0.8.0

🚀 Enhancement

• Add rudimentary duecredit support using zenodo's dandi-cli DOI #285 (@yarikoptic)

🐛 Bug Fix

• BF: add h5py.version into the list of tokens for caching #284 (@yarikoptic)
• change from disease to disorder #291 (@satra)

🏠 Internal

• Fix publish-schemata workflow #290 (@jwodder)
• updated just models #287 (@satra)
• Add workflow for publishing model schemata to dandi/schema #276 (@jwodder)
• DOC: strip away duplicate with the handbook information #279 (@yarikoptic)

Authors: 3

• John T. Wodder II (@jwodder)
• Satrajit Ghosh (@satra)
• Yaroslav Halchenko (@yarikoptic)

0.7.2

🐛 Bug Fix

• Support h5py 3.0 #275 (@jwodder)
• Include item path in "Multiple files found for item" message #271 (@jwodder)
• Copy files with cp --reflink=auto where supported #269 (@jwodder)
• Make keyring lookup more flexible #267 (@jwodder)

🏠 Internal

• Add healthchecks for the Postgres and minio Docker containers #272 (@jwodder)

Authors: 1

• John T. Wodder II (@jwodder)

0.7.1

🐛 Bug Fix

• Use oldest file when race condition causes multiple files per item #265 (@jwodder)

🏠 Internal

• Set up workflow with auto for releasing & PyPI uploads #257 (@jwodder)

📝 Documentation

• Remove unused link from CHANGELOG.md #266 (@jwodder)

Authors: 1

• John T. Wodder II (@jwodder)

[0.7.0] - 2020-11-04

Added

• Files are now stored in temporary directories while downloading alongside metadata for use in resuming interrupted downloads

Changed

• Moved code for navigating Dandi Archive into new dandiarchive submodule
• YAML output now sorts keys
• dandiset.yaml is no longer uploaded to the archive
• Restrict h5py dependency to pre-v3.0

[0.6.4] - 2020-09-04

Primarily a range of bugfixes to ensure correct operation with current state
of other components of DANDI, and use of the client on Windows OS.

Added
- Initial DANDI schema files
- More tests for various code paths
- `download`: new option `--download [assets,dandiset.yaml,all]`

Fixed
- `download` - account for changes in DANDI API (relevant only for released
  datasets, of which we do not have any "real" ones yet)
- `upload` - various Windows specific fixes

Note: [0.6.3] was released under missing some of the fixes, so overall
abandoned.

[0.6.2] - 2020-08-19

Fixed
- `organize` treatment of paths on window (gh-204)

[0.6.1] - 2020-08-18

Changed

- CLI modules RF to avoid circular imports
- `pytest` default traceback style is short and shows 10 slowest tsts

Fixed

- `download` of draft datasets from Windows (gh-202)
- `upload` and other tests to account for new web UI

0.6.0

A variety of improvements and bug fixes, with major changes toward su…

0.6.0-rc1: This a Release Candidate which is based on a good number of changes

which are in `master` branch and might still be released as 0.5.1
release.  Those changes are yet to be documented.

Added
- `download` now can download from "published" (versioned) dandisets.
  The entire `download` was refactored and new UI also uses pyout (as
  `upload` and `ls`) so there will be no tqdm progress bar indicators.

[0.5.0] - 2020-06-04

Added
- `metadata` and `organize`: extract and use `probe_ids` metadata to
  disambiguate (if needed)
- `organize`: `--devel-debug` option to perform metadata extraction serially
- `upload`:
  - `--allow-any-path` development option to allow upload of DANDI
    not yet 'unsupported' file types/paths
  - compute 4 digests (all are checksums ATM): md5, sha1, sha256, sha512
    and upload as a part of the metadata record
- `download`:
  - use the "fastest" available digest (sha1) to validate correctness of the
    download
  - follow redirections from arbitrary redirector (e.g., bit.ly). Succeeds
    only if the final URL is known to DANDI.

Fixed
- `upload`: a crash while issuing a record to update about deleted empty item

Refactored
- `organize`: disambiguation process now could use a flexible list of metadata
  fields (ATM only `probe_ids` and `obj_id`)
- `download`: handling of redirection - now uses `HEAD` request instead of `GET`