Fix: Add missing python-dateutil dependency

Contents

Dependencies

• Add missing python-dateutil dependency

Tweak semantic version incrementing rules

Contents

Operations

• Make non-feature/breaking changes require a patch version increase
• Disable major version increments while package is in beta

Enable continuous deployment with semantic versions

Contents

Operations

• Run release workflow on merge of any branch into main
• Add Conventional Commits pre-commit hook
• Replace check-version-consistency job with check-semantic-version job, which checks that the version in setup.py is the same as the semantic version expected by git-mkver given the Conventional Commits since the last tag
• Add the update-pull-request workflow that auto-generates part of the PR description on each commit
• Run publish test job on all branches and make it dependent on check-semantic-version job passing
• Rename tests job to run-tests

Release/0.1.19

Contents

New Features

• Make Datafiles and Datasets labelable
• Use new version of tags in all Taggables
• Replace string tags in a TagSet with key-value pairs in a TagDict
• Add new Taggable mixin for providing the new tags interface
• Add FilterDict, allowing filtering of key-value pairs by their values
• Allow nested attribute/dictionary filtering in the filter containers FilterSet, FilterList and FilterDict
• Allow any number of filters to be specified when filtering in filter containers
• Allow ignoring of filterables missing the filtered-for attribute in a filter container instead of raising an error
• Add filter container one method?
• Allow ordering by nested attributes in all FilterContainers
• Allow gs:// paths to be used in Datafile, Dataset, and Manifest
• Allow gs:// paths to be used in storage client
• Add datetime filters
• Add in-range filters to str, datetime, and Number filters

Breaking changes

• Use new format for manifests' datasets in twine.json files
• Convert old Taggable mixin to Labelable mixin
• Convert old Tag class to Label class
• Convert TagSet to LabelSet
• Use key-value pairs for filter names and values when filtering Filterables
• Stop logging in Serialisable
• Always exclude logger field in Serialisable
• Simplify tag name pattern to ^[a-z0-9][a-z0-9_]*(?<!_)$
• Simplify label pattern to ^[a-z0-9][a-z0-9-]*(?<!-)$
• Store tags as key-value pairs in GCS custom metadata
• Unbase TagDict and LabelSet from filter containers
• JSON-encode cloud storage custom metadata again
• Store tags in tags field of cloud metadata again
• Close #165: prefix GCS custom metadata fields with "octue__"

Minor improvements

• Remove filters field from manifest strand in twines
• Allow tags to be added via kwargs in Taggable.add_tags
• Remove unused _FILTERSET_ATTRIBUTE class variables
• Base Label on str
• Support non-English characters in case-insensitive filtering
• Add octue-sdk-python version to datafile metadata
• Base filter containers on new FilterContainer abstract class
• Move filter and order methods into FilterContainer
• Use OctueJSONDecoder in Serialisable and GoogleCloudStorageClient
• Add de/serialisation of datetime objects to de/encoders
• Clarify name of some GoogleCloudStorageClient methods
• Add set and UserString encoding to OctueJSONEncoder
• Use OctueJSONDecoder
• Add set and datetime decoding to OctueJSONDecoder
• Remove unnecessary methods from LabelSet
• Rename add_labels method and add add method to LabelSet
• Automatically generate complementary (not) filters from other filters
• Remove a line of duplicated code in Datafile

Fixes

• Handle timestamps from cloud with/without timezone information
• Fix OctueJSONDecoder
• Make it harder to add invalid labels to LabelSet

Dependencies

• Use new version of twined that distinguishes tags from labels

Testing

• Use latest GCS emulator
• Only run deployment test if RUN_DEPLOYMENT_TESTS envvar is True

Quality Checklist

• New features are fully tested (No matter how much Coverage Karma you have)
• [v0.2 onward] New features are included in the documentation

Release/0.1.18

Contents

New features

• Allow decimal points in tags

Minor improvements

• Close #162: make timestamp an optional parameter for Datafile

Quality Checklist

• New features are fully tested (No matter how much Coverage Karma you have)

Release/0.1.17

Contents

New Features

• Allow Datafile to be used as a context manager for changes to local datafiles
• Allow Datafile.from_cloud to be used as a context manager for changes to cloud datafiles
• Allow Datafile to remember where in the cloud it came from
• Add the following methods to Datafile:
  • get_cloud_metadata
  • update_cloud_metadata
  • clear_from_file_cache
  • _get_cloud_location
  • _store_cloud_location
  • _check_for_attribute_conflict
• Avoid re-uploading Datafile file or metadata if they haven't changed
• Raise error if implicit cloud location is missing from Datafile
• Add GoogleCloudStorageClient.update_metadata method
• Allow option to not update cloud metadata in Datafile cloud methods
• Allow tags to contain capitals and forward slashes (but not start or end in a forward slash)
• Allow datetime and posix timestamps for Datafile.timestamp
• Add Datafile.posix_timestamp property

Breaking changes

• Close #148: remove hash_value from Datafile GCS metadata
• When hashing Datafiles, only hash represented file (i.e. stop hashing metadata)
• When hashing Datasets and Manifests, only hash the files contained (i.e. stop hashing metadata)
• Make hash of Hashable instance with _ATTRIBUTES_TO_HASH=None the empty string hash value "AAAAAA=="

Minor improvements

• Simplify output of GoogleCloudStorageClient.get_metadata
• Make Hashable instances re-calculate their hash_value every time unless an immutable_hash_value is explicitly provided (e.g. for cloud datafiles where you don't have the file locally to hash)
• Add private Identifiable._set_id method
• Close #147: pull metadata gathering for Datafile into method
• Get datetime objects directly from GCS blob instead of parsing string serialisations
• Add time utils module
• Add hash preparation function to Hashable for datetime instances
• Use the empty string hash value for Datafile if GCS crc32c metadata isn't present
• Stop serialising hash value of Manifest, Dataset, and Datafile

Fixes

• Close #146: Stop serialising GCS metadata as JSON. This avoids strings in the metadata appearing in two sets of quotation marks on Google Cloud Storage. This is a breaking change for any files already persisted with JSON-encoded metadata.
• Remove ability to set custom hash value via kwargs when using Datafile.from_cloud

Testing

• Factor out cloud datafile creation in datafile tests

Quality Checklist

• New features are fully tested (No matter how much Coverage Karma you have)

Release/0.1.16

Contents

Breaking changes

• Rename Service.__init__ parameter id to service_id to avoid built-in name clash
• Move deployment package into cloud package

Dependencies

• Use newest version of twined to support python>=3.6

Minor improvements

• Remove duplicate code and unnecessary comments from Runner
• Raise error if SERVICE_ID envvar is missing from deployment environment
• Disallow non-None empty values as Service IDs
• Add base class for service backends; update docstrings

Fixes

• Use OctueJSONEncoder in JSON serialisation inside Service.answer to ensure numpy arrays are serialised

Testing

• Add tests for Topic and Subscription
• Add extra test for Service
• Shorten runtime of cli.start test

Release/0.1.15

Contents

Fixes

• Add from_string option to Serialisable.deserialise

Testing

• Mock Google Pub/Sub Service, Topic, Subscription, Publisher and Subscriber in tests
• Remove unneeded cleanup code from Service tests

Release/0.1.14

Contents

Breaking changes

• Remove TagSet.__str__

Fixes

• Use TagSet to deserialise tags in Datafile.from_cloud
• Add custom (de)serialise methods to TagSet
• Return subtags of a Tag in order using a FilterList
• Remove separate dependencies copy/cache steps in Google Cloud Run Dockerfile so that it works for older versions of docker

Minor improvements

• Remove absolute path from Dataset and Manifest serialisation
• Add Serialisable.deserialise method
• Add filter method to TagSet to avoid e.g. taggable.tags.tags.filter

Operations

• Improve description of release workflow

Release/0.1.13

Contents

New features

• Support setup.py and requirements-dev.txt in Cloud Run Dockerfile
• Retrieve credentials from Google Cloud Secret Manager and inject into environment in Runner.run
• Add ability to retrieve and update cloud files via the Datafile.download or Datafile.open methods
• Allow cloud file attributes to be updated via Datafile.to_cloud method
• Allow instantiation of TagSets from JSON-encoded lists

Breaking changes

• Raise error if the datasets of the input manifest passed to Service.ask aren't all cloud-based

Fixes

• Fix Dataset construction from serialised form in Manifest
• Fix Datafile construction from serialised form in Dataset
• Fix Datafile.deserialise
• Adjust usages of tempfile.NamedTemporaryFile to also work on Windows
• Add timeout and retry to Service.answer
• Add retry to Service.wait_for_answer
• Add 60 second timeout for answering question in Cloud Run deployment
• Use correct environment variable for service ID in Cloud Run Dockerfile
• Set _last_modified, size_bytes, and _hash_value to null values if a Datafile representing a cloud file is instantiated for a hypothetical cloud location (i.e. not synced to a cloud file at that point in time)
• Allow Dataset.get_file_sequence use with no filter

Dependencies

• Use new twined version that supports validation of credentials strand
• Use newest version of gcp-storage-emulator

Minor improvements

• Make path a positional argument of Datafile
• Move gunicorn requirement into octue requirements
• Raise warning instead of error if Google Cloud credentials environment variable is not found and return None as credentials
• Move cloud code into new cloud subpackage
• Raise TimeoutError in Service.wait_for_answer if no response is received by end of retries
• Only look for deployment_configuration.json file in docker container /app directory
• Ensure deployment_configuration.json file is always loaded correctly in docker container
• Pass credentials strand into Runner instance in Cloud Run deployment
• Add name attribute to Identifiable mixin
• Add Google Cloud metadata to Datafile serialisation
• Add deserialise method to Datafile
• Add ability to add metadata to a Datafile instantiated from a regular cloud file
• Use CRC32C hash value from Google Cloud when instantiating a Datafile from the cloud
• Add ability to name Datafiles
• Add ability to check whether a Datafile, all Datafiles in a Dataset, or all Datasets in a Manifest are located in Google Cloud
• Use Datafile.deserialise when instantiating a Dataset from a dictionary
• Add representation to GCPPubSubBackend
• Load credentials strand JSON in Runner initialisation
• Add location searched to message of error raised when app module can't be found in Runner.run
• Ignore E203 flake8 warning

Testing

• Remove subjective Service test test_serve_with_timeout
• Use temporary file rather than temporary directory for tests where possible
• Test Dataset.deserialise

Quality Checklist

• New features are fully tested (No matter how much Coverage Karma you have)

Coverage Karma

• If your PR decreases test coverage, do you feel you have built enough Coverage Karma* to justify it?