Resolwe

==================

Changed
-------
- Update description text of ``cutadapt-star-htseq`` descriptor schema to
better describe the difference between gene/transcript-type analyses
- Speed-up management command for inserting mappings


==================

==================

Added
-----
- Add R, tabix, and CheMut R library to ``resolwebio/dnaseq`` Docker image
- Add ``SRA Toolkit`` to ``resolwebio/rnaseq`` Docker image

Changed
-------
- Require Resolwe 6.x
- Extend pathway map with species and source field
- Move template and logo for multi-sample report into ``resolwebio/latex``
Docker image
- Refactor ``amplicon-report`` process to contain all relevant inputs for
``amplicon-archive-multi-report``
- Refactor ``amplicon-archive-multi-report``
- Use ``resolwebio/dnaseq:1.2.0`` Docker image in ``filtering_chemut`` process

Fixed
-----
- Enable DEBUG setting in tests using Django's ``LiveServerTestCase``
- Wait for ElasticSeach to index the data in ``KBBioProcessTestCase``
- Remove unused parameters in TopHat (2.0.13) process and Chip-seq workflow


==================

==================

Changed
-------
- Remove runtime directory during general data purge instead of immediately
after each process finishes
- Only process the Data object (and its children) for which the dispatcher's
``communicate()`` was triggered
- Propagate logging messages from executors to the listener
- Use process' slug instead of data id when logging errors in listener
- Improve log messages in dispatcher

Added
-----
- Add ``descriptor_completed`` field to the ``Entity`` filter
- Validate manager semaphors after each test case, to ease debugging of tests
which execute processes

Fixed
-----
- Don't set Data object's status to error if executor is run multiple times to
mitigate the Celery issue of tasks being run multiple times
- Make management commands respect the set verbosity level


==================

==================

Fixed
-----
- Make manager more robust to ORM/database failures during data object
processing
- Rebuild the ElasticSearch index after permission is removed from an object
- Trim ``Data.process_error``, ``Data.process_warning`` and
``Data.process_info`` fields before saving them
- Make sure values in ``Data.process_error``, ``Data.process_warning`` and
``Data.process_info`` cannot be overwritten
- Handle missing ``Data`` objects in ``hydrate_input_references`` function
- Make executor fail early when executed twice on the same data directory


==================

==================

Changed
-------
- **BACKWARD INCOMPATIBLE:** ``FLOW_DOCKER_LIMIT_DEFAULTS`` has been renamed
to ``FLOW_PROCESS_RESOURCE_DEFAULTS`` and ``FLOW_DOCKER_LIMIT_OVERRIDES``
has been renamed to ``FLOW_PROCESS_RESOURCE_OVERRIDES``
- **BACKWARD INCOMPATIBLE:** ``Process.PERSISTENCE_TEMP`` is not used for
execution priority anymore
- **BACKWARD INCOMPATIBLE:** There is only one available manager class, which
includes dispatch logic; custom manager support has been removed and their
role subsumed into the new connector system
- **BACKWARD INCOMPATIBLE:** Removed ``FLOW_DOCKER_MAPPINGS`` in favor of new
``FLOW_DOCKER_VOLUME_EXTRA_OPTIONS`` and ``FLOW_DOCKER_EXTRA_VOLUMES``
- Parent relations are kept even after the parent is deleted and are deleted
when the child is deleted
- Dependency resolver in manager is sped up by use of parent relations
- Validation of ``Data`` inputs is performed on save instead of on create

Added
-----
- Support for the SLURM workload manager
- Support for dispatching ``Data`` objects to different managers
- Support for passing secrets to processes in a controlled way using a newly
defined ``basic:secret`` input type
- ``is_testing`` test helper function, which returns ``True`` when invoked in
tests and ``False`` otherwise
- Add ``collecttools`` Django command for collecting tools' files in single
location defined in ``FLOW_TOOLS_ROOT`` Django setting which is used for
mapping tools in executor when ``DEBUG`` is set to ``False`` (but not in
tests)

Fixed
-----
- Fix ``Data`` object preparation race condition in ``communicate()``
- Set correct executor in flow manager
- Make executors more robust to unhandled failures
- Calculate ``Data.size`` by summing ``total_size`` of all file-type outputs
- Don't change slug explicitly defined by user - raise an error instead
- Objects are locked while updated over API, so concurrent operations don't
override each other
- Make manager more robust to unhandled failures during data object processing
- Fix manager deadlock during tests
- Fix ctypes cache clear during tests
- Don't raise ``ChannelFull`` error in manager's communicate call
- Don't trim predefined slugs in ``ResolweSlugField``


==================

==================

Added
-----
- Database-side JSON projections for ``Storage`` models
- Compute total size (including refs size) for file-type outputs
- Add ``size`` field to ``Data`` model and migrate all existing objects

Change
------
- Change Test Runner's test directory creation so it always creates a
subdirectory in ``FLOW_EXECUTOR``'s ``DATA_DIR``, ``UPLOAD_DIR`` and
``RUNTIME_DIR`` directories

Fixed
-----
- Do not report additional failure when testing a tagged process errors or
fails
- Fix Test Runner's ``changes-only`` mode when used together with a Git
repository in detached ``HEAD`` state
- Fix handling of tags and test labels together in Test Runner's
``changes-only`` mode
- Fix parallel test execution where more test processes than databases were
created during tests

==================