Change Log

All notable changes to this project are documented in this file. This project adheres to Semantic Versioning.

Unreleased

Added

• Add BackgroundTask model to enable support for observable long running backend tasks

• Add BackgroundTaskConsumer class to handle starting background jobs

Changed

• BACKWARD INCOMPATIBLE Remove the duplicated collection subfield from the entity field when serializing Data objects

• BACKWARD INCOMPATIBLE: Replace duplicate methods on Data, Entity and

Collection objects with the call to the corresponding background duplicate job.

• BACKWARD INCOMPATIBLE: The duplicate endpoints on Data, Entity and Collection return the details of the BackgroundTask instance that is used to track the duplication status

• When subscribing to an object for notifications the CREATE notifications are also sent along. This is necessary to receive CREATE notifications triggered from the data inside the container on the container.

• Remove the redundant prefetch in the BaseCollectionViewSet

• Annotate the queryset in the BaseCollectionViewSet with data_statuses, entity_count and data_count attributes to reduce the number of queries

• Remove the redundant prefetch in the EntityViewSet

• Annotate the queryset in the EntityViewSet with data_statuses and and data_count attributes to reduce the number of queries

• Modify the DataViewSet to remove redundant queries

• Add duplicate_data, duplicate_entity and duplicate_collection methods to BackgroundTaskConsumer class

• Set is_active to True for all available processes every time register management command is run even if the same version of the process already exists in the database

Fixed

• Send observer notifications when data object is automatically assigned to the entity via entity_type in the process

• Remove observer from subscription if object is deleted

• Handle empty iterable in observe_instance_container method

• Remove call to close in observers consumer that caused RuntimeError

33.0.0 - 2023-03-13

Added

• Add suppress_observer_notifications and move_to_container decorators to observers

• Add types ChannelsMessage and WebsocketMessage describing type of observers messages

• Add property containers on the PermissionObject that returns all the containers for the given instance

• Add observe_instance_containers method to observer models

Changed

• BACKWARD INCOMPATIBLE: Drop support for Python 3.6, Python 3.7, Python 3.8 and Python 3.9. The processing container continues to support Python >= 3.6.

• Add source attribute to observer notifications

• When subscribing to observer endpoint without explicit ids subscribe to CREATE and DELETE notifications by default

• Send CREATE and DELETE notifications to the containers of the given object when permissions to it are gained/lost or it is moved between containers. The notifications also contain the source and source_id context containing content type and id of the object that triggered the change.

32.1.0 - 2023-02-13

Added

• Add new entity annotation framework

• Add set_permission method on PermissionQuerySet

• Add notify_create method on Subscription model in observers to enable sending notification when object is created

• Allow ordering data endpoint by entity name

• Add with_superusers argument to users_with_permission method

Changed

• When slug collision occurs in the listener when creating new objects retry up to ten times before raising the exception

• Enable overriding user and group id of the processing container with environmental variable

Fixed

• Notify user when object (data, entity) is created in the container

• Do not delete observers in use by other subscriptions when unsubscribing

• Notify subscribers to the collection if object inside them is modified

• Notify superusers without explicit permissions when object is created

32.0.0 - 2022-11-14

Changed

• BACKWARD INCOMPATIBLE: Drop support for Python 3.4 in the processing container

• BACKWARD INCOMPATIBLE: Rewrite the listener to allow more than one of them to run at the same time

• Rewrite the processing container code to make it more stable

• Rewrite the commnication container code to make it more stable

• Add resolve_url call to the re_import.sh bash script

Changed

• Migrate base docker images to Ubuntu 22.04 and Fedora 36

31.3.1 - 2022-10-19

Fix

• Make sure signals are triggered for data objects when their sample is moved to a new collection

31.3.0 - 2022-10-17

Added

• Add handler resolve_url to listener

• Add support for resolving urls in Python processes

• Add url method to the BaseConnector class and override it in S3 and local connectors

31.2.1 - 2022-09-20

Changed

• Add cleanup_callback to retry decorator

• Clean kubernetes temporary files with credentials on config load error

31.2.0 - 2022-09-19

Added

• Add descriptor and descriptor_schema fields to the Relation model

• Setting KUBERNETES_DISPATCHER_CONFIG_LOCATION specifying the location of the kubernetes config to load in the Kubernetes workload connector

Changed

• Retry loading Kubernetes configuration couple of times before giving up in the kubernetes workload connector

• Add model observers that notify clients about model changes via a websocket connection

31.1.0 - 2022-08-23

Fixed

• Clear custom_messages array on auditlog reset call

31.0.0 - 2022-07-18

Changed

• BACKWARD INCOMPATIBLE: Always try to load kubernetes configuration via load_kube_config befofe falling back to load_incluster_config

• Rename docker-compose.yml to compose.yaml

• Listener can always modify the data object that it is processing

• Add original objects to post_duplicate call

• Install asgiref version based on the version of the installed Python interpreter

Fixed

• Fix a typo in listener permission handling when creating data model: a check was performed on the wrong object type

Added

• Make requests and limits for the communication container configurable

30.3.0 - 2022-06-13

Added

• Support custom user model in Python processes

30.2.0 - 2022-05-16

Added

• Send custom signal post_duplicate when objects are duplicated since regular Django signals are not sent

• Add auditlog application to log user actions

30.1.0 - 2022-04-15

Added

• Custom signal resolwe.flow.signals.before_processing is sent when data object is ready for processing

• Support setting descriptor and DescriptorSchema on Data during process runtime

• Support filtering Data, Entity and Collections based on permissions (view, edit, owner, group, shared_with_me)

• Support filtering Data and Entity objects by relation id

• Create upload_config API endpoint that specifies upload connector type and credentials client can use for optimized upload

Fixed

• Do not return multiple version of the same process while checking for permissions in Python processes

• Change misleading error message when importing file if the response with status code indicating error was received from the upstream server

Changed

• Use data_id instead of data.id when notifying dispatcher to avoid potential database query inside async context

30.0.0 - 2022-03-14

Added

• Add support for Python 3.10

• Add health checks support for deploy in Kubernetes

• Add collecttools_kubernetes management command

• Add COMMUNICATION_CONTAINER_LISTENER_CONNECTION to separate settings for listener (where to bind to) and containers (where to connect to)

• Support docker_volume setting in connector config

• Support use of named volumes as processing or input volume in Docker executor

• Support SSL connection to Redis

Changed

• BACKWARD INCOMPATIBLE: Require Django 3.2

• BACKWARD INCOMPATIBLE: Require Django Priority Batch version 4

• Do not prepare tools configmaps in Kubernetes workload connector

• Enable Docker containers to connect to the custom network

• Auto-delete completed jobs in Kubernetes after 5 minutes

• Optionally add affinity to the Kubernetes job

• Remove support for setting permissions using old syntax

29.3.0 - 2022-02-15

Added

• Add MD5 checksum to DataBrowseView view

29.2.0 - 2022-01-17

Added

• Support range parameter in fields of Python proces

Changed

• Flush stdout/stderr on Python processes on every write

Fixed

• Add attribute as a field to a resolwe.process.fields.GroupField in Python process only if it is an instance of resolwe.process.fields.Field.

29.1.0 - 2021-12-12

Changed

• Do not fail in case of missing files in UriResolverView

Fixed

• Remove references to temporary export files from the database and make sure they are not created anymore

• Wrap move_to_collection in transaction and only call method if collection has changed

29.0.0 - 2021-11-11

Added

• BACKWARD INCOMPATIBLE: New permission architecture: it is not based on Guardian anymore. The main benefits of new architecture are speed gains in common operations, such as setting a permission and retrieving objects with the given permission.

• Allow overriding process resources in data object

Changed

• Allow mounting connectors into pods as persistent volume claim instead of volume of type hostPath

Fixed

• use the same connector inside pod to handle files and directories

• When data object was deleted listener did not receive the terminate message and pod did not terminate immediatelly

28.5.0 - 2021-09-13

Added

• Add compare_models_and_csv management script to check if all ``ReferencedPath``s point to a valid file in the aws database

• Add method get_latest to Process class in Python Processes returning the latest version of the process with the given slug

• Support assuming role in S3 connector

Changed

• Set hashes during upload to avoid creating multiple versions of the object in S3 bucket with enabled versioning

28.4.0 - 2021-08-16

Changed

• Remove dependency on EFS/NFS when running on Kubernetes

• When running on Kubernetes the runtime volume configuration can be omitted

28.3.0 - 2021-07-20

Fixed

• Prepare release 28.3.0 due to preexisting 28.3.0a1 pre-release

28.2.1 - 2021-07-13

Fixed

• Speed up deleting storage locations by considering only referenced paths belonging to the given storage location

• Temporary pin asteval to version 0.9.23 due to compatibility issues with Python 3.6

Changed

• Improve logging in cleanup manager

28.2.0 - 2021-06-15

Fixed

• Create ReferencedPath objects during transfer only when needed

Changed

• Retry data transfer if botocore.exceptions.ClientError is raised during transfer

Added

• Add FLOW_PROCESS_MAX_MEM Django setting to limit the ammount of memory used by a process

• Support disabled fields in Python processes

• Add method get_latest to the Process class in Python processes which retrieves latest version of the process with the given slug

28.1.0 - 2021-05-17

Fixed

• Do not raise exception when terminating runlistener management command

• Change concurrency issue in the listener causing processes to sometitimes get incorrect value for RUNTIME_VOLUME_MAPS settings

Changed

• Make S3 connectors use system credentials when they are not explicitely given in settings

Added

• Make it possible to rewrite container image names in kubernetes workload connector

28.0.4 - 2021-05-04

Fixed

• Use per process storage overrides

28.0.3 - 2021-05-04

Changed

• Make gp2 default EBS storage class

28.0.2 - 2021-05-03

Fixed

• Log peer activity on every received message to avoid declaring otherwise healthy node as failed

• Fix possible data loss caused by parallel command processing when uploading log files interfered with processing command from a script

• Fix deadlock when uploading empty files

28.0.1 - 2021-04-28

Changed

• Make logger level inside init and communication containers configurable via environmental variable

• Change default logger level inside init and communication containers for AWS S3 and Google Cloud Storage components to WARNING

Fixed

• Stop timer that uploads log files in the processing container immediatelly after the processing is finished to avoid timing issues that could cause the data object to be marked as failed

28.0.0 - 2021-04-19

Changed

• BACKWARD INCOMPATIBLE: Require Django 3.1.x

• BACKWARD INCOMPATIBLE: Require Django Channels version 3.0.x

• BACKWARD INCOMPATIBLE: Require asgiref version 3.3.x

• BACKWARD INCOMPATIBLE: Require jsonschema version 3.2.x

• BACKWARD INCOMPATIBLE: Require Sphinx version 3.5.x

• BACKWARD INCOMPATIBLE: Require django-guardian version 2.3.x

• Refresh versions of the other dependencies

• Replace django-versionfield2 with django-versionfield

• Overhaul of the storage configuration

• Remove per-process runtime directory

• Increase socket timeouts in the processing and the communication container

Added

• Add multipart upload capability to LocalFilesystemConnector and AwsS3Connector

• Support uploading files to LocalFilesystemConnector or AwsS3Connector

• Add support for Python 3.9

27.1.2 - 2021-03-22

Fixed

• Bump version of upload-dir process to use the default version of processing image instead of the previous one

• Use Signature Version 4 when generating presigned URLs in S3 connector

• Fix possible socket timeout when uploading files in the processing container

• Remove static rnaseq image from list of docker images

Changed

• Use tagged base image in upload-file process

• Allow to change Entity descriptor in Python process.

27.1.1 - 2021-03-21

Fixed

• Fix connection timeout in communication container when sending initial message to the listener

27.1.0 - 2021-03-15

Fixed

• Account for file system overhead when processing Data objects with large inputs

Changed

• Improve storage manager to only process applicable storage locations instead of iterating through all of them

• Skip hash computation when connector itself provides data integrity check

• Remove job prefix from kubernetes job name

• Make error messages in Python processes more useful

Added

• Add label job_type to Kubernetes job to separate interactive jobs from batch jobs

27.0.0 - 2021-02-22

Fixed

• Fixed progress reporting in Python processes

• Do not override content-type of S3 object when storing hashes

• Support upload of files larger than 80G to AWS S3

Changed

• Download input data in init container

• Storage objects are sent to the listener over socket instead of using files on the shared filesystem

• Make it possible to run the platform without shared filesystem. All inputs for processed data object are prepared in input container and all outputs are uploaded to the chosen storage connector when they are referenced.

• Overcommit CPU in kubertenes processing container by 20%

• Move docker images from Docker Hub to Amazon ECR

Added

• Make automatic removal of Docker containers configurable

• Terminate processing immediately when data object is deleted

• Make default processing image configurable

26.0.0 - 2021-01-20

Changed

• BACKWARD INCOMPATIBLE: Remove resolwe/upload-tab-file and resolwe/archiver Docker images

• BACKWARD INCOMPATIBLE: Remove obsolete processes: archiver, upload-tab-file and upload-image-file

• BACKWARD INCOMPATIBLE: Python process syntax has changed: all the attributes of Data object are now available in Python process and therefore accessing outputs using syntax data_object.output_name is no longer valid since output_name could be the name of the attribute. The new syntax is data_object.output.output_name.

• BACKWARD INCOMPATIBLE: Communication between the processing script and listener has changed from printing to stdout to sending messages over sockets. Messages printed to stdout or sent using old version of the resolwe-runtime-utils (YAML processes) are no longer processed. YAML processes need new version of resolwe-runtime-utils while Python processes require a rewrite to the new syntax and Python version 3.4 or higher in the container (resolwe-runtime-utils package is no longer needed).

• Use Github Actions to run the tests

• Listener communicates with containers through ZeroMQ instead of Redis

• Start two containers for each process instead of one: the second one is used to communicate with the listener

• Move settings for Python processes from files to environmental variables

Added

• Add Kubernetes workload connector

• Support running process instant termination

• Support registering custom command handlers in listener and exposing data objects (possibly defined in other Django applications) to Python processes

• Support Django-like syntax in Python processes to create, filter or access attributes of the exposed data objects

• Support creating new base classes for Python processes

25.2.0 - 2020-12-15

Fixed

• Allow retrieval of Storage object that was linked to more than one Data object

Changed

• Migrate docker images to Fedora 33 and Ubuntu 20.04

25.1.0 - 2020-11-16

Added

• Support Python processes in Sphinx autoprocess*:: directive

25.0.0 - 2020-10-16

Added

• BACKWARD INCOMPATIBLE: Only copy parent relations when duplicating Data objects

• Add duplicate data dependency to indicate from which object the Data object was duplicated

• Support accessing Data name in Python processes through self.name

• Add permission filter to collection, entity and data that returns only objects on which current user has given permission

Changed

• Make relations in collection visible to public user if he has view permissions on the collection

24.0.0 - 2020-09-14

Changed

• BACKWARD INCOMPATIBLE: Terminate Python process immediately after self.error method is called inside the process

Fixed

• Make sure to terminate Docker container before executor exits

• Speed-up duplication of Data objects, Enteties, and Collections

• Lock inputs’ storage locations while the process is waiting and processing to make sure that they are not deleted

• Don’t validate input objects when Data object is marked as done as they may already be deleted at that point

23.0.0 - 2020-08-17

Fixed

• Fix ordering options in Data viewset to enable ordering by process__name and process__type

• Handle exception when processing deleted Data object

Changed

• BACKWARD INCOMPATIBLE: Delete elastic application

• Don’t pass undefined values to steps of workflows

Added

• Add relations property to data: and list:data: fields to support relations on client

• Add entity_id property do DataField in Python processes

• Add relations in Python processes

22.1.3 - 2020-07-13

Fixed

• When deciding which StorageLocation objects will be deleted consider only completed StorageLocation objects.

• Add google.resumable_media.common.DataCorruption exception to transfer_exceptions tuple.

22.1.2 - 2020-06-30

Fixed

• Celery sometimes starts more than one worker for a given Data object. In such case the download and purge part of the worker must be skipped or errors processing Data objects might occur.

22.1.1 - 2020-06-16

Changed

• Remove asgiref version pin due to new release that fixed previous regression.

22.1.0 - 2020-06-15

Changed

• Rename transfer_rec to transfer_objects and change its signature to accept dictionary objects with information about name, size and hashes of objects to transfer

• Move part of Data object validation to listener

• Improve loading time of collection endpoint

Added

• Add move_to_collection method to Data viewset

• Report registration failure in ProcessTestCase

• Add a pseudo Python process to serve as a template

• Add validate_urls method to storage BaseConnector class

• Validate storage connector settings on registraton

• Add transfer_data method to StorageLocation class

• Remove data when StorageLocation object is deleted

• Store file hashes inside ReferencedPath model and connect it to StorageLocation model

• Add get_hashes method to storage connectors

• Add open_stream method to storage connectors

• Add compute_hashes function to storage.connectors.hasher module

• Use threads when transfering files with AwsS3Connector

• Add duplicate method to storage connectors

• Add pre-processing and post-processing hooks to storage connectors

• Use multiple threads for file transfer

Fixed

• Add missing decorator validate_url to AwsS3Connector

• Always import exceptions from requests library

• Fix bug that sometimes caused objects inside workflow to fail with Failed to transfer data.

• Fix dependency handling bug in listener: when checking for missing data listener must only consider depencies with kind KIND_IO instead of all depencies.

• Raise exception when data transfer failed.

22.0.0 - 2020-05-18

Changed

• BACKWARD INCOMPATIBLE: Move purge code inside worker, remove old purge code

• Various code fixes to make code work with the new storage model

• Use storage connectors in workers to download data not available locally

Added

• Add resolwe.storage application, a framework for storage management

• Add storage connectors for Google Cloud Storage, Amazon Simple Storage service and local filesystem.

• Add migrations to move from old storage model to the new one

• Add storage manager

• Add management command to start storage manager

• Add cleanup manager for removing unreferenced data

• Add isnull related lookup filter

• Add entity_count to the Collection serializer

• Add inherit_collection to Data viewset

• Add entity_always_create in Process serializer

21.1.0 - 2020-04-14

Added

• Add support for the allow_custom_choice field property in Python processes

• Add ordering by contributor’s first and last name to Collection and Data viewsets

• Add data_count and status fields to the Collection serializer

Fixed

• Enable all top-level class definitions in Python processes

• Make filtering via foreign key more ‘Django like’: when foreign key does not exist return empty set instead of raising validation exception. Also when filtering using list of foreign keys do not raise validation exception if some foreign keys in the list do not exist.

• Reduce number of database queries in API viewsets by prefetching all required data

21.0.0 - 2020-03-16

Changed

• BACKWARD INCOMPATIBLE: Use Postgres filtering instead of Elasticsearch on API endpoints

• BACKWARD INCOMPATIBLE: Remove filtering by year, month, day, hour, minute and second on API endpoints

• Migrate docker images to Fedora 31

• Use DictRelatedField for collection field in RelationSerializer. In practice this fixes inconsistency comparing with how other serializers handle collections field.

Added

• Add a custom database connector to optimize queries and enable them to use database indexes

• Add database indexes to improve search performance

• Add database fields and triggers for full-text search in Postgres

• Add support for annotating entities in processes

• Add support for Python 3.8

20.2.0 - 2020-02-17

Added

• Support workflows as inputs to Python processes

• Support retrieval of Data.name in Python process

• Add name_contains, contributor_name, and owners_name collection and data filtering fields on API

• Add username to current_user_permissions field of objects on API

• Add delete_chunked method to Collection, Entity and Storage managers

Fixed

• Delete orphane Storage objects in chunks in purge command to prevent running out of memory

20.1.0 - 2019-12-16

Added

• Add description field to Collection full-text search

20.0.0 - 2019-11-18

Changed

• BACKWARD INCOMPATIBLE: Remove download permission from Data objects, samples and collections and add permission from samples and collections

• BACKWARD INCOMPATIBLE: Remove Entity.descriptor_completed field

Fixed

• Fix Docker executor command with --cpus limit option. This solves the issue where process is killed before the timeout 30s is reached

19.1.0 - 2019-09-17

Added

• Support filtering by process_slug in DataViewSet

Fixed

• Fix DictRelatedField so it can be used in browsable-API

• Fix access to subfields of empty GroupField in Python processes

19.0.0 - 2019-08-20

Changed

• BACKWARD INCOMPATIBLE: Change relations between Data, Entity and Collection from ManyToMany to ManyToOne. In practice this means that Data.entity, Data.colllecton and Entity.collection are now ForeignKey-s. This also implies the following changes:

  • CollectionViewSet methods add_data and remove_data are removed

  • EntityViewset methods add_data,``remove_data``, add_to_collection and remove_from_collection are removed

  • EntityQuerySet and Entity method duplicate argument inherit_collections is renamed to inherit_collection.

  • EntityFilter FilterSet field collections is renamed to collection.

• BACKWARD INCOMPATIBLE: Change following fields in DataSerializer:

  • process_slug, process_name, process_type, process_input_schema, process_output_schema are removed and moved in process field which is now DictRelatedField that uses ProcessSerializer for representation

  • Remove entity_names and collection_names fields

  • add entity and colection fields which are DictRelatedField-s that use corresponding serializers for representation

  • Remove support for hydrate_entities and hydrate_collections query parameters

• BACKWARD INCOMPATIBLE: Remove data field in EntitySerializer and CollectionSerializer. This implies that parameter hydrate_data is no longer supported.

• BACKWARD INCOMPATIBLE: Remove delete_content paremeter in delete method of EntityViewset and CollectionViewSet. From now on, when Entity/Collection is deleted, all it’s objects are removed as well

• Gather all Data creation logic into DataQuerySet.create method

Added

• Enable sharing based on user email

• Support running tests with live Resolwe host on non-linux platforms

• Add inherit_entity and inherit_collection arguments to Data.duplicate and DataQuerySet.duplicate method

• Implement DictRelatedField

18.0.0 - 2019-07-15

Changed

• BACKWARD INCOMPATIBLE: Remove parents and children query filters from Data API endpoint

Added

• /api/data/:id/parents and /api/data/:id/children API endpoints for listing parents and children Data objects of the object with given id

• Add entity_always_create field to Process model

Fixed

• Make sure that Elasticsearch index exists before executing a search query

17.0.0 - 2019-06-17

Changed

• BACKWARD INCOMPATIBLE: Use Elasticsearch version 6.x

• BACKWARD INCOMPATIBLE: Bump Django requirement to version 2.2

• BACKWARD INCOMPATIBLE: Remove not used django-mathfilters requirement

Added

• Support Python 3.7

• Support forward and reverse many-to-one relations in Elasticsearch

• Add collection_names field to DataSerializer

• Add test methods to ProcessTestCase that assert directory structure and content: assertDirExists, assertDir, and assertDirStructure

• Add upload-dir process

16.0.1 - 2019-04-29

Fixed

• Pin django-priority-batch to version 1.1 to fix compatibility issues

16.0.0 - 2019-04-16

Changed

• BACKWARD INCOMPATIBLE: Access to DataField members (in Python process input) changed from dict to Python objects. For example, input_field.file_field['name'] changed to input_field.file_field.path.

• BACKWARD INCOMPATIBLE: Filters that are based on django-filter FilterSet now use dict-declaring-syntax. This requires that subclasses of respective filters modify their syntax too.

• Interactively save results in Python processes

Added

• Add get_data_id_by_slug method to Python processes’ Process class

• Python process syntax enhancements:

  • Support .entity_name in data inputs

  • Easy access to process resources through self.resources

• Raise error if ViewSet receives invalid filter parameter(s)

• Report process error for exceptions in Python processes

• Report process error if spawning fails

• Automatically export files for spawned processes (in Python process syntax)

• Import files of Python process FileField inputs (usage: inputs.src.import_file())

Fixed

• Interactively write to standard output within Python processes

• Fix writing to integer and float output fields

• Allow non-required DataField as Python process input

15.0.1 - 2019-03-19

Fixed

• Fix storage migration to use less memory

15.0.0 - 2019-03-19

Changed

• Log plumbum commands to standard output

• Change storage data relation from many-to-one to many-to-many

• Moved purged field from Data to DataLocation model

Added

• Add run_process method to Process to support triggering of a new process from the running Python process

• Add DataLocation model and pair it with Data model to handle data location

• Add entity_names field to DataSerializer

• Support duplication of Data, Entity and Collection

• Support moving entities between collections

• Support relations requirement in process syntax

14.4.0 - 2019-03-07

Changed

• Purge processes only not jet purged Data objects

Fixed

• Allow references to missing Data objects in the output of finished Data objects, as we don’t have the control over what (and when) is deleted

14.3.0 - 2019-02-19

Added

• Add scheduled field to Data objects to store the date when object was dispatched to the scheduling system

• Add purge field to Data model that indicates whether Data object was processed by purge

Fixed

• Make Elasticsearch build arguments cache thread-safe and namespace cache keys to make sure they don’t interfere

• Trigger the purge outside of the transaction, to make sure the Data object is commited in the database when purge worger grabs it

14.2.0 - 2019-01-28

Added

• Add input Jinja filter to access input fields

14.1.0 - 2019-01-17

Added

• Add assertFilesExist method to ProcessTestCase

• Add clean_test_dir management command that removes files created during testing

Fixed

• Support registration of Python processes inherited from process.Process

• Skip docker image pull if image exists locally. This solves the issue where pull would fail if process uses an image that is only used locally.

14.0.1 - 2018-12-17

Fixed

• Make sure that tmp dir exists in Docker executor

14.0.0 - 2018-12-17

Changed

• BACKWARD INCOMPATIBLE: Run data purge in a separate worker to make sure that listener replies to the executor within 60 seconds

• Use batcher for spawned processes in listener

• Increase Docker’s memory limit for 100MB to make sure processes are not killed when using all available memory and tune Docker memory limits to avoid OOM.

Added

• Raise an exception in Docker executor if container doesn’t start for 60 seconds

• Set TMPDIR environment variable in Docker executor to .tmp dir in data directory to prevent filling up container’s local storage

Fixed

• Process SIGTERM signal in executor as expected - set the Data status to error and set the process_error field

• Clear cached Django settings from the manager’s shared state on startup

13.3.0 - 2018-11-20

Changed

• Switch channels_redis dependency to upstream version

Added

• Python execution engine

• Support multiple entity types

• Support extending viewsets with custom filter methods

• Add tags attribute to ProcessTestCase.run_process method which adds listed tag to the created Data object

• Copy Data objects tags from parent objects for spawned Data objects and Data objects created by workflows

Fixed

• Fix manager shutdown in the test runner. If an unrecoverable exception occurred while running a test, and never got caught (e.g. an unpicklable exception in a parallel test worker), the listener would not get terminated properly, leading to a hang.

• Data and collection name API filters were fixed to work as expected (ngrams was switched to raw).

13.2.0 - 2018-10-23

Added

• Use prioritized batcher in listener

13.1.0 - 2018-10-19

Added

• Use batching for ES index builds

Fixed

• Fix handling of M2M dependencies in ES indexer

13.0.0 - 2018-10-10

Changed

• BACKWARD INCOMPATIBLE: Remove Data descriptors from Entity Elasticsearch index

• Support searching by slug and descriptor_data in entity viewset text search

Added

• Add tags to collections

12.0.0 - 2018-09-18

Changed

• BACKWARD INCOMPATIBLE: Switch Collection and Entity API viewsets to use Elasticsearch

• BACKWARD INCOMPATIBLE: Refactor Relation model, which includes:

  • renaming position to partition

  • renaming label to category and making it required

  • adding unit

  • making collection field required

  • requiring unique combination of collection and category

  • renaming partition’s position to label

  • adding (integer) position to partition (used for sorting)

  • deleting Relation when the last Entity is removed

• BACKWARD INCOMPATIBLE: Remove rarely used parameters of the register command --path and --schemas.

• Omit current_user_permissions field in serialization if only a subset of fields is requested

• Allow slug to be null on update to enable slug auto-generation

• Retire obsolete processes. We have added the is_active field to the Process model. The field is read-only on the API and can only be changed through Django ORM. Inactive processes can not be executed. The register command was extended with the --retire flag that removes old process versions which do not have associated data. Then it finds the processes that have been registered but do not exist in the code anymore, and:

  • If they do not have data: removes them

  • If they have data: flags them not active (is_active=False)

Added

• Add support for URLs in basic:file: fields in Django tests

• Add collections and entities fields to Data serializer, with optional hydration using hydrate_collections and/or hydrate_entities

• Support importing large files from Google Drive in re-import

• Add python3-plumbum package to resolwe/base:ubuntu-18.04 image

Fixed

• Prevent mutation of input_ parameter in ProcessTestCase.run_process

• Return 400 instead of 500 error when slug already exists

• Add trailing colon to process category default

• Increase stdout buffer size in the Docker executor

11.0.0 - 2018-08-13

Changed

• BACKWARD INCOMPATIBLE: Remove option to list all objects on Storage API endpoint

• Make the main executor non-blocking by using Python asyncio

• Debug logs are not send from executors to the listener anymore to limit the amount of traffic on Redis

Added

• Add size to Data serializer

• Implement ResolweSlugRelatedField. As a result, DescriptorSchema objects can only be referenced by slug (instead of id)

• Add options to filter by type and scheduling_class on Process API endpoint

Fixed

• Inherit collections from Entity when adding Data object to it

10.1.0 - 2018-07-16

Changed

• Lower the level of all INFO logs in elastic app to DEBUG

Added

• Add load tracking to the listener with log messages on overload

• Add job partition selection in the SLURM workload connector

• Add slug Jinja filter

• Set Data status to ERROR if executor is killed by the scheduling system

Fixed

• Include the manager in the documentation, make sure all references work and tidy the content up a bit

10.0.1 - 2018-07-07

Changed

• Convert the listener to use asyncio

• Switched to channels_redis_persist temporarily to mitigate connection storms

Fixed

• Attempt to reconnect to Redis in the listener in case of connection errors

10.0.0 - 2018-06-19

Changed

• BACKWARD INCOMPATIBLE: Drop support for Python 3.4 and 3.5

• BACKWARD INCOMPATIBLE: Start using Channels 2.x

Added

• Add the options to skip creating of fresh mapping after dropping ES indices with elastic_purge management command

• Add dirname and relative_path Jinja filters

9.0.0 - 2018-05-15

Changed

• Make sorting by contributor case insensitive in Elasticsearch endpoints

• Delete ES documents in post delete signal instead of pre delete one

Added

• BACKWARD INCOMPATIBLE: Add on-register validation of default values in process and schemas

• BACKWARD INCOMPATIBLE: Validate that field names in processes and schemas start with a letter and only contain alpha-numeric characters

• Support Python 3.6

• Add range parameter and related validation to fields of type basic:integer:, basic:decimal, list:basic:integer: and list:basic:decimal

• Support filtering and sorting by process_type parameter on Data API endpoint

• Add dirname Jinja filter

• Add relative_path Jinja filter

Fixed

• Add missing list:basic:decimal type to JSON schema

• Don’t crash on empty in lookup

• Fix {{ requirements.resources.* }} variables in processes to take in to the account overrides specified in Django settings

• Create Elasticsearch mapping even if there is no document to push

8.0.0 - 2018-04-11

Changed

• BACKWARD INCOMPATIBLE: Use Elasticsearch version 5.x

• BACKWARD INCOMPATIBLE: Raise an error if an invalid query argument is used in Elasticsearch viewsets

• BACKWARD INCOMPATIBLE: Switch Data API viewset to use Elasticsearch

• Terminate the executor if listener response with error message

• verbosity setting is no longer propagated to the executor

• Only create Elasticsearch mappings on first push

Added

• Add sort argument to assertFile and assertFiles methods in ProcessTestCase to sort file lines before asserting the content

• Add process_slug field to DataSerializer

• Improve log messages in executor and workload connectors

• Add process_memory and process_cores fields to Data model and DataSerializer

• Support lookup expressions (lt, lte, gt, gte, in, exact) in ES viewsets

• Support for easier dynamic composition of type extensions

• Add elastic_mapping management command

Fixed

• Fix Elasticsearch index rebuilding after a dependant object is deleted

• Send response to executor even if data object was already deleted

• Correctly handle reverse m2m relations when processing ES index dependencies

7.0.0 - 2018-03-12

Changed

• BACKWARD INCOMPATIBLE: Remove Ubuntu 17.04 base Docker image due to end of lifetime

• BACKWARD INCOMPATIBLE: Remove support for Jinja in DescriptorSchema’s default values

• BACKWARD INCOMPATIBLE: Remove CONTAINER_IMAGE configuration option from the Docker executor; if no container image is specified for a process using the Docker executor, the same pre-defined default image is used (currently this is resolwe/base:ubuntu-16.04)

• Add mechanism to change test database name from the environment, appending a _test suffix to it; this replaces the static name used before

Added

• Add Ubuntu 17.10 and Ubuntu 18.04 base Docker images

• Add database migration operations for process schema migrations

• Add delete_chunked method to Data objects queryset which is needed due to Django’s extreme memory usage when deleting a large count of Data objects

• Add validate_process_types utility function, which checks that all registered processes conform to their supertypes

• Add FLOW_CONTAINER_VALIDATE_IMAGE setting which can be used to validate container image names

• Only pull Docker images at most once per process in list_docker_images

• Add FLOW_PROCESS_MAX_CORES Django setting to limit the number of CPU cores used by a process

Fixed

• Make parallel test suite worker threads clean up after initialization failures

• Add mechanism to override the manager’s control channel prefix from the environment

• Fix deletion of a Data objects which belongs to more than one Entity

• Hydrate paths in refs of basic:file:, list:basic:file:, basic:dir: and list:basic:dir: fields before processing Data object

6.1.0 - 2018-02-21

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

6.0.1 - 2018-01-29

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

6.0.0 - 2018-01-17

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

5.1.0 - 2017-12-12

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

5.0.0 - 2017-11-28

Changed

• BACKWARD INCOMPATIBLE: The keep_data() method in TransactionTestCase is no longer supported. Use the --keep-data option to the test runner instead.

• BACKWARD INCOMPATIBLE: Convert the manager to Django Channels

• BACKWARD INCOMPATIBLE: Refactor executors into standalone programs

• BACKWARD INCOMPATIBLE: Drop Python 2 support, require Python 3.4+

• Move common test environment preparation to TestCaseHelpers mixin

Fixed

• Fix parents/children filter on Data objects

• Correctly handle removed processes in the changes-only mode of the Resolwe test runner

4.0.0 - 2017-10-25

Added

• BACKWARD INCOMPATIBLE: Add option to build only subset of specified queryset in Elasticsearch index builder

• --pull option to the list_docker_images management command

• Test profiling and process tagging

• New test runner, which supports partial test suite execution based on changed files

• Add all and any Jinja filters

Changed

• BACKWARD INCOMPATIBLE: Bump Django requirement to version 1.11.x

• BACKWARD INCOMPATIBLE: Make ProcessTestCase non-transactional

• Pull Docker images after process registration is complete

• Generalize Jinja filters to accept lists of Data objects

• When new Data object is created, permissions are copied from collections and entity to which it belongs

Fixed

• Close schema (YAML) files after register command stops using them

• Close schema files used for validating JSON schemas after they are no longer used

• Close stdout used to retrieve process results in executor after the process is finished

• Remove unrelated permissions occasionally listed among group permissions on permissions endpoint

• Fix ResolwePermissionsMixin to work correctly with multi-words model names, i.e. DescriptorSchema

• Fix incorrect handling of offset/limit in Elasticsearch viewsets

3.1.0 - 2017-10-05

Added

• resolwe/base Docker image based on Ubuntu 17.04

• Support different dependency kinds between data objects

Fixed

• Serialize current_user_permissions field in a way that is compatible with DRF 3.6.4+

• API requests on single object endpoints are allowed to all users if object has appropriate public permissions

3.0.1 - 2017-09-15

Fixed

• Correctly relabel SELinux contexts on user/group files

3.0.0 - 2017-09-13

Added

• Add filtering by id on descriptor_schma API endpoint

• Support assigning descriptor schema by id (if set value is of type int) on Collection, Data and Entity endpoints

• assertAlmostEqualGeneric test case helper, which enables recursive comparison for almost equality of floats in nested containers

Changed

• BACKWARD INCOMPATIBLE: Run Docker containers as non-root user

Fixed

• Use per-process upload dir in tests to avoid race conditions

2.0.0 - 2017-08-24

Added

• descriptor jinja filter to get the descriptor (or part of it) in processes

• Ubuntu 14.04/16.04 based Docker images for Resolwe

• Add list_docker_images management command that lists all Docker images required by registered processes in either plain text or YAML

• Data status is set to ERROR and error message is appended to process_error if value of basic:storage: field is set to a file with invalid JSON

Changed

• BACKWARD INCOMPATIBLE: Quote all unsafe strings when evaluating expressions in Bash execution engine

• BACKWARD INCOMPATIBILE: Rename permissions attribute on API endpoints to current_user_permissions

• API permissions endpoint raises error if no owner is assigned to the object after applied changes

• owner permission cannot be assigned to a group

• Objects with public permissions are included in list API views for logged-in users

• Owner permission is assigned to the contributor of the processes and descriptor schemas in the register management command

• The base image Dockerfile is renamed to Dockerfile.fedora-26

Fixed

• Add basic:url:link field to the JSON schema

• Return more descriptive error if non-existing permission is sent to the permissions endpoint

• Handle errors occurred while processing Elasticsearch indices and log them

• Return 400 error with a descriptive message if permissions on API are assigned to a non-existing user/group

1.5.1 - 2017-07-20

Changed

• Add more descriptive message if user has no permission to add Data object to the collection when the object is created

Fixed

• Set contributor of Data object to public user if it is created by not authenticated user

• Remove remaining references to calling pip with --process-dependency-links argument

1.5.0 - 2017-07-04

Added

• Add Resolwe test framework

• Add with_custom_executor and with_resolwe_host test decorators

• Add isort linter to check order of imports

• Support basic test case based on Django’s TransactionTestCase

• Support ES test case based on Django’s TransactionTestCase

• Support process test case based on Resolwe’s TransactionTestCase

• Add ability to set a custom command for the Docker executor via the FLOW_DOCKER_COMMAND setting.

• get_url jinja filter

• When running register management command, permissions are automatically granted based on the permissions of previous latest version of the process or descriptor schema.

• Set parent relation in spawned Data objects and workflows

• Relations between entities

• Resolwe toolkit Docker images

• Archive file process

• File upload processes

• Resolwe process tests

• Add SET_ENV setting to set environment variables in executor

• Support ordering by version for descriptor schema

• Add NullExecutor

• If choices are defined in JSON schema, value of field is validated with them

• Add cpu core, memory and network resource limits

• Add scheduling class for processes (interactive, batch), which replaces the previously unused process priority field

• Add share_content flag to the collection and entity permissions endpoint to also share the content of the coresponding object

• Add delete_content flag to the collection and entity destroy method on API to also delete the content of the coresponding object

Changed

• Support running tests in parallel

• Split flow.models module to multiple files

• Remove ability to set a custom executor command for any executor via the FLOW_EXECUTOR['COMMAND'] setting.

• Rename RESOLWE_API_HOST setting and environment variable in executor to RESOLWE_HOST_URL

• Remove keep_failed function in tests.

• Rename keep_all function to keep_data.

• Manager is automatically run when new Data object is created

• Outputs of Data objects with status Error are not validated

• Superusers are no longer included in response in permissions endpoint of resources

• Remove public_processes field from the Collection model as it is never used

• Public users can create new Data objects with processes and descriptor schemas on which they have appropriate permissions

• Add custom ResolweSlugField and use it instead of django-autoslug

Fixed

• SECURITY: Prevent normal users from creating new Processes over API

• Configure parallel tests

• Isolate Elasticsearch indices for parallel tests

• Fix Docker container name for parallel tests

• Generate temporary names for upload files in tests

• Fix permissions in Elasticsearch tests

• Do not purge data in tests

• Remove primary keys before using cached schemas’ in process tests

• Set appropriate SELinux labels when mounting tools in Docker containers

• Data objects created by the workflow inherit its permissions

• If user doesn’t have permissions on the latest versions of processes and descriptor schemas, older ones are used or error is returned

• Support data: and list:data: types

• Set Data object status to error if worker cannot update the object in the database

• Data objects returned in CollectionViewset and EntityViewset are filtered by permissions of the user in request

• Public permissions are taken into account in elastic app

• Treat None field value as if the field is missing

• Copy parent’s permissions to spawned Data objects

1.4.1 - 2017-01-27

Fixed

• Update instructions on preparing a release to no longer build the wheel distribution which currently fails to install Resolwe’s dependency links

1.4.0 - 2017-01-26

Added

• Auto-process style, type tree and category index

• Support loading JSON from a file if the string passed to the basic:json: field is a file.

• list:basic:integer: field

• Data object’s checksum is automatically calculated on save

• get_or_create end point for Data objects

• basic:file:html: field for HTML files

• Helper function for comparing JSON fields in tests

• Purge directories not belonging to any data objects

• Ordering options to API endpoints

• Workflow execution engine

• data_by_slug filter for jinja expression engine

• Export RESOLWE_API_HOST environment variable in executor

• Add check_installed() test utility function

• Add support for configuring the network mode of Docker executor

• Add with_docker_executor test utility decorator

• Support for Docker image requirements

• Support version in descriptor schema YAML files

• Add Entity model that allows grouping of Data objects

• Introduce priority of Data objects

• Data objects created with processes with temporary persistence are given high priority.

• Add resolwe.elastic application, a framework for advanced indexing of Django models with ElasticSearch

Changed

• Refactor linters, check PEP 8 and PEP 257

• Split expression engines into expression engines and execution engines

• Use Jinja2 instead of Django Template syntax

• Expression engine must be declared in requirements

• Set Docker Compose’s project name to resolwe to avoid name clashes

• Expose check_docker() test utility function

• Update versionfield to 0.5.0

• Support Django 1.10 and update filters

• Executor is no longer serialized

• Put Data objects with high priority into hipri Celery queue.

Fixed

• Fix pylint warnings (PEP 8)

• Fix pydocstyle warnings (PEP 257)

• Take last version of process for spawned objects

• Use default values for descriptor fields that are not given

• Improve handling of validation errors

• Ignore file size in assertFields

• Order data objects in CollectionViewSet

• Fix tests for Django 1.10

• Add quotes to paths in a test process test-save-file

1.3.1 - 2016-07-27

Added

• Sphinx extension autoprocess for automatic process documentation

1.3.0 - 2016-07-27

Added

• Ability to pass certain information to the process running in the container via environment variables (currently, user’s uid and gid)

• Explicitly set working directory inside the container to the mapped directory of the current Data’s directory

• Allow overriding any FLOW_EXECUTOR setting for testing

• Support GET request on /api/<model>/<id>/permissons/ url

• Add OWNER permissions

• Validate JSON fields before saving Data object

• Add basic:dir field

• RESOLWE_CUSTOM_TOOLS_PATHS setting to support custom paths for tools directories

• Add test coverage and track it with Codecov

• Implement data purge

• Add process_fields.name custom tamplate tag

• Return contributor information together with objects

• Added permissions filter to determine Storage permissions based on referenced Data object

Changed

• Move filters to separate file and systemize them

• Unify file loading in tests

• Simplify ProcessTestCase by removing the logic for handling different uid/gid of the user running inside the Docker container

• Upgrade to django-guardian 1.4.2

• Rename FLOW_EXECUTOR['DATA_PATH'] setting to FLOW_EXECUTOR['DATA_DIR']

• Rename FLOW_EXECUTOR['UPLOAD_PATH'] setting to FLOW_EXECUTOR['UPLOAD_DIR']

• Rename proc.data_path system variable to proc.data_dir

• Rename test project’s data and upload directories to .test_data and .test_upload

• Serve permissions in new format

• Rename assertFiles method in ProcessTestCase to assertFile and add new assertFiles method to check list:basic:file field

• Make flow.tests.run_process function also handle file paths

• Use Travis CI to run the tests

• Include all necessary files for running the tests in source distribution

• Exclude tests from built/installed version of the package

• Put packaging tests in a separate Tox testing environment

• Put linters (pylint, pep8) into a separate Tox testing environment

• Drop django-jenkins package since we no longer use Jenkins for CI

• Move testing utilities from resolwe.flow.tests to resolwe.flow.utils.test and from resolwe.permissions.tests.base to resolwe.permissions.utils.test

• Add Tox testing environment for building documentation

• Extend Reference documentation

Fixed

• Spawn processors (add data to current collection)

• Set collection name to avoid warnings in test output

• Improve Python 3 compatibility

• Fix setting descriptor schema on create

1.2.1 - 2016-05-15

Added

• Add docker-compose configuration for PostgreSQL

• Processes can be created on API

• Enable spawned processes

Changed

• Move logic from Collection model to the BaseCollection abstract model and make it its parent

• Remove all logic for handling flow_collection

• Change default database user and port in test project’s settings

• Keep track of upload files created during tests and purge them afterwards

Fixed

• Test processes location agnostic

• Test ignore timezone support

1.2.0 - 2016-05-06

Changed

• Rename assertFileExist to assertFileExists

• Drop --process-dependency-links from Tox’s pip configuration

• Improve documentation on preparing a new release

Added

• Ability to use a custom executor command by specifying the FLOW_EXECUTOR['COMMAND'] setting

• Make workload manager configurable in settings

Fixed

• Make Resolwe work with Python 3 again

• Fix tests

• Render data name again after inputs are resolved

• Ensure Tox installs the package from sdist

• Pass all Resolwe’s environment variables to Tox’s testing environment

• Ensure tests gracefully handle unavailability of Docker

1.1.0 - 2016-04-18

Changed

• Rename process_register manage.py command to register

• Reference process by slug when creating new Data object

• Run manager when new Data object is created through API

• Include full DescriptorSchema object when hydrating Data and Collection objects

• Add djangorestframework-filters package instead of django-filters

Added

• Tox tests for ensuring high-quality Python packaging

• Timezone support in executors

• Generating slugs with django-autoslug package

• Auto-generate Data name on creation based on template defined in Process

• Added endpoint for adding/removeing Data objects to/from Collection

Fixed

• Pass all Resolwe’s environment variables to Tox’s testing environment

• Include all source files and supplementary package data in sdist

• Make Celery engine work

• Add all permissions to creator of flow_collection Colection

• Set DescriptorSchema on creating Data objects and Collections

• Loading DescriptorSchema in tests

• Handle Exceptions if input field doesn’t match input schema

• Trigger ORM signals on Data status updates

• Don’t set status od Data object to error status if return code of tool is 0

1.0.0 - 2016-03-31

Changed

• Renamed Project to Collection

• Register processes from packages and custom paths

• Removed support for Python 3.3

Added

• Permissions

• API for flow

• Docker executor

• Expression engine support

• Celery engine

• Purge command

• Framework for testing processors

• Processor finders

• Support for Django 1.9

• Support for Python 3.5

• Initial migrations

• Introductory documentation

0.9.0 - 2015-04-09

Added

Initial release.