Releases: TileDB-Inc/TileDB-Cloud-Py
v0.14.4
What's Changed
- tiledb-cloud is incompatible with urllib3 >= 2.6.0 and requires a version < 2.6.0 (gh-733)
- add batch resources to query by @spirpinias in (gh-732)
Full Changelog: v0.14.2...v0.14.4
v0.14.3
What's Changed
- add batch resources to query by @spirpinias in #732
Full Changelog: v0.14.2...v0.14.3
v0.14.2
What's Changed
- manual override for group_fragments by @spirpinias in #731
New Contributors
- @spirpinias made their first contribution in #731
Full Changelog: v0.14.1...v0.14.2
v0.14.1
What's Changed
- Add support for passing resources for all tasks in VCF Ingestion by @Shelnutt2 in #730
Full Changelog: v0.14.0...v0.14.1
v0.14.0
What's Changed
- Update schedule for tiledb-cloud-py.yaml by @sgillies in #694
- Add a udf mark, to skip tests that can't pass without a 3.12 image by @sgillies in #695
- Exclude tiledb-py versions 0.33.0-0.33.2 by @sgillies in #696
- Add tiledb main branch to CI matrix by @sgillies in #697
- Remove dest_config from kwargs by @ktsitsi in #699
- Update runners to ubuntu-24.04 by @sgillies in #700
- CI updates and noise reduction by @sgillies in #702
- Exclude more unpredictable tests by @sgillies in #703
- Fix error when passing
output_array_nametotiledb.cloud.sql.execby @nickvigilante in #704 - Expose get_group_contents() via high level API by @sgillies in #705
- Raise ValueError when asset.info's uri doesn't resolve by @sgillies in #706
- Provide a default name for arrays when registering by @sgillies in #707
- Avoid udf name contention between different CI jobs by @sgillies in #708
- Remove upper limit for tblib requirement spec by @sgillies in #711
- Register DAG by @spencerseale in #692
- Permit null concat with non-null in distributed vcf query by @spencerseale in #712
- Give tasks and registered datasets unique names by @sgillies in #713
- Skip one more wheel test by @sgillies in #715
- Summarize XPASS-ing tests in CI by @sgillies in #714
- Fix upload and restore of test durations by @sgillies in #716
- Workflow register, run, and resume by @gspowley in #698
- Allow exclude_metadata to be callable by @ktsitsi in #710
- Resolve sc-49443 for workaround on member-list performance by @johnkerl in #721
- Set VCF record count to 0 when the index is bad by @alancleary in #724
- Updated VCF multiple sample check to split on comma by @alancleary in #725
- Add a dag_name parameter to vcf.read(...) by @alancleary in #726
- Add consolidation and manifest resource parameters for VCF ingestion by @Shelnutt2 in #729
New Contributors
- @alancleary made their first contribution in #724
Full Changelog: v0.13.0...v0.14.0
v0.13.0
General availability of cloudpickle 2.2 vendoring
Vendoring cloudpickle 2.2 for internal use allows a user of this package to install any version of cloudpickle that they like. Cloudpickle version 3 can now be used with, for example, Dask in delayed functions.
What's Changed
- Fill in missing API docs by @nickvigilante in #687
- Rename vcf.ingestionq to vcf.ingestion by @nickvigilante in #688
- Vendor cloudpickle by @sgillies in #686
- Fix credentials discrepancy between Bioimaging/VCF and SOMA registration UDFs by @sgillies in #685
- Register SOMA dataset using the carried credentials by @sgillies in #689
- Use fnmatch for SOMA pattern matching by @sgillies in #690
Full Changelog: v0.12.31...v0.13.0
v0.13rc1
Vendoring cloudpickle 2.2
Vendoring cloudpickle 2.2 for internal use allows cloudpickle version 3 to be used with, for example, Dask in delayed functions.
What's Changed
- Fill in missing API docs by @nickvigilante in #687
- Rename vcf.ingestionq to vcf.ingestion by @nickvigilante in #688
- Vendor cloudpickle by @sgillies in #686
- Fix credentials discrepancy between Bioimaging/VCF and SOMA registration UDFs by @sgillies in #685
Full Changelog: v0.12.31...v0.13rc1
v0.12.31
Run VCF queries as TileDB batch task graphs
This option enables the tiledb.cloud.vcf.read distributed query to use batch task graphs (as opposed to realtime), which is slower but more robust.
What's Changed
Full Changelog: v0.12.30...v0.12.31
v0.12.30
New read/write/delete organization role, SOMA ingestion improvements, and better user experience for the tiledb namespace
Version 0.12.30 allows use of the TileDB platform's new "read_write_delete" (permission to read, write, and delete assets) organization role. It's one step less privileged than the existing "admin" role.
The tiledb.cloud.soma.ingest module has been refactored and a defect in registering multiple SOMA groups has been fixed. The cloud storage location of output groups has changed slightly. All groups are now output to OUTPUT_URI/NAME, where OUTPUT_URI is the given parameter, like s3://bucket/prefix/, and NAME is the base/stem of an H5ad file.
Lastly, tiledb is a namespace package, and in your environment may include modules from tiledb, tiledb.vector_search and tiledb.cloud distributions. Previously, importing tiledb or tiledb.vector_search in an environment where tiledb.cloud is also installed would trigger a warning about your need to login, even if you don't have any intent to use tiledb.cloud methods. This warning no longer happens needlessly and only shows up when appropriate.
What's Changed
- Re-generate generated API to add new r/w/d organization role. by @NullHypothesis in #644
- fix obs and var defaults to return all values if none spec'd by @spencerseale in #676
- Move test requirements, adjust and streamline CI YAML by @sgillies in #674
- Register ingested groups using the correct URI by @sgillies in #679
- Defer loading of cloud.json file by @sgillies in #681
- typ{o,e} fixes by @ryan-williams in #680
- Patch config._config, not config.config by @sgillies in #682
- Expose deadline argument for DAG by @ktsitsi in #678
Full Changelog: v0.12.29...v0.12.30
v0.12.29
Bio-imaging and SOMA ingestion improvements
Finer control over the resources used to run UDFs during Bio-imaging and SOMA data ingestion, and full support for ingesting folders of multiple H5ad files are highlights of this release.
The Bio-imaging, SOMA, and VCF ingestion entry points now all take an ingest_resources argument. Its value is the same as the existing resources argument. The new ingest_resources argument isn't consumed by task graph construction and is propagated to the most resource-intensive tasks of ingestion graphs.
The previous version of this package allowed ingestion of single H5ad files. Now, folders of multiple files can be ingested by passing the URI of the folder and a pattern argument, such as *.h5ad, that matches only H5ad files in the folder.
What's Changed
- Adding PNG support and enabling experimental reader in kwargs by @ktsitsi in #654
- Add Panel dashboard wrapper for Voila by @gspowley in #667
- Add a tag name check for releases and some words about releases by @sgillies in #668
- Propagate resources to ingest_h5ad node by @sgillies in #669
- Skip sub-folders and make pattern matching more useful by @sgillies in #670
- Add ingest_resources for use with UI by @sgillies in #671
- Remove value registration wrappers by @sgillies in #672
- Tests have been xpass-ing, no longer need xfail marks by @sgillies in #673
Full Changelog: v0.12.28...v0.12.29