🔆 Executive Summary: HSDS Version 1.0.2

Bug Fixes:

• PUT to an object's attributes collection now returns 400 when the request contains
  no attributes, instead of reporting success without writing anything

• A read that fails on the data node returns an error instead of a 200 with an
  empty or partial body

• A range read that runs past the end of an object keeps the bytes it read
  instead of returning zeros

• Chunks and contiguous datasets are limited to the data node chunk cache size,
  and compact datasets to 64 KiB, so a large chunk can no longer exhaust a data
  node's memory

• The head node refuses to start when TARGET_SN_COUNT or TARGET_DN_COUNT is
  unset or 0, instead of leaving the cluster stuck in WAITING

• Dataset creation no longer rejects filters that give a class but omit id or name

• Link and attribute names containing { or } no longer get a 404

• Looking up a committed datatype by path no longer returns 500

• GET on a linked dataset created by HSDS 0.9.x reports its reference layout
  instead of a plain chunked one

• Reading a few elements of an H5D_CONTIGUOUS_REF dataset no longer fetches the
  whole dataset from its file

Dependencies:

• h5json 2.0.1 or later is required, for the H5D_CONTIGUOUS_REF fix above

Acknowledgements:

We would like to thank the HSDS community members who contributed to this release.

Changelog and release information:

The full list of changes in this release is in
CHANGELOG.md.

What's Changed

• fix: preserve and test reference layouts in GET_Dataset response by @mattjala in #469
• fix: preserve reference layouts in GET_Dataset response by @joaopaulosr95 in #468
• Fix curly brace chars invalidating link/attr names by @mattjala in #471
• Fix unrecognized datatype ids in httpUtil helpers by @mattjala in #472
• Document remaining v1.0.0 breaking changes in README by @mattjala in #474
• Fix/put attributes single body by @mattjala in #478
• Report read errors correctly by @mattjala in #479
• Fix dropped bytes in short range reads by @mattjala in #480
• Bump urllib3 from 2.7.0 to 2.8.0 by @dependabot[bot] in #476
• Bump pyjwt from 2.13.0 to 2.15.0 by @dependabot[bot] in #477
• Explicitly limit chunk size to the data node chunk cache by @mattjala in #481
• Remove unreachable symbolic-link block from getObjectIdByPath by @mattjala in #482
• Refuse to start the head node without target node counts by @mattjala in #483
• Accept filter specs that omit id or name when creating datasets by @mattjala in #484
• Bump version to 1.0.2 by @mattjala in #485

Full Changelog: v1.0.1...v1.0.2

🔆 Executive Summary: HSDS Version 1.0.1

Significant Advancements:

• Fixed issue introduced in 1.0.0 where headless Kubernetes deployments could not serve requests.
  Every Kubernetes manifest HSDS ships runs without a head node, and on those deployments every node had
  been stuck in WAITING for the lifetime of the process, answering 503 to every
  request.

Enhanced Features:

• A new Kubernetes CI job that runs the full integration suite against a real k3s cluster,
  preventing silent breakage of Kubernetes functionality.

Bug Fixes:

• A domain deleted while a bucket scan was queued is no longer partly re-created by that
  scan after garbage collection has swept it

• A PUT to the POSIX backend no longer returns 500 when the key it just wrote is
  removed concurrently

Dependencies:

• aiohttp upgraded from 3.9.4 to 3.14.3

Acknowledgements:

We would like to thank the HSDS community members who contributed to this release.

Changelog and release information:

The full list of changes in this release is in
CHANGELOG.md.

What's Changed

• Reach READY in headless Kubernetes deployments, and test k8s in CI by @mattjala in #460
• fix CI log capture and races by @mattjala in #462
• Bump aiohttp from 3.9.4 to 3.14.3 by @dependabot[bot] in #461
• add a CHANGELOG and generate release notes from it by @mattjala in #463
• Bump version to 1.0.1 by @mattjala in #466

Full Changelog: v1.0.0...v1.0.1

What's Changed

• assign container rank via registration order by @jreadey in #439
• Osmos by @jreadey in #440
• Bump azure-core from 1.35.0 to 1.38.0 by @dependabot[bot] in #443
• Nullreq by @jreadey in #444
• Bump idna from 3.10 to 3.15 by @dependabot[bot] in #445
• feat: json log format and W3C trace context propagation by @joaopaulosr95 in #446
• feat: prometheus metrics endpoint on SN and DN nodes by @joaopaulosr95 in #447
• H5json by @mattjala in #450
• Skip DockerHub login on Dependabot PRs by @mattjala in #448
• Split CI into per-OS HSDS and h5pyd workflows, add status badges by @mattjala in #451
• Remove duplicate test suite execution by @mattjala in #452
• Don't corrupt response stream on GET_Value errors by @mattjala in #449
• Remove race from test data in CI by @mattjala in #454
• Build and publish HSDS to PyPI on release by @mattjala in #453
• Publish to PyPI only after the GitHub release succeeds by @mattjala in #455
• Lift the requests upper bound by @mattjala in #456
• Regenerate requirements.txt from the declared dependencies by @mattjala in #457
• Depend on h5json from PyPI by @mattjala in #458

New Contributors

• @joaopaulosr95 made their first contribution in #446

Full Changelog: v0.9.4...v1.0.0

Update dependencies, remove Python 3.9 support, and update version to 0.9.4.

Python Support:
* Removed support for Python 3.9 in pyproject.toml.
Version Update:
*Updated HSDS_VERSION to 0.9.4 in hsds/basenode.py.
*Updated project version to 0.9.4 in pyproject.toml.
Dependencies:
* Updated numpy to >=2.0.0 in pyproject.toml.
* Replaced requests-unixsocket2 with requests-unixsocket in pyproject.toml.
* Updated requirements.txt with specific versions for all dependencies.

• Replace socket package - #430
• Fix Numpy 2.0 compatibility issues - #427

Updated:

• Improved support for links to contigous HDF5 datasets
• Added support for fancy selection with multiple dimensions
• Numpy 2.0 support
• Codespaces integration
• Support "Where in" queries
• Support Hyperchunks for multidimensional datasets
• Multi-link methods - get/update multiple links in one request
• Multi-attribute methods - get/update multiple attributes in one request
• Support bithshuffle compression
• Added support for Fixed width UTF8 string
• Enable dataset shape reduction
• Support broadcast operations for dataset reads and writes

Plus - various bug fixes

Minor update to fix PATH issue for running HSDS on Windows

Fixes:

• Issue with listing more than 1000 domains in Azure has been resolved (#257)

Features:

• Updated setup script
• Improved support for run hsds app on command line
• Intelligent range get requests (for 1-dimensional datasets)
• Base64 encoding option for fill value
• Enable query by creation order for links and attributes

Features:

• Support for path traversal in getObjectByPath #219
• Added setup instructions for Tencent Cloud
• Add ChunkInitializer support
• Removed RangegetProxy node

Fixes:

• Fix GET datatype including bucket in domain #223
• Fix arg order in HTTPRequestEntityTooLarge #206
• Window compatibility update #209
• Fix for k8sutil #209
• Fix for high cpu usage on k8s - #199