informative error is now signaled when the UI backend, which cannot display a question, is asked to do so. (#4553)

0.12.6 (April 23, 2020) – .

Major refactoring and deprecations

  • The value of is no longer considered. The variable will be removed in a later release. (#4409)


  • Staring with v0.12.0, datalad save recorded the current branch of a parent dataset as the branch value in the .gitmodules entry for a subdataset. This behavior is problematic for a few reasons and has been reverted. (#4375)

  • The default for the --jobs option, “auto”, instructed DataLad to pass a value to git-annex’s --jobs equal to min(8, max(3, <number   of CPUs>)), which could lead to issues due to the large number of child processes spawned and file descriptors opened. To avoid this behavior, --jobs=auto now results in git-annex being called with --jobs=1 by default. Configure the new option datalad.runtime.max-annex-jobs to control the maximum value that will be considered when --jobs='auto'. (#4409)

  • Various commands have been adjusted to better handle the case where a remote’s HEAD ref points to an unborn branch. (#4370)

  • search

    • learned to use the query as a regular expression that restricts the keys that are shown for --show-keys short. (#4354)

    • gives a more helpful message when query is an invalid regular expression. (#4398)

  • The code for parsing Git configuration did not follow Git’s behavior of accepting a key with no value as shorthand for key=true. (#4421)

  • needed a compatibility update for a change in how git-annex reports file names. (#4431)

  • create-sibling-github did not gracefully handle a token that did not have the necessary permissions. (#4400)

Enhancements and new features

  • search learned to use the query as a regular expression that restricts the keys that are shown for --show-keys short. (#4354)

  • datalad <subcommand> learned to point to the datalad-container extension when a subcommand from that extension is given but the extension is not installed. (#4400) (#4174)

0.12.5 (Apr 02, 2020) – a small step for datalad …

Fix some bugs and make the world an even better place.


  • Our log_progress helper mishandled the initial display and step of the progress bar. (#4326)

  • AnnexRepo.get_content_annexinfo is designed to accept init=None, but passing that led to an error. (#4330)

  • Update a regular expression to handle an output change in Git v2.26.0. (#4328)

  • We now set LC_MESSAGES to ‘C’ while running git to avoid failures when parsing output that is marked for translation. (#4342)

  • The helper for decoding JSON streams loaded the last line of input without decoding it if the line didn’t end with a new line, a regression introduced in the 0.12.0 release. (#4361)

  • The clone command failed to git-annex-init a fresh clone whenever it considered to add the origin of the origin as a remote. (#4367)

0.12.4 (Mar 19, 2020) – Windows?!

The main purpose of this release is to have one on PyPi that has no associated wheel to enable a working installation on Windows (#4315).


  • The description of the log.outputs config switch did not keep up with code changes and incorrectly stated that the output would be logged at the DEBUG level; logging actually happens at a lower level. (#4317)

0.12.3 (March 16, 2020) – .

Updates for compatibility with the latest git-annex, along with a few miscellaneous fixes

Major refactoring and deprecations

  • All spots that raised a NoDatasetArgumentFound exception now raise a NoDatasetFound exception to better reflect the situation: it is the dataset rather than the argument that is not found. For compatibility, the latter inherits from the former, but new code should prefer the latter. (#4285)


  • Updates for compatibility with git-annex version 8.20200226. (#4214)

  • datalad export-to-figshare failed to export if the generated title was fewer than three characters. It now queries the caller for the title and guards against titles that are too short. (#4140)

  • Authentication was requested multiple times when git-annex launched parallel downloads from the datalad special remote. (#4308)

  • At verbose logging levels, DataLad requests that git-annex display debugging information too. Work around a bug in git-annex that prevented that from happening. (#4212)

  • The internal command runner looked in the wrong place for some configuration variables, including datalad.log.outputs, resulting in the default value always being used. (#4194)

  • publish failed when trying to publish to a git-lfs special remote for the first time. (#4200)

  • AnnexRepo.set_remote_url is supposed to establish shared SSH connections but failed to do so. (#4262)

Enhancements and new features

  • The message provided when a command cannot determine what dataset to operate on has been improved. (#4285)

  • The “aws-s3” authentication type now allows specifying the host through “aws-s3_host”, which was needed to work around an authorization error due to a longstanding upstream bug. (#4239)

  • The xmp metadata extractor now recognizes “.wav” files.

0.12.2 (Jan 28, 2020) – Smoothen the ride

Mostly a bugfix release with various robustifications, but also makes the first step towards versioned dataset installation requests.

Major refactoring and deprecations

  • The minimum required version for GitPython is now 2.1.12. (#4070)


  • The class for handling configuration values, ConfigManager, inappropriately considered the current working directory’s dataset, if any, for both reading and writing when instantiated with dataset=None. This misbehavior is fairly inaccessible through typical use of DataLad. It affects datalad.cfg, the top-level configuration instance that should not consider repository-specific values. It also affects Python users that call Dataset with a path that does not yet exist and persists until that dataset is created. (#4078)

  • update saved the dataset when called with --merge, which is unnecessary and risks committing unrelated changes. (#3996)

  • Confusing and irrelevant information about Python defaults have been dropped from the command-line help. (#4002)

  • The logic for automatically propagating the ‘origin’ remote when cloning a local source didn’t properly account for relative paths. (#4045)

  • Various fixes to file name handling and quoting on Windows. (#4049) (#4050)

  • When cloning failed, error lines were not bubbled up to the user in some scenarios. (#4060)

Enhancements and new features

  • clone (and thus install)

    • now propagates the reckless mode from the superdataset when cloning a dataset into it. (#4037)

    • gained support for ria+<protocol>:// URLs that point to RIA stores. (#4022)

    • learned to read “@version” from ria+ URLs and install that version of a dataset (#4036) and to apply URL rewrites configured through Git’s url.*.insteadOf mechanism (#4064).

    • now copies datalad.get.subdataset-source-candidate-<name> options configured within the superdataset into the subdataset. This is particularly useful for RIA data stores. (#4073)

  • Archives are now (optionally) handled with 7-Zip instead of patool. 7-Zip will be used by default, but patool will be used on non-Windows systems if the datalad.runtime.use-patool option is set or the 7z executable is not found. (#4041)

0.12.1 (Jan 15, 2020) – Small bump after big bang

Fix some fallout after major release.


  • Revert incorrect relative path adjustment to URLs in clone. (#3538)

  • Various small fixes to internal helpers and test to run on Windows (#2566) (#2534)

0.12.0 (Jan 11, 2020) – Krakatoa

This release is the result of more than a year of development that includes fixes for a large number of issues, yielding more robust behavior across a wider range of use cases, and introduces major changes in API and behavior. It is the first release for which extensive user documentation is available in a dedicated DataLad Handbook. Python 3 (3.5 and later) is now the only supported Python flavor.

Major changes 0.12 vs 0.11

  • save fully replaces add (which is obsolete now, and will be removed in a future release).

  • A new Git-annex aware status command enables detailed inspection of dataset hierarchies. The previously available diff command has been adjusted to match status in argument semantics and behavior.

  • The ability to configure dataset procedures prior and after the execution of particular commands has been replaced by a flexible “hook” mechanism that is able to run arbitrary DataLad commands whenever command results are detected that match a specification.

  • Support of the Windows platform has been improved substantially. While performance and feature coverage on Windows still falls behind Unix-like systems, typical data consumer use cases, and standard dataset operations, such as create and save, are now working. Basic support for data provenance capture via run is also functional.

  • Support for Git-annex direct mode repositories has been removed, following the end of support in Git-annex itself.

  • The semantics of relative paths in command line arguments have changed. Previously, a call datalad save --dataset /tmp/myds some/relpath would have been interpreted as saving a file at /tmp/myds/some/relpath into dataset /tmp/myds. This has changed to saving $PWD/some/relpath into dataset /tmp/myds. More generally, relative paths are now always treated as relative to the current working directory, except for path arguments of Dataset class instance methods of the Python API. The resulting partial duplication of path specifications between path and dataset arguments is mitigated by the introduction of two special symbols that can be given as dataset argument: ^ and ^., which identify the topmost superdataset and the closest dataset that contains the working directory, respectively.

  • The concept of a “core API” has been introduced. Commands situated in the module datalad.core (such as create, save, run, status, diff) receive additional scrutiny regarding API and implementation, and are meant to provide longer-term stability. Application developers are encouraged to preferentially build on these commands.

Major refactoring and deprecations since 0.12.0rc6

  • clone has been incorporated into the growing core API. The public --alternative-source parameter has been removed, and a clone_dataset function with multi-source capabilities is provided instead. The --reckless parameter can now take literal mode labels instead of just being a binary flag, but backwards compatibility is maintained.

  • The get_file_content method of GitRepo was no longer used internally or in any known DataLad extensions and has been removed. (#3812)

  • The function get_dataset_root has been replaced by rev_get_dataset_root. rev_get_dataset_root remains as a compatibility alias and will be removed in a later release. (#3815)

  • The add_sibling module, marked obsolete in v0.6.0, has been removed. (#3871)

  • mock is no longer declared as an external dependency because we can rely on it being in the standard library now that our minimum required Python version is 3.5. (#3860)

  • download-url now requires that directories be indicated with a trailing slash rather than interpreting a path as directory when it doesn’t exist. This avoids confusion that can result from typos and makes it possible to support directory targets that do not exist. (#3854)

  • The dataset_only argument of the ConfigManager class is deprecated. Use source="dataset" instead. (#3907)

  • The --proc-pre and --proc-post options have been removed, and configuration values for datalad.COMMAND.proc-pre and datalad.COMMAND.proc-post are no longer honored. The new result hook mechanism provides an alternative for proc-post procedures. (#3963)

Fixes since 0.12.0rc6

  • publish crashed when called with a detached HEAD. It now aborts with an informative message. (#3804)

  • Since 0.12.0rc6 the call to update in siblings resulted in a spurious warning. (#3877)

  • siblings crashed if it encountered an annex repository that was marked as dead. (#3892)

  • The update of rerun in v0.12.0rc3 for the rewritten diff command didn’t account for a change in the output of diff, leading to rerun --report unintentionally including unchanged files in its diff values. (#3873)

  • In 0.12.0rc5 download-url was updated to follow the new path handling logic, but its calls to AnnexRepo weren’t properly adjusted, resulting in incorrect path handling when the called from a dataset subdirectory. (#3850)

  • download-url called git annex addurl in a way that failed to register a URL when its header didn’t report the content size. (#3911)

  • With Git v2.24.0, saving new subdatasets failed due to a bug in that Git release. (#3904)

  • With DataLad configured to stop on failure (e.g., specifying --on-failure=stop from the command line), a failing result record was not rendered. (#3863)

  • Installing a subdataset yielded an “ok” status in cases where the repository was not yet in its final state, making it ineffective for a caller to operate on the repository in response to the result. (#3906)

  • The internal helper for converting git-annex’s JSON output did not relay information from the “error-messages” field. (#3931)

  • run-procedure reported relative paths that were confusingly not relative to the current directory in some cases. It now always reports absolute paths. (#3959)

  • diff inappropriately reported files as deleted in some cases when to was a value other than None. (#3999)

  • An assortment of fixes for Windows compatibility. (#3971) (#3974) (#3975) (#3976) (#3979)

  • Subdatasets installed from a source given by relative path will now have this relative path used as ‘url’ in their .gitmodules record, instead of an absolute path generated by Git. (#3538)

  • clone will now correctly interpret ‘~/…’ paths as absolute path specifications. (#3958)

  • run-procedure mistakenly reported a directory as a procedure. (#3793)

  • The cleanup for batched git-annex processes has been improved. (#3794) (#3851)

  • The function for adding a version ID to an AWS S3 URL doesn’t support URLs with an “s3://” scheme and raises a NotImplementedError exception when it encounters one. The function learned to return a URL untouched if an “s3://” URL comes in with a version ID. (#3842)

  • A few spots needed to be adjusted for compatibility with git-annex’s new --sameas feature, which allows special remotes to share a data store. (#3856)

  • The swallow_logs utility failed to capture some log messages due to an incompatibility with Python 3.7. (#3935)

  • siblings

    • crashed if --inherit was passed but the parent dataset did not have a remote with a matching name. (#3954)

    • configured the wrong pushurl and annexurl values in some cases. (#3955)

Enhancements and new features since 0.12.0rc6

  • By default, datasets cloned from local source paths will now get a configured remote for any recursively discoverable ‘origin’ sibling that is also available from a local path in order to maximize automatic file availability across local annexes. (#3926)

  • The new result hooks mechanism allows callers to specify, via local Git configuration values, DataLad command calls that will be triggered in response to matching result records (i.e., what you see when you call a command with -f json_pp). (#3903)

  • The command interface classes learned to use a new _examples_ attribute to render documentation examples for both the Python and command-line API. (#3821)

  • Candidate URLs for cloning a submodule can now be generated based on configured templates that have access to various properties of the submodule, including its dataset ID. (#3828)

  • DataLad’s check that the user’s Git identity is configured has been sped up and now considers the appropriate environment variables as well. (#3807)

  • The tag method of GitRepo can now tag revisions other than HEAD and accepts a list of arbitrary git tag options. (#3787)

  • When get clones a subdataset and the subdataset’s HEAD differs from the commit that is registered in the parent, the active branch of the subdataset is moved to the registered commit if the registered commit is an ancestor of the subdataset’s HEAD commit. This handling has been moved to a more central location within GitRepo, and now applies to any update_submodule(..., init=True) call. (#3831)

  • The output of datalad -h has been reformatted to improve readability. (#3862)

  • unlock has been sped up. (#3880)

  • run-procedure learned to provide and render more information about discovered procedures, including whether the procedure is overridden by another procedure with the same base name. (#3960)

  • save now (#3817)

    • records the active branch in the superdataset when registering a new subdataset.

    • calls git annex sync when saving a dataset on an adjusted branch so that the changes are brought into the mainline branch.

  • subdatasets now aborts when its dataset argument points to a non-existent dataset. (#3940)

  • wtf now

    • reports the dataset ID if the current working directory is visiting a dataset. (#3888)

    • outputs entries deterministically. (#3927)

  • The ConfigManager class

    • learned to exclude .datalad/config as a source of configuration values, restricting the sources to standard Git configuration files, when called with source="local". (#3907)

    • accepts a value of “override” for its where argument to allow Python callers to more convenient override configuration. (#3970)

  • Commands now accept a dataset value of “^.” as shorthand for “the dataset to which the current directory belongs”. (#3242)

0.12.0rc6 (Oct 19, 2019) – some releases are better than the others

bet we will fix some bugs and make a world even a better place.

Major refactoring and deprecations

  • DataLad no longer supports Python 2. The minimum supported version of Python is now 3.5. (#3629)

  • Much of the user-focused content at has been removed in favor of more up to date and complete material available in the DataLad Handbook. Going forward, the plan is to restrict to technical documentation geared at developers. (#3678)

  • update used to allow the caller to specify which dataset(s) to update as a PATH argument or via the the --dataset option; now only the latter is supported. Path arguments only serve to restrict which subdataset are updated when operating recursively. (#3700)

  • Result records from a get call no longer have a “state” key. (#3746)

  • update and get no longer support operating on independent hierarchies of datasets. (#3700) (#3746)

  • The run update in 0.12.0rc4 for the new path resolution logic broke the handling of inputs and outputs for calls from a subdirectory. (#3747)

  • The is_submodule_modified method of GitRepo as well as two helper functions in, kwargs_to_options and split_remote_branch, were no longer used internally or in any known DataLad extensions and have been removed. (#3702) (#3704)

  • The only_remote option of GitRepo.is_with_annex was not used internally or in any known extensions and has been dropped. (#3768)

  • The get_tags method of GitRepo used to sort tags by committer date. It now sorts them by the tagger date for annotated tags and the committer date for lightweight tags. (#3715)

  • The rev_resolve_path substituted resolve_path helper. (#3797)


  • Correctly handle relative paths in publish. (#3799) (#3102)

  • Do not erroneously discover directory as a procedure. (#3793)

  • Correctly extract version from manpage to trigger use of manpages for --help. (#3798)

  • The cfg_yoda procedure saved all modifications in the repository rather than saving only the files it modified. (#3680)

  • Some spots in the documentation that were supposed appear as two hyphens were incorrectly rendered in the HTML output en-dashs. (#3692)

  • create, install, and clone treated paths as relative to the dataset even when the string form was given, violating the new path handling rules. (#3749) (#3777) (#3780)

  • Providing the “^” shortcut to --dataset didn’t work properly when called from a subdirectory of a subdataset. (#3772)

  • We failed to propagate some errors from git-annex when working with its JSON output. (#3751)

  • With the Python API, callers are allowed to pass a string or list of strings as the cfg_proc argument to create, but the string form was mishandled. (#3761)

  • Incorrect command quoting for SSH calls on Windows that rendered basic SSH-related functionality (e.g., sshrun) on Windows unusable. (#3688)

  • Annex JSON result handling assumed platform-specific paths on Windows instead of the POSIX-style that is happening across all platforms. (#3719)

  • path_is_under() was incapable of comparing Windows paths with different drive letters. (#3728)

Enhancements and new features

  • Provide a collection of “public” call_git* helpers within GitRepo and replace use of “private” and less specific _git_custom_command calls. (#3791)

  • status gained a --report-filetype. Setting it to “raw” can give a performance boost for the price of no longer distinguishing symlinks that point to annexed content from other symlinks. (#3701)

  • save disables file type reporting by status to improve performance. (#3712)

  • subdatasets (#3743)

    • now extends its result records with a contains field that lists which contains arguments matched a given subdataset.

    • yields an ‘impossible’ result record when a contains argument wasn’t matched to any of the reported subdatasets.

  • install now shows more readable output when cloning fails. (#3775)

  • SSHConnection now displays a more informative error message when it cannot start the ControlMaster process. (#3776)

  • If the new configuration option datalad.log.result-level is set to a single level, all result records will be logged at that level. If you’ve been bothered by DataLad’s double reporting of failures, consider setting this to “debug”. (#3754)

  • Configuration values from datalad -c OPTION=VALUE ... are now validated to provide better errors. (#3695)

  • rerun learned how to handle history with merges. As was already the case when cherry picking non-run commits, re-creating merges may results in conflicts, and rerun does not yet provide an interface to let the user handle these. (#2754)

  • The fsck method of AnnexRepo has been enhanced to expose more features of the underlying git fsck command. (#3693)

  • GitRepo now has a for_each_ref_ method that wraps git   for-each-ref, which is used in various spots that used to rely on GitPython functionality. (#3705)

  • Do not pretend to be able to work in optimized (python -O) mode, crash early with an informative message. (#3803)

0.12.0rc5 (September 04, 2019) – .

Various fixes and enhancements that bring the 0.12.0 release closer.

Major refactoring and deprecations

  • The two modules below have a new home. The old locations still exist as compatibility shims and will be removed in a future release.

    • datalad.distribution.subdatasets has been moved to datalad.local.subdatasets (#3429)

    • has been moved to (#3444)

  • The lock method of AnnexRepo and the options parameter of AnnexRepo.unlock were unused internally and have been removed. (#3459)

  • The get_submodules method of GitRepo has been rewritten without GitPython. When the new compat flag is true (the current default), the method returns a value that is compatible with the old return value. This backwards-compatible return value and the compat flag will be removed in a future release. (#3508)

  • The logic for resolving relative paths given to a command has changed (#3435). The new rule is that relative paths are taken as relative to the dataset only if a dataset instance is passed by the caller. In all other scenarios they’re considered relative to the current directory.

    The main user-visible difference from the command line is that using the --dataset argument does not result in relative paths being taken as relative to the specified dataset. (The undocumented distinction between “rel/path” and “./rel/path” no longer exists.)

    All commands under datalad.core and datalad.local, as well as unlock and addurls, follow the new logic. The goal is for all commands to eventually do so.


  • The function for loading JSON streams wasn’t clever enough to handle content that included a Unicode line separator like U2028. (#3524)

  • When unlock was called without an explicit target (i.e., a directory or no paths at all), the call failed if any of the files did not have content present. (#3459)

  • AnnexRepo.get_content_info failed in the rare case of a key without size information. (#3534)

  • save ignored --on-failure in its underlying call to status. (#3470)

  • Calling remove with a subdirectory displayed spurious warnings about the subdirectory files not existing. (#3586)

  • Our processing of git-annex --json output mishandled info messages from special remotes. (#3546)

  • create

    • didn’t bypass the “existing subdataset” check when called with --force as of 0.12.0rc3 (#3552)

    • failed to register the up-to-date revision of a subdataset when --cfg-proc was used with --dataset (#3591)

  • The base downloader had some error handling that wasn’t compatible with Python 3. (#3622)

  • Fixed a number of Unicode py2-compatibility issues. (#3602)

  • AnnexRepo.get_content_annexinfo did not properly chunk file arguments to avoid exceeding the command-line character limit. (#3587)

Enhancements and new features

  • New command create-sibling-gitlab provides an interface for creating a publication target on a GitLab instance. (#3447)

  • subdatasets (#3429)

    • now supports path-constrained queries in the same manner as commands like save and status

    • gained a --contains=PATH option that can be used to restrict the output to datasets that include a specific path.

    • now narrows the listed subdatasets to those underneath the current directory when called with no arguments

  • status learned to accept a plain --annex (no value) as shorthand for --annex basic. (#3534)

  • The .dirty property of GitRepo and AnnexRepo has been sped up. (#3460)

  • The get_content_info method of GitRepo, used by status and commands that depend on status, now restricts its git calls to a subset of files, if possible, for a performance gain in repositories with many files. (#3508)

  • Extensions that do not provide a command, such as those that provide only metadata extractors, are now supported. (#3531)

  • When calling git-annex with --json, we log standard error at the debug level rather than the warning level if a non-zero exit is expected behavior. (#3518)

  • create no longer refuses to create a new dataset in the odd scenario of an empty .git/ directory upstairs. (#3475)

  • As of v2.22.0 Git treats a sub-repository on an unborn branch as a repository rather than as a directory. Our documentation and tests have been updated appropriately. (#3476)

  • addurls learned to accept a --cfg-proc value and pass it to its create calls. (#3562)

0.12.0rc4 (May 15, 2019) – the revolution is over

With the replacement of the save command implementation with rev-save the revolution effort is now over, and the set of key commands for local dataset operations (create, run, save, status, diff) is now complete. This new core API is available from datalad.core.local (and also via datalad.api, as any other command).

Major refactoring and deprecations

  • The add command is now deprecated. It will be removed in a future release.


  • Remove hard-coded dependencies on POSIX path conventions in SSH support code (#3400)

  • Emit an add result when adding a new subdataset during save (#3398)

  • SSH file transfer now actually opens a shared connection, if none exists yet (#3403)

Enhancements and new features

  • SSHConnection now offers methods for file upload and download (get(), put(). The previous copy() method only supported upload and was discontinued (#3401)

0.12.0rc3 (May 07, 2019) – the revolution continues

Continues API consolidation and replaces the create and diff command with more performant implementations.

Major refactoring and deprecations

  • The previous diff command has been replaced by the diff variant from the datalad-revolution extension. (#3366)

  • rev-create has been renamed to create, and the previous create has been removed. (#3383)

  • The procedure setup_yoda_dataset has been renamed to cfg_yoda (#3353).

  • The --nosave of addurls now affects only added content, not newly created subdatasets (#3259).

  • Dataset.get_subdatasets (deprecated since v0.9.0) has been removed. (#3336)

  • The .is_dirty method of GitRepo and AnnexRepo has been replaced by .status or, for a subset of cases, the .dirty property. (#3330)

  • AnnexRepo.get_status has been replaced by AnnexRepo.status. (#3330)


  • status

    • reported on directories that contained only ignored files (#3238)

    • gave a confusing failure when called from a subdataset with an explicitly specified dataset argument and “.” as a path (#3325)

    • misleadingly claimed that the locally present content size was zero when --annex basic was specified (#3378)

  • An informative error wasn’t given when a download provider was invalid. (#3258)

  • Calling rev-save PATH saved unspecified untracked subdatasets. (#3288)

  • The available choices for command-line options that take values are now displayed more consistently in the help output. (#3326)

  • The new pathlib-based code had various encoding issues on Python 2. (#3332)

Enhancements and new features

  • wtf now includes information about the Python version. (#3255)

  • When operating in an annex repository, checking whether git-annex is available is now delayed until a call to git-annex is actually needed, allowing systems without git-annex to operate on annex repositories in a restricted fashion. (#3274)

  • The load_stream on helper now supports auto-detection of compressed files. (#3289)

  • create (formerly rev-create)

    • learned to be speedier by passing a path to status (#3294)

    • gained a --cfg-proc (or -c) convenience option for running configuration procedures (or more accurately any procedure that begins with “cfg_”) in the newly created dataset (#3353)

  • AnnexRepo.set_metadata now returns a list while AnnexRepo.set_metadata_ returns a generator, a behavior which is consistent with the add and add_ method pair. (#3298)

  • AnnexRepo.get_metadata now supports batch querying of known annex files. Note, however, that callers should carefully validate the input paths because the batch call will silently hang if given non-annex files. (#3364)

  • status

    • now reports a “bytesize” field for files tracked by Git (#3299)

    • gained a new option eval_subdataset_state that controls how the subdataset state is evaluated. Depending on the information you need, you can select a less expensive mode to make status faster. (#3324)

    • colors deleted files “red” (#3334)

  • Querying repository content is faster due to batching of git   cat-file calls. (#3301)

  • The dataset ID of a subdataset is now recorded in the superdataset. (#3304)

  • GitRepo.diffstatus

    • now avoids subdataset recursion when the comparison is not with the working tree, which substantially improves performance when diffing large dataset hierarchies (#3314)

    • got smarter and faster about labeling a subdataset as “modified” (#3343)

  • GitRepo.get_content_info now supports disabling the file type evaluation, which gives a performance boost in cases where this information isn’t needed. (#3362)

  • The XMP metadata extractor now filters based on file name to improve its performance. (#3329)

0.12.0rc2 (Mar 18, 2019) – revolution!


  • GitRepo.dirty does not report on nested empty directories (#3196).

  • reports results on deleted files.

Enhancements and new features

  • Absorb a new set of core commands from the datalad-revolution extension:

    • rev-status: like git status, but simpler and working with dataset hierarchies

    • rev-save: a 2-in-1 replacement for save and add

    • rev-create: a ~30% faster create

  • JSON support tools can now read and write compressed files.

0.12.0rc1 (Mar 03, 2019) – to boldly go …

Major refactoring and deprecations

  • Discontinued support for git-annex direct-mode (also no longer supported upstream).

Enhancements and new features

  • Dataset and Repo object instances are now hashable, and can be created based on pathlib Path object instances

  • Imported various additional methods for the Repo classes to query information and save changes.

0.11.8 (Oct 11, 2019) – annex-we-are-catching-up


  • Our internal command runner failed to capture output in some cases. (#3656)

  • Workaround in the tests around python in cPython >= 3.7.5 ‘;’ in the filename confusing mimetypes (#3769) (#3770)

Enhancements and new features

  • Prepared for upstream changes in git-annex, including support for the latest git-annex

    • 7.20190912 auto-upgrades v5 repositories to v7. (#3648) (#3682)

    • 7.20191009 fixed treatment of (larger/smaller)than in .gitattributes (#3765)

  • The cfg_text2git procedure, as well the --text-no-annex option of create, now configure .gitattributes so that empty files are stored in git rather than annex. (#3667)

0.11.7 (Sep 06, 2019) – python2-we-still-love-you-but-…

Primarily bugfixes with some optimizations and refactorings.


  • addurls

    • now provides better handling when the URL file isn’t in the expected format. (#3579)

    • always considered a relative file for the URL file argument as relative to the current working directory, which goes against the convention used by other commands of taking relative paths as relative to the dataset argument. (#3582)

  • run-procedure

    • hard coded “python” when formatting the command for non-executable procedures ending with “.py”. sys.executable is now used. (#3624)

    • failed if arguments needed more complicated quoting than simply surrounding the value with double quotes. This has been resolved for systems that support shlex.quote, but note that on Windows values are left unquoted. (#3626)

  • siblings now displays an informative error message if a local path is given to --url but --name isn’t specified. (#3555)

  • sshrun, the command DataLad uses for GIT_SSH_COMMAND, didn’t support all the parameters that Git expects it to. (#3616)

  • Fixed a number of Unicode py2-compatibility issues. (#3597)

  • download-url now will create leading directories of the output path if they do not exist (#3646)

Enhancements and new features

  • The annotate-paths helper now caches subdatasets it has seen to avoid unnecessary calls. (#3570)

  • A repeated configuration query has been dropped from the handling of --proc-pre and --proc-post. (#3576)

  • Calls to git annex find now use --in=. instead of the alias --in=here to take advantage of an optimization that git-annex (as of the current release, 7.20190730) applies only to the former. (#3574)

  • addurls now suggests close matches when the URL or file format contains an unknown field. (#3594)

  • Shared logic used in the files of DataLad and its extensions has been moved to modules in the _datalad_build_support/ directory. (#3600)

  • Get ready for upcoming git-annex dropping support for direct mode (#3631)

0.11.6 (Jul 30, 2019) – am I the last of 0.11.x?

Primarily bug fixes to achieve more robust performance


  • Our tests needed various adjustments to keep up with upstream changes in Travis and Git. (#3479) (#3492) (#3493)

  • AnnexRepo.is_special_annex_remote was too selective in what it considered to be a special remote. (#3499)

  • We now provide information about unexpected output when git-annex is called with --json. (#3516)

  • Exception logging in the __del__ method of GitRepo and AnnexRepo no longer fails if the names it needs are no longer bound. (#3527)

  • addurls botched the construction of subdataset paths that were more than two levels deep and failed to create datasets in a reliable, breadth-first order. (#3561)

  • Cloning a type=git special remote showed a spurious warning about the remote not being enabled. (#3547)

Enhancements and new features

  • For calls to git and git-annex, we disable automatic garbage collection due to past issues with GitPython’s state becoming stale, but doing so results in a larger .git/objects/ directory that isn’t cleaned up until garbage collection is triggered outside of DataLad. Tests with the latest GitPython didn’t reveal any state issues, so we’ve re-enabled automatic garbage collection. (#3458)

  • rerun learned an --explicit flag, which it relays to its calls to [run][[]]. This makes it possible to call rerun in a dirty working tree (#3498).

  • The metadata command aborts earlier if a metadata extractor is unavailable. (#3525)

0.11.5 (May 23, 2019) – stability is not overrated

Should be faster and less buggy, with a few enhancements.


  • create-sibling (#3318)

    • Siblings are no longer configured with a post-update hook unless a web interface is requested with --ui.

    • git submodule update --init is no longer called from the post-update hook.

    • If --inherit is given for a dataset without a superdataset, a warning is now given instead of raising an error.

  • The internal command runner failed on Python 2 when its env argument had unicode values. (#3332)

  • The safeguard that prevents creating a dataset in a subdirectory that already contains tracked files for another repository failed on Git versions before 2.14. For older Git versions, we now warn the caller that the safeguard is not active. (#3347)

  • A regression introduced in v0.11.1 prevented save from committing changes under a subdirectory when the subdirectory was specified as a path argument. (#3106)

  • A workaround introduced in v0.11.1 made it possible for save to do a partial commit with an annex file that has gone below the annex.largefiles threshold. The logic of this workaround was faulty, leading to files being displayed as typechanged in the index following the commit. (#3365)

  • The resolve_path() helper confused paths that had a semicolon for SSH RIs. (#3425)

  • The detection of SSH RIs has been improved. (#3425)

Enhancements and new features

  • The internal command runner was too aggressive in its decision to sleep. (#3322)

  • The “INFO” label in log messages now retains the default text color for the terminal rather than using white, which only worked well for terminals with dark backgrounds. (#3334)

  • A short flag -R is now available for the --recursion-limit flag, a flag shared by several subcommands. (#3340)

  • The authentication logic for create-sibling-github has been revamped and now supports 2FA. (#3180)

  • New configuration option datalad.ui.progressbar can be used to configure the default backend for progress reporting (“none”, for example, results in no progress bars being shown). (#3396)

  • A new progress backend, available by setting datalad.ui.progressbar to “log”, replaces progress bars with a log message upon completion of an action. (#3396)

  • DataLad learned to consult the NO_COLOR environment variable and the new datalad.ui.color configuration option when deciding to color output. The default value, “auto”, retains the current behavior of coloring output if attached to a TTY (#3407).

  • clean now removes annex transfer directories, which is useful for cleaning up failed downloads. (#3374)

  • clone no longer refuses to clone into a local path that looks like a URL, making its behavior consistent with git clone. (#3425)

  • wtf

    • Learned to fall back to the dist package if platform.dist, which has been removed in the yet-to-be-release Python 3.8, does not exist. (#3439)

    • Gained a --section option for limiting the output to specific sections and a --decor option, which currently knows how to format the output as GitHub’s <details> section. (#3440)

0.11.4 (Mar 18, 2019) – get-ready

Largely a bug fix release with a few enhancements


  • 0.11.x series will be the last one with support for direct mode of git-annex which is used on crippled (no symlinks and no locking) filesystems. v7 repositories should be used instead.


  • Extraction of .gz files is broken without p7zip installed. We now abort with an informative error in this situation. (#3176)

  • Committing failed in some cases because we didn’t ensure that the path passed to git read-tree --index-output=... resided on the same filesystem as the repository. (#3181)

  • Some pointless warnings during metadata aggregation have been eliminated. (#3186)

  • With Python 3 the LORIS token authenticator did not properly decode a response (#3205).

  • With Python 3 downloaders unnecessarily decoded the response when getting the status, leading to an encoding error. (#3210)

  • In some cases, our internal command Runner did not adjust the environment’s PWD to match the current working directory specified with the cwd parameter. (#3215)

  • The specification of the pyliblzma dependency was broken. (#3220)

  • search displayed an uninformative blank log message in some cases. (#3222)

  • The logic for finding the location of the aggregate metadata DB anchored the search path incorrectly, leading to a spurious warning. (#3241)

  • Some progress bars were still displayed when stdout and stderr were not attached to a tty. (#3281)

  • Check for stdin/out/err to not be closed before checking for .isatty. (#3268)

Enhancements and new features

  • Creating a new repository now aborts if any of the files in the directory are tracked by a repository in a parent directory. (#3211)

  • run learned to replace the {tmpdir} placeholder in commands with a temporary directory. (#3223)

  • duecredit support has been added for citing DataLad itself as well as datasets that an analysis uses. (#3184)

  • The eval_results interface helper unintentionally modified one of its arguments. (#3249)

  • A few DataLad constants have been added, changed, or renamed (#3250):

    • HANDLE_META_DIR is now DATALAD_DOTDIR. The old name should be considered deprecated.

    • METADATA_DIR now refers to DATALAD_DOTDIR/metadata rather than DATALAD_DOTDIR/meta (which is still available as OLDMETADATA_DIR).

    • The new DATASET_METADATA_FILE refers to METADATA_DIR/dataset.json.

    • The new DATASET_CONFIG_FILE refers to DATALAD_DOTDIR/config.


0.11.3 (Feb 19, 2019) – read-me-gently

Just a few of important fixes and minor enhancements.


  • The logic for setting the maximum command line length now works around Python 3.4 returning an unreasonably high value for SC_ARG_MAX on Debian systems. (#3165)

  • DataLad commands that are conceptually “read-only”, such as datalad ls -L, can fail when the caller lacks write permissions because git-annex tries merging remote git-annex branches to update information about availability. DataLad now disables annex.merge-annex-branches in some common “read-only” scenarios to avoid these failures. (#3164)

Enhancements and new features

  • Accessing an “unbound” dataset method now automatically imports the necessary module rather than requiring an explicit import from the Python caller. For example, calling Dataset.add no longer needs to be preceded by from datalad.distribution.add import Add or an import of datalad.api. (#3156)

  • Configuring the new variable datalad.ssh.identityfile instructs DataLad to pass a value to the -i option of ssh. (#3149) (#3168)

0.11.2 (Feb 07, 2019) – live-long-and-prosper

A variety of bugfixes and enhancements

Major refactoring and deprecations

  • All extracted metadata is now placed under git-annex by default. Previously files smaller than 20 kb were stored in git. (#3109)

  • The function datalad.cmd.get_runner has been removed. (#3104)


  • Improved handling of long commands:

    • The code that inspected SC_ARG_MAX didn’t check that the reported value was a sensible, positive number. (#3025)

    • More commands that invoke git and git-annex with file arguments learned to split up the command calls when it is likely that the command would fail due to exceeding the maximum supported length. (#3138)

  • The setup_yoda_dataset procedure created a malformed .gitattributes line. (#3057)

  • download-url unnecessarily tried to infer the dataset when --no-save was given. (#3029)

  • rerun aborted too late and with a confusing message when a ref specified via --onto didn’t exist. (#3019)

  • run:

    • run didn’t preserve the current directory prefix (“./”) on inputs and outputs, which is problematic if the caller relies on this representation when formatting the command. (#3037)

    • Fixed a number of unicode py2-compatibility issues. (#3035) (#3046)

    • To proceed with a failed command, the user was confusingly instructed to use save instead of add even though run uses add underneath. (#3080)

  • Fixed a case where the helper class for checking external modules incorrectly reported a module as unknown. (#3051)

  • add-archive-content mishandled the archive path when the leading path contained a symlink. (#3058)

  • Following denied access, the credential code failed to consider a scenario, leading to a type error rather than an appropriate error message. (#3091)

  • Some tests failed when executed from a git worktree checkout of the source repository. (#3129)

  • During metadata extraction, batched annex processes weren’t properly terminated, leading to issues on Windows. (#3137)

  • add incorrectly handled an “invalid repository” exception when trying to add a submodule. (#3141)

  • Pass GIT_SSH_VARIANT=ssh to git processes to be able to specify alternative ports in SSH urls

Enhancements and new features

  • search learned to suggest closely matching keys if there are no hits. (#3089)

  • create-sibling

  • Interface classes can now override the default renderer for summarizing results. (#3061)

  • run:

    • --input and --output can now be shortened to -i and -o. (#3066)

    • Placeholders such as “{inputs}” are now expanded in the command that is shown in the commit message subject. (#3065)

    • gained an extra_inputs argument so that wrappers like datalad-container can specify additional inputs that aren’t considered when formatting the command string. (#3038)

    • “–” can now be used to separate options for run and those for the command in ambiguous cases. (#3119)

  • The utilities create_tree and ok_file_has_content now support “.gz” files. (#3049)

  • The Singularity container for 0.11.1 now uses nd_freeze to make its builds reproducible.

  • A publications page has been added to the documentation. (#3099)

  • GitRepo.set_gitattributes now accepts a mode argument that controls whether the .gitattributes file is appended to (default) or overwritten. (#3115)

  • datalad --help now avoids using man so that the list of subcommands is shown. (#3124)

0.11.1 (Nov 26, 2018) – v7-better-than-v6

Rushed out bugfix release to stay fully compatible with recent git-annex which introduced v7 to replace v6.


  • install: be able to install recursively into a dataset (#2982)

  • save: be able to commit/save changes whenever files potentially could have swapped their storage between git and annex (#1651) (#2752) (#3009)

  • [aggregate-metadata][]:

    • dataset’s itself is now not “aggregated” if specific paths are provided for aggregation (#3002). That resolves the issue of -r invocation aggregating all subdatasets of the specified dataset as well

    • also compare/verify the actual content checksum of aggregated metadata while considering subdataset metadata for re-aggregation (#3007)

  • annex commands are now chunked assuming 50% “safety margin” on the maximal command line length. Should resolve crashes while operating of too many files at ones (#3001)

  • run sidecar config processing (#2991)

  • no double trailing period in docs (#2984)

  • correct identification of the repository with symlinks in the paths in the tests (#2972)

  • re-evaluation of dataset properties in case of dataset changes (#2946)

  • [text2git][] procedure to use ds.repo.set_gitattributes (#2974) (#2954)

  • Switch to use plain os.getcwd() if inconsistency with env var $PWD is detected (#2914)

  • Make sure that credential defined in env var takes precedence (#2960) (#2950)

Enhancements and new features

  • shub://datalad/datalad:git-annex-dev provides a Debian buster Singularity image with build environment for git-annex. tools/bisect-git-annex provides a helper for running git bisect on git-annex using that Singularity container (#2995)

  • Added .zenodo.json for better integration with Zenodo for citation

  • run-procedure now provides names and help messages with a custom renderer for (#2993)

  • Documentation: point to datalad-revolution extension (prototype of the greater DataLad future)

  • run

    • support injecting of a detached command (#2937)

  • annex metadata extractor now extracts annex.key metadata record. Should allow now to identify uses of specific files etc (#2952)

  • Test that we can install from

  • Proper rendering of CommandError (e.g. in case of “out of space” error) (#2958)

0.11.0 (Oct 23, 2018) – Soon-to-be-perfect

git-annex 6.20180913 (or later) is now required - provides a number of fixes for v6 mode operations etc.

Major refactoring and deprecations

  • datalad.consts.LOCAL_CENTRAL_PATH constant was deprecated in favor of datalad.locations.default-dataset configuration variable (#2835)

Minor refactoring

  • "notneeded" messages are no longer reported by default results renderer

  • run no longer shows commit instructions upon command failure when explicit is true and no outputs are specified (#2922)

  • get_git_dir moved into GitRepo (#2886)

  • _gitpy_custom_call removed from GitRepo (#2894)

  • GitRepo.get_merge_base argument is now called commitishes instead of treeishes (#2903)


  • update should not leave the dataset in non-clean state (#2858) and some other enhancements (#2859)

  • Fixed chunking of the long command lines to account for decorators and other arguments (#2864)

  • Progress bar should not crash the process on some missing progress information (#2891)

  • Default value for jobs set to be "auto" (not None) to take advantage of possible parallel get if in -g mode (#2861)

  • wtf must not crash if git-annex is not installed etc (#2865), (#2865), (#2918), (#2917)

  • Fixed paths (with spaces etc) handling while reporting annex error output (#2892), (#2893)

  • __del__ should not access .repo but ._repo to avoid attempts for reinstantiation etc (#2901)

  • Fix up submodule .git right in GitRepo.add_submodule to avoid added submodules being non git-annex friendly (#2909), (#2904)

  • run-procedure (#2905)

    • now will provide dataset into the procedure if called within dataset

    • will not crash if procedure is an executable without .py or .sh suffixes

  • Use centralized .gitattributes handling while setting annex backend (#2912)

  • GlobbedPaths.expand(..., full=True) incorrectly returned relative paths when called more than once (#2921)

Enhancements and new features

  • Report progress on clone when installing from “smart” git servers (#2876)

  • Stale/unused sth_like_file_has_content was removed (#2860)

  • Enhancements to search to operate on “improved” metadata layouts (#2878)

  • Output of git annex init operation is now logged (#2881)

  • New

    • GitRepo.cherry_pick (#2900)

    • GitRepo.format_commit (#2902)

  • run-procedure (#2905)

    • procedures can now recursively be discovered in subdatasets as well. The uppermost has highest priority

    • Procedures in user and system locations now take precedence over those in datasets. (Sep 13, 2018) – Nothing-is-perfect

Emergency bugfix to address forgotten boost of version in datalad/

0.10.3 (Sep 13, 2018) – Almost-perfect

This is largely a bugfix release which addressed many (but not yet all) issues of working with git-annex direct and version 6 modes, and operation on Windows in general. Among enhancements you will see the support of public S3 buckets (even with periods in their names), ability to configure new providers interactively, and improved egrep search backend.

Although we do not require with this release, it is recommended to make sure that you are using a recent git-annex since it also had a variety of fixes and enhancements in the past months.


  • Parsing of combined short options has been broken since DataLad v0.10.0. (#2710)

  • The datalad save instructions shown by datalad run for a command with a non-zero exit were incorrectly formatted. (#2692)

  • Decompression of zip files (e.g., through datalad   add-archive-content) failed on Python 3. (#2702)

  • Windows:

    • colored log output was not being processed by colorama. (#2707)

    • more codepaths now try multiple times when removing a file to deal with latency and locking issues on Windows. (#2795)

  • Internal git fetch calls have been updated to work around a GitPython BadName issue. (#2712), (#2794)

  • The progress bar for annex file transferring was unable to handle an empty file. (#2717)

  • datalad add-readme halted when no aggregated metadata was found rather than displaying a warning. (#2731)

  • datalad rerun failed if --onto was specified and the history contained no run commits. (#2761)

  • Processing of a command’s results failed on a result record with a missing value (e.g., absent field or subfield in metadata). Now the missing value is rendered as “N/A”. (#2725).

  • A couple of documentation links in the “Delineation from related solutions” were misformatted. (#2773)

  • With the latest git-annex, several known V6 failures are no longer an issue. (#2777)

  • In direct mode, commit changes would often commit annexed content as regular Git files. A new approach fixes this and resolves a good number of known failures. (#2770)

  • The reporting of command results failed if the current working directory was removed (e.g., after an unsuccessful install). (#2788)

  • When installing into an existing empty directory, datalad install removed the directory after a failed clone. (#2788)

  • datalad run incorrectly handled inputs and outputs for paths with spaces and other characters that require shell escaping. (#2798)

  • Globbing inputs and outputs for datalad run didn’t work correctly if a subdataset wasn’t installed. (#2796)

  • Minor (in)compatibility with git 2.19 - (no) trailing period in an error message now. (#2815)

Enhancements and new features

  • Anonymous access is now supported for S3 and other downloaders. (#2708)

  • A new interface is available to ease setting up new providers. (#2708)

  • Metadata: changes to egrep mode search (#2735)

    • Queries in egrep mode are now case-sensitive when the query contains any uppercase letters and are case-insensitive otherwise. The new mode egrepcs can be used to perform a case-sensitive query with all lower-case letters.

    • Search can now be limited to a specific key.

    • Multiple queries (list of expressions) are evaluated using AND to determine whether something is a hit.

    • A single multi-field query (e.g., pa*:findme) is a hit, when any matching field matches the query.

    • All matching key/value combinations across all (multi-field) queries are reported in the query_matched result field.

    • egrep mode now shows all hits rather than limiting the results to the top 20 hits.

  • The documentation on how to format commands for datalad run has been improved. (#2703)

  • The method for determining the current working directory on Windows has been improved. (#2707)

  • datalad --version now simply shows the version without the license. (#2733)

  • datalad export-archive learned to export under an existing directory via its --filename option. (#2723)

  • datalad export-to-figshare now generates the zip archive in the root of the dataset unless --filename is specified. (#2723)

  • After importing datalad.api, help(datalad.api) (or datalad.api? in IPython) now shows a summary of the available DataLad commands. (#2728)

  • Support for using datalad from IPython has been improved. (#2722)

  • datalad wtf now returns structured data and reports the version of each extension. (#2741)

  • The internal handling of gitattributes information has been improved. A user-visible consequence is that datalad create   --force no longer duplicates existing attributes. (#2744)

  • The “annex” metadata extractor can now be used even when no content is present. (#2724)

  • The add_url_to_file method (called by commands like datalad   download-url and datalad add-archive-content) learned how to display a progress bar. (#2738)

0.10.2 (Jul 09, 2018) – Thesecuriestever

Primarily a bugfix release to accommodate recent git-annex release forbidding file:// and http://localhost/ URLs which might lead to revealing private files if annex is publicly shared.


  • fixed testing to be compatible with recent git-annex (6.20180626)

  • download-url will now download to current directory instead of the top of the dataset

Enhancements and new features

  • do not quote ~ in URLs to be consistent with quote implementation in Python 3.7 which now follows RFC 3986

  • run support for user-configured placeholder values

  • documentation on native git-annex metadata support

  • handle 401 errors from LORIS tokens

  • yoda procedure will instantiate

  • --discover option added to run-procedure to list available procedures

0.10.1 (Jun 17, 2018) – OHBM polish

The is a minor bugfix release.


  • Be able to use backports.lzma as a drop-in replacement for pyliblzma.

  • Give help when not specifying a procedure name in run-procedure.

  • Abort early when a downloader received no filename.

  • Avoid rerun error when trying to unlock non-available files.

0.10.0 (Jun 09, 2018) – The Release

This release is a major leap forward in metadata support.

Major refactoring and deprecations

  • Metadata

    • Prior metadata provided by datasets under .datalad/meta is no longer used or supported. Metadata must be reaggregated using 0.10 version

    • Metadata extractor types are no longer auto-guessed and must be explicitly specified in datalad.metadata.nativetype config (could contain multiple values)

    • Metadata aggregation of a dataset hierarchy no longer updates all datasets in the tree with new metadata. Instead, only the target dataset is updated. This behavior can be changed via the –update-mode switch. The new default prevents needless modification of (3rd-party) subdatasets.

    • Neuroimaging metadata support has been moved into a dedicated extension:

  • Crawler

  • export_tarball plugin has been generalized to export_archive and can now also generate ZIP archives.

  • By default a dataset X is now only considered to be a super-dataset of another dataset Y, if Y is also a registered subdataset of X.


A number of fixes did not make it into the 0.9.x series:

  • Dynamic configuration overrides via the -c option were not in effect.

  • save is now more robust with respect to invocation in subdirectories of a dataset.

  • unlock now reports correct paths when running in a dataset subdirectory.

  • get is more robust to path that contain symbolic links.

  • symlinks to subdatasets of a dataset are now correctly treated as a symlink, and not as a subdataset

  • add now correctly saves staged subdataset additions.

  • Running datalad save in a dataset no longer adds untracked content to the dataset. In order to add content a path has to be given, e.g. datalad save .

  • wtf now works reliably with a DataLad that wasn’t installed from Git (but, e.g., via pip)

  • More robust URL handling in simple_with_archives crawler pipeline.

Enhancements and new features

  • Support for DataLad extension that can contribute API components from 3rd-party sources, incl. commands, metadata extractors, and test case implementations. See for a demo extension.

  • Metadata (everything has changed!)

    • Metadata extraction and aggregation is now supported for datasets and individual files.

    • Metadata query via search can now discover individual files.

    • Extracted metadata can now be stored in XZ compressed files, is optionally annexed (when exceeding a configurable size threshold), and obtained on demand (new configuration option datalad.metadata.create-aggregate-annex-limit).

    • Status and availability of aggregated metadata can now be reported via metadata --get-aggregates

    • New configuration option datalad.metadata.maxfieldsize to exclude too large metadata fields from aggregation.

    • The type of metadata is no longer guessed during metadata extraction. A new configuration option datalad.metadata.nativetype was introduced to enable one or more particular metadata extractors for a dataset.

    • New configuration option to enable the storage of aggregated metadata for dataset content (i.e. file-based metadata) in contrast to just metadata describing a dataset as a whole.

  • search was completely reimplemented. It offers three different modes now:

    • ‘egrep’ (default): expression matching in a plain string version of metadata

    • ‘textblob’: search a text version of all metadata using a fully featured query language (fast indexing, good for keyword search)

    • ‘autofield’: search an auto-generated index that preserves individual fields of metadata that can be represented in a tabular structure (substantial indexing cost, enables the most detailed queries of all modes)

  • New extensions:

    • addurls, an extension for creating a dataset (and possibly subdatasets) from a list of URLs.

    • export_to_figshare

    • extract_metadata

  • add_readme makes use of available metadata

  • By default the wtf extension now hides sensitive information, which can be included in the output by passing --senstive=some or --senstive=all.

  • Reduced startup latency by only importing commands necessary for a particular command line call.

  • create:

    • -d <parent> --nosave now registers subdatasets, when possible.

    • --fake-dates configures dataset to use fake-dates

  • run now provides a way for the caller to save the result when a command has a non-zero exit status.

  • datalad rerun now has a --script option that can be used to extract previous commands into a file.

  • A DataLad Singularity container is now available on Singularity Hub.

  • More casts have been embedded in the use case section of the documentation.

  • datalad --report-status has a new value ‘all’ that can be used to temporarily re-enable reporting that was disable by configuration settings.

0.9.3 (Mar 16, 2018) – pi+0.02 release

Some important bug fixes which should improve usability


  • datalad-archives special remote now will lock on acquiring or extracting an archive - this allows for it to be used with -J flag for parallel operation

  • relax introduced in 0.9.2 demand on git being configured for datalad operation - now we will just issue a warning

  • datalad ls should now list “authored date” and work also for datasets in detached HEAD mode

  • datalad save will now save original file as well, if file was “git mv”ed, so you can now datalad run git mv old new and have changes recorded

Enhancements and new features

  • --jobs argument now could take auto value which would decide on # of jobs depending on the # of available CPUs. git-annex > 6.20180314 is recommended to avoid regression with -J.

  • memoize calls to RI meta-constructor – should speed up operation a bit

  • DATALAD_SEED environment variable could be used to seed Python RNG and provide reproducible UUIDs etc (useful for testing and demos)

0.9.2 (Mar 04, 2018) – it is (again) better than ever

Largely a bugfix release with a few enhancements.


  • Execution of external commands (git) should not get stuck when lots of both stdout and stderr output, and should not loose remaining output in some cases

  • Config overrides provided in the command line (-c) should now be handled correctly

  • Consider more remotes (not just tracking one, which might be none) while installing subdatasets

  • Compatibility with git 2.16 with some changed behaviors/annotations for submodules

  • Fail remove if annex drop failed

  • Do not fail operating on files which start with dash (-)

  • URL unquote paths within S3, URLs and DataLad RIs (///)

  • In non-interactive mode fail if authentication/access fails

  • Web UI:

    • refactored a little to fix incorrect listing of submodules in subdirectories

    • now auto-focuses on search edit box upon entering the page

  • Assure that extracted from tarballs directories have executable bit set

Enhancements and new features

  • A log message and progress bar will now inform if a tarball to be downloaded while getting specific files (requires git-annex > 6.20180206)

  • A dedicated datalad rerun command capable of rerunning entire sequences of previously run commands. Reproducibility through VCS. Use ``run`` even if not interested in ``rerun``

  • Alert the user if git is not yet configured but git operations are requested

  • Delay collection of previous ssh connections until it is actually needed. Also do not require ‘:’ while specifying ssh host

  • AutomagicIO: Added proxying of isfile, lzma.LZMAFile and

  • Testing:

    • added DATALAD_DATASETS_TOPURL= to run tests against another website to not obscure access stats

    • tests run against temporary HOME to avoid side-effects

    • better unit-testing of interactions with special remotes

  • describes how to setup and use git-hub tool to “attach” commits to an issue making it into a PR

  • DATALAD_USE_DEFAULT_GIT env variable could be used to cause DataLad to use default (not the one possibly bundled with git-annex) git

  • Be more robust while handling not supported requests by annex in special remotes

  • Use of swallow_logs in the code was refactored away – less mysteries now, just increase logging level

  • wtf plugin will report more information about environment, externals and the system

0.9.1 (Oct 01, 2017) – “DATALAD!”(JBTM)

Minor bugfix release


  • Should work correctly with subdatasets named as numbers of bool values (requires also GitPython >= 2.1.6)

  • Custom special remotes should work without crashing with git-annex >= 6.20170924

0.9.0 (Sep 19, 2017) – isn’t it a lucky day even though not a Friday?

Major refactoring and deprecations

  • the files argument of save has been renamed to path to be uniform with any other command

  • all major commands now implement more uniform API semantics and result reporting. Functionality for modification detection of dataset content has been completely replaced with a more efficient implementation

  • publish now features a --transfer-data switch that allows for a disambiguous specification of whether to publish data – independent of the selection which datasets to publish (which is done via their paths). Moreover, publish now transfers data before repository content is pushed.


  • drop no longer errors when some subdatasets are not installed

  • install will no longer report nothing when a Dataset instance was given as a source argument, but rather perform as expected

  • remove doesn’t remove when some files of a dataset could not be dropped

  • publish

    • no longer hides error during a repository push

    • publish behaves “correctly” for --since= in considering only the differences the last “pushed” state

    • data transfer handling while publishing with dependencies, to github

  • improved robustness with broken Git configuration

  • search should search for unicode strings correctly and not crash

  • robustify git-annex special remotes protocol handling to allow for spaces in the last argument

  • UI credentials interface should now allow to Ctrl-C the entry

  • should not fail while operating on submodules named with numerics only or by bool (true/false) names

  • crawl templates should not now override settings for largefiles if specified in .gitattributes

Enhancements and new features

  • Exciting new feature run command to protocol execution of an external command and rerun computation if desired. See screencast

  • save now uses Git for detecting with sundatasets need to be inspected for potential changes, instead of performing a complete traversal of a dataset tree

  • add looks for changes relative to the last committed state of a dataset to discover files to add more efficiently

  • diff can now report untracked files in addition to modified files

  • [uninstall][] will check itself whether a subdataset is properly registered in a superdataset, even when no superdataset is given in a call

  • subdatasets can now configure subdatasets for exclusion from recursive installation (datalad-recursiveinstall submodule configuration property)

  • precrafted pipelines of [crawl][] now will not override annex.largefiles setting if any was set within .gitattribues (e.g. by datalad create --text-no-annex)

  • framework for screencasts: tools/cast* tools and sample cast scripts under doc/casts which are published at

  • new project YouTube channel

  • tests failing in direct and/or v6 modes marked explicitly

0.8.1 (Aug 13, 2017) – the best birthday gift



  • Do not attempt to update a not installed sub-dataset

  • In case of too many files to be specified for get or copy_to, we will make multiple invocations of underlying git-annex command to not overfill command line

  • More robust handling of unicode output in terminals which might not support it

Enhancements and new features

  • Ship a copy of numpy.testing to facilitate [test][] without requiring numpy as dependency. Also allow to pass to command which test(s) to run

  • In get and copy_to provide actual original requested paths, not the ones we deduced need to be transferred, solely for knowing the total

0.8.0 (Jul 31, 2017) – it is better than ever

A variety of fixes and enhancements


  • publish would now push merged git-annex branch even if no other changes were done

  • publish should be able to publish using relative path within SSH URI (git hook would use relative paths)

  • publish should better tollerate publishing to pure git and git-annex special remotes

Enhancements and new features

  • plugin mechanism came to replace export. See export_tarball for the replacement of export. Now it should be easy to extend datalad’s interface with custom functionality to be invoked along with other commands.

  • Minimalistic coloring of the results rendering

  • publish/copy_to got progress bar report now and support of --jobs

  • minor fixes and enhancements to crawler (e.g. support of recursive removes)

0.7.0 (Jun 25, 2017) – when it works - it is quite awesome!

New features, refactorings, and bug fixes.

Major refactoring and deprecations

Enhancements and new features

  • siblings can now be used to query and configure a local repository by using the sibling name here

  • siblings can now query and set annex preferred content configuration. This includes wanted (as previously supported in other commands), and now also required

  • New metadata command to interface with datasets/files meta-data

  • Documentation for all commands is now built in a uniform fashion

  • Significant parts of the documentation of been updated

  • Instantiate GitPython’s Repo instances lazily


  • API documentation is now rendered properly as HTML, and is easier to browse by having more compact pages

  • Closed files left open on various occasions (Popen PIPEs, etc)

  • Restored basic (consumer mode of operation) compatibility with Windows OS

0.6.0 (Jun 14, 2017) – German perfectionism

This release includes a huge refactoring to make code base and functionality more robust and flexible

  • outputs from API commands could now be highly customized. See --output-format, --report-status, --report-type, and --report-type options for datalad command.

  • effort was made to refactor code base so that underlying functions behave as generators where possible

  • input paths/arguments analysis was redone for majority of the commands to provide unified behavior

Major refactoring and deprecations

  • add-sibling and rewrite-urls were refactored in favor of new siblings command which should be used for siblings manipulations

  • ‘datalad.api.alwaysrender’ config setting/support is removed in favor of new outputs processing


  • Do not flush manually git index in pre-commit to avoid “Death by the Lock” issue

  • Deployed by publish post-update hook script now should be more robust (tolerate directory names with spaces, etc.)

  • A variety of fixes, see list of pull requests and issues closed for more information

Enhancements and new features

  • new annotate-paths plumbing command to inspect and annotate provided paths. Use --modified to summarize changes between different points in the history

  • new clone plumbing command to provide a subset (install a single dataset from a URL) functionality of install

  • new diff plumbing command

  • new siblings command to list or manipulate siblings

  • new subdatasets command to list subdatasets and their properties

  • drop and remove commands were refactored

  • benchmarks/ collection of Airspeed velocity benchmarks initiated. See reports at

  • crawler would try to download a new url multiple times increasing delay between attempts. Helps to resolve problems with extended crawls of Amazon S3

  • CRCNS crawler pipeline now also fetches and aggregates meta-data for the datasets from datacite

  • overall optimisations to benefit from the aforementioned refactoring and improve user-experience

  • a few stub and not (yet) implemented commands (e.g. move) were removed from the interface

  • Web frontend got proper coloring for the breadcrumbs and some additional caching to speed up interactions. See

  • Small improvements to the online documentation. See e.g. summary of differences between git/git-annex/datalad

0.5.1 (Mar 25, 2017) – cannot stop the progress

A bugfix release


  • add was forcing addition of files to annex regardless of settings in .gitattributes. Now that decision is left to annex by default

  • tools/testing/run_doc_examples used to run doc examples as tests, fixed up to provide status per each example and not fail at once

  • doc/examples

  • progress bars

    • should no longer crash datalad and report correct sizes and speeds

    • should provide progress reports while using Python 3.x

Enhancements and new features

  • doc/examples

    • new example to demonstrate how new super- and sub- datasets were established as a part of our datasets collection

0.5.0 (Mar 20, 2017) – it’s huge

This release includes an avalanche of bug fixes, enhancements, and additions which at large should stay consistent with previous behavior but provide better functioning. Lots of code was refactored to provide more consistent code-base, and some API breakage has happened. Further work is ongoing to standardize output and results reporting (#1350)

Most notable changes

  • requires git-annex >= 6.20161210 (or better even >= 6.20161210 for improved functionality)

  • commands should now operate on paths specified (if any), without causing side-effects on other dirty/staged files

  • save

    • -a is deprecated in favor of -u or --all-updates so only changes known components get saved, and no new files automagically added

    • -S does no longer store the originating dataset in its commit message

  • add

    • can specify commit/save message with -m

  • add-sibling and create-sibling

    • now take the name of the sibling (remote) as a -s (--name) option, not a positional argument

    • --publish-depends to setup publishing data and code to multiple repositories (e.g. github + webserve) should now be functional see this comment

    • got --publish-by-default to specify what refs should be published by default

    • got --annex-wanted, --annex-groupwanted and --annex-group settings which would be used to instruct annex about preferred content. publish then will publish data using those settings if wanted is set.

    • got --inherit option to automagically figure out url/wanted and other git/annex settings for new remote sub-dataset to be constructed

  • publish

    • got --skip-failing refactored into --missing option which could use new feature of create-sibling --inherit


  • More consistent interaction through ssh - all ssh connections go through sshrun shim for a “single point of authentication”, etc.

  • More robust ls operation outside of the datasets

  • A number of fixes for direct and v6 mode of annex

Enhancements and new features

  • New drop and remove commands

  • clean

    • got --what to specify explicitly what cleaning steps to perform and now could be invoked with -r

  • datalad and git-annex-remote* scripts now do not use setuptools entry points mechanism and rely on simple import to shorten start up time

  • Dataset is also now using Flyweight pattern, so the same instance is reused for the same dataset

  • progressbars should not add more empty lines

Internal refactoring

  • Majority of the commands now go through _prep for arguments validation and pre-processing to avoid recursive invocations

0.4.1 (Nov 10, 2016) – CA release

Requires now GitPython >= 2.1.0


  • save

    • to not save staged files if explicit paths were provided

  • improved (but not yet complete) support for direct mode

  • update to not crash if some sub-datasets are not installed

  • do not log calls to git config to avoid leakage of possibly sensitive settings to the logs

Enhancements and new features

  • New rfc822-compliant metadata format

  • save

    • -S to save the change also within all super-datasets

  • add now has progress-bar reporting

  • create-sibling-github to create a :term:sibling of a dataset on github

  • OpenfMRI crawler and datasets were enriched with URLs to separate files where also available from openfmri s3 bucket (if upgrading your datalad datasets, you might need to run git annex enableremote datalad to make them available)

  • various enhancements to log messages

  • web interface

    • populates “install” box first thus making UX better over slower connections

0.4 (Oct 22, 2016) – Paris is waiting

Primarily it is a bugfix release but because of significant refactoring of the install and get implementation, it gets a new minor release.


  • be able to get or install while providing paths while being outside of a dataset

  • remote annex datasets get properly initialized

  • robust detection of outdated git-annex

Enhancements and new features

  • interface changes

    • get --recursion-limit=existing to not recurse into not-installed subdatasets

    • get -n to possibly install sub-datasets without getting any data

    • install --jobs|-J to specify number of parallel jobs for annex get call could use (ATM would not work when data comes from archives)

  • more (unit-)testing

  • documentation: see for basic principles and useful shortcuts in referring to datasets

  • various webface improvements: breadcrumb paths, instructions how to install dataset, show version from the tags, etc.

0.3.1 (Oct 1, 2016) – what a wonderful week

Primarily bugfixes but also a number of enhancements and core refactorings


  • do not build manpages and examples during installation to avoid problems with possibly previously outdated dependencies

  • install can be called on already installed dataset (with -r or -g)

Enhancements and new features

  • complete overhaul of datalad configuration settings handling (see Configuration documentation), so majority of the environment. Now uses git format and stores persistent configuration settings under .datalad/config and local within .git/config variables we have used were renamed to match configuration names

  • create-sibling does not now by default upload web front-end

  • export command with a plug-in interface and tarball plugin to export datasets

  • in Python, .api functions with rendering of results in command line got a _-suffixed sibling, which would render results as well in Python as well (e.g., using search_ instead of search would also render results, not only output them back as Python objects)

  • get

    • --jobs option (passed to annex get) for parallel downloads

    • total and per-download (with git-annex >= 6.20160923) progress bars (note that if content is to be obtained from an archive, no progress will be reported yet)

  • install --reckless mode option

  • search

    • highlights locations and fieldmaps for better readability

    • supports -d^ or -d/// to point to top-most or centrally installed meta-datasets

    • “complete” paths to the datasets are reported now

    • -s option to specify which fields (only) to search

  • various enhancements and small fixes to meta-data handling, ls, custom remotes, code-base formatting, downloaders, etc

  • completely switched to tqdm library (progressbar is no longer used/supported)

0.3 (Sep 23, 2016) – winter is coming

Lots of everything, including but not limited to

0.2.3 (Jun 28, 2016) – busy OHBM

New features and bugfix release

0.2.2 (Jun 20, 2016) – OHBM we are coming!

New feature and bugfix release

  • greatly improved documentation

  • publish command API RFing allows for custom options to annex, and uses –to REMOTE for consistent with annex invocation

  • variety of fixes and enhancements throughout

0.2.1 (Jun 10, 2016)

  • variety of fixes and enhancements throughout

0.2 (May 20, 2016)

Major RFing to switch from relying on rdf to git native submodules etc

0.1 (Oct 14, 2015)

Release primarily focusing on interface functionality including initial publishing