Changelog

The format is based on Keep a Changelog.

Versioning currently follows X.Y.Z semantic versioning. Historically, before 1.14.0, a different versioning scheme was used.

Minor breaking changes are still possible in 1.X.Y but we try to announce them with DeprecationWarnings in the previous Y-1 release.

[Unreleased]

[1.22.0] - 2026-08-25

Added

• upload_files resumes interrupted transfers by default, keeping each file’s tus URL under ~/.bfabric/resume per server, keyed by the file’s MD5 and source path. Pass resume_cache to choose another path, or resume_cache=None to keep no state and never resume.

Changed

• Breaking: new bfabric.BaseUrl, a str subclass replacing the pydantic HttpUrl on BfabricClientConfig.base_url, Entity.bfabric_instance and EntityUriComponents.bfabric_instance, which always ends with /bfabric. UrlTokenContext.base_url, TokenValidationSettings and WebappIntegrationSettings hold their instance URLs as BaseUrl too.

• connect_oauth / connect_pkce / connect_device_code / connect_pat and WebappClient.create normalise host case and a default port, and reject a non-HTTP URL with a ValueError.

• Breaking: on_duplicate="link" requires a B-Fabric that reports linkable and only links a duplicate reported as linkable; one whose resource is still pending or failed is uploaded instead. DuplicateResult carries linkable and existingResourceStatus.

• DuplicateResult.action is a Literal["upload", "skip", "link", "unsupported"]; an action the client does not recognise is normalised to unsupported and refused per file.

• upload_files raises WorkunitCompletionError, carrying the UploadSummary, when every transfer succeeded but marking the workunit available failed.

• A resumed upload continues into its original workunit, resource and tracked job instead of creating a second one, and an interrupted upload that saved a resume URL leaves its workunit processing rather than failed.

• upload_files documents that a successful return means the transfer completed, not that the file is stored — a caller that deletes its local copy must re-read the resource status and require available first.

Fixed

• The SUDS WSDL URL, and the show.html links printed by bfabric_read and bfabric-cli api read, no longer contain a doubled slash.

Removed

• Breaking: bfabric.transfer.api_to_rest_url — use client.config.base_url.

[1.21.0] - 2026-08-20

Added

• create_workunit attaches a dataset (a WorkunitDataset of name + base64-encoded csv/tsv/parquet) as the workunit’s output dataset, and references an existing one as its input via input_dataset_id. A dataset alone counts as workunit data, so neither resources nor files are required.

• on_duplicate="link" registers a duplicate as a resource pointing at the existing bytes instead of omitting the file, transferring nothing. See FileInfo.link_from_resource_id.

• upload_files / collect_file_infos accept exclude_names, dropping files by basename at any depth (e.g. .DS_Store).

• upload_files accepts an on_url callback, reporting each file’s resumable tus URL as soon as it exists (including for a transfer that then fails), so a caller can persist it for a later send_to_sink(resume_url=...).

• Config: new CLI-only scope field recording the scope requested at login, so a login can be replayed from disk; config_writer.clear_environment_credentials strips inline secrets while keeping the environment configured.

Changed

• upload_files(client, params) takes its files as UploadFilesParams.files, a list of UploadFileParam(path=..., on_duplicate=...); the force / link_duplicates booleans are gone.

• on_duplicate defaults to upload, so the duplicate check no longer runs unless asked for — pass skip for the old behaviour.

• UploadSummary holds one list per outcome — uploads, skips, failures, links — replacing the 1.20.0 counters; skips carry the duplicate they lost out to.

• A workunit whose files were all linked now completes instead of failing the “nothing uploaded” check.

• Two input files mapping to the same resource name are rejected before the workunit is created, not after.

• create_workunit accepts a plain mapping for params, validated internally so an invalid mapping raises ValidationError before any write.

• The OAuth module moved from bfabric._oauth to bfabric.oauth.

• import bfabric no longer imports polars eagerly (~296 ms → ~184 ms); it loads on first use.

• PKCE’s printed-URL fallback and timeout error now name the loopback redirect target and point at the device-code flow.

• use_client logs the reported error’s traceback at DEBUG; the Error: <message> line and exit code 1 are unchanged.

Removed

• Breaking: the zeep engine, which had stopped working against B-Fabric. EngineSUDS is the only engine, BfabricAPIEngineType.ZEEP is gone, and so is the bfabric[zeep] extra. An environment setting engine: ZEEP no longer loads — drop the key.

• Breaking: the engine parameter of Bfabric.from_config (which never applied it) and of BfabricClientConfig.copy_with.

Fixed

• An unreachable B-Fabric instance raises BfabricUnavailableError (a BfabricRequestError) naming the instance and the transport failure, instead of leaking suds.transport.TransportError, urllib.error.URLError or httpx.TransportError. Covers the SOAP engine and the OAuth/REST calls (JWKS, device code, registration, token exchange, introspection, PKCE).

• ResultContainer.assert_success raises BfabricRequestError instead of a bare RuntimeError; it remains a RuntimeError subclass, so existing except RuntimeError handlers keep working.

• MultiQuery.read_multi reserves query elements for obj’s other fields when chunking, instead of overflowing the API’s 100-element limit. A query that already exhausts it raises ValueError.

• Uploading a folder with sub-directories now works against servers that echo the resource name verbatim.

• A nested re-upload reports renamed_duplicate rather than exact_duplicate; both carry the skip action, so branch on action, not category.

• Device-code login reports a non-JSON token-endpoint response (e.g. an app-server 404 page during a redeploy) as BfabricOAuthError instead of a traceback.

• write_environment_to_config merges into an existing environment instead of replacing it, so unrelated keys survive a re-login; auth-owned keys are still replaced wholesale.

• Bfabric.connect() errors clearly when an auth_method: oauth config has no env_name.

• created_by / modified_by raise a ValueError naming the login instead of returning None.

• setup_script_logging no longer skips setup in subprocesses, which fell back to loguru’s verbose DEBUG default; its repeat guard is now process-local.

• The “could not find the config file” and “empty list provided for deletion” diagnostics go through loguru at WARNING instead of print(), so they honour the configured log level and sink.

• Packaging: project.readme points at a package-local README.md, fixing builds from source with hatchling 1.32.0.

• Internal: owner typing corrected across the entities/core descriptors and mixins; HasMany(bfabric_field=...) is now required and class-level access raises like HasOne.

[1.20.0] - 2026-08-03

• OAuth 2.0 authentication: connect_oauth (client-credentials grant, auto-refresh + optional token cache), connect_pkce (interactive browser login, PKCE), connect_device_code (headless device grant), and connect_pat (Personal Access Token). connect auto-routes to OAuth when an environment sets auth_method: oauth. All three OAuth entry points require an explicit client_id and scope — the core library bakes in no defaults (those are CLI policy; bfabric-cli auth … supplies its own), and connect() raises if an auth_method: oauth environment has no client_id. Token-acquisition failures (expired/revoked refresh token, unreachable token endpoint) raise BfabricOAuthError instead of leaking an authlib/requests traceback, and a failed PKCE login renders a distinct “Login failed” callback page showing the provider’s error (e.g. a two-factor-enrollment requirement) rather than always claiming success. Adds the authlib / joserfc dependencies.

• WebappClient — a dual-identity client pairing a token-derived user identity with a service (client-credentials) identity, for apps launched from B-Fabric.

• New bfabric.transfer layer for moving files to/from B-Fabric: fetch_to_path (batch download over rsync / scp / cp / symlink / HTTP, checksum-verified) and send_to_sink (resumable tus upload, behind the bfabric[transfer] extra), plus the B-Fabric bindings (ssh_source / http_source, UploadRestClient, token_provider, OAuth scope checks). Ported from the bfabric-tus-client prototype.

• bfabric.operations.workunit.upload_files — create or reuse a workunit and upload files as resources over tus (duplicate-skipping, optional track_job, failure cleanup). Exposed as bfabric-cli workunit upload.

• Config: new auth_method (password/oauth) and client_id environment fields, plus write_environment_to_config and DEFAULT_CONFIG_FILE. PAT environments store the token under a pat key with auth_method: pat (not login: __oauth__ / password:), so they no longer make older (≤1.19.0) clients reject the shared ~/.bfabricpy.yml.

• EntityReader lookups (read_id / read_ids / query / query_one) accept an entity class in place of the endpoint string — e.g. client.reader.read_id(Resource, id) — inferring both the endpoint and the result type; the string form (with optional expected_type) still works. The id/URI lookups (read_ids / read_uris) now return an EntityResult, a dict[EntityUri, Entity | None] subclass with .present (found entities as a list) and .by_id (found entities re-keyed by integer id), so callers write reader.read_ids(Resource, ids).present instead of wrapping the result in a helper.

• EntityUri.normalize — normalize a B-Fabric web URL (as copied from the browser) into a canonical entity URI, dropping extra query parameters (e.g. &tab=details) and the fragment, and normalizing host case and a default port. The EntityUri constructor stays strict and now hints at normalize when it rejects such a URL. URI validation is also no longer regex-based (urllib.parse instead), so its error messages name the actual problem. This leaves the canonical form the constructor accepts unchanged, except that credentials in the URL (https://user:pw@host/bfabric/…) are now rejected and the localhost exemption is case-insensitive.

• zeep is now optional — install bfabric[zeep] to use engine="zeep"; the default suds engine is unaffected.

• Use polars (with the rtcompat extra on macOS x86_64) instead of polars-lts-cpu, fixing a clash with pandera[polars]; minimum polars now 1.34.

• ResultContainer.assert_success now raises RuntimeError with a single message instead of a (message, errors) tuple.

• The @use_client CLI decorator now catches ValueError / RuntimeError and prints a concise Error: ... (exiting non-zero) instead of a traceback.

• ResultContainer.to_polars() returns an empty DataFrame for an empty result set instead of raising polars.exceptions.NoDataError, fixing a crash in bfabric-cli api read when a query matched no records.

• flatten_relations (and ResultContainer.to_polars(flatten=True)) recursively flattens nested struct columns instead of only one level, so a nested b.inner.p field becomes the b_inner_p column. Also shipped as the 1.19.1 hotfix.

• Internal: HasOne descriptors are now generic over their return type, so to-one accesses type-check to the related entity (no more cast / # pyright: ignore), with no runtime change; class→endpoint inference is centralized in import_entity.entity_type_of (the inverse of import_entity); removed dead required-relationship guards; ruff now treats pydantic.BaseModel as runtime-evaluated (dropping # noqa: TC00x workarounds).

[1.19.0] - 2026-06-10

Added

• Job entity wrapping the B-Fabric job endpoint.

• BfabricTokenExpiredError and BfabricTokenInvalidError in bfabric.errors: specific subclasses of BfabricTokenValidationFailedError raised instead of the generic error.

• bfabric.operations module: the canonical location for named write operations against B-Fabric.

  • operations.workunit.create_workunit — creates a workunit with its resources, rolling back on failure; accepts caller-supplied audit_attributes.

  • operations.dataset — create_dataset, update_dataset, preview_dataset_update, plus polars_to_dataset_dict and identify_changes.

  • operations.update_custom_attributes.

Deprecated

• bfabric.experimental.upload_dataset — moved to bfabric.operations.dataset; deprecation shim emits warnings and will be removed in the next release.

• bfabric.experimental.update_custom_attributes — moved to bfabric.operations.

Removed

• bfabric.experimental.upload_dataset.bfabric_save_csv2dataset (inner function) — removed outright; accessing it raises AttributeError with a migration hint.

• bfabric.experimental.dataset_column_types — internal implementation detail of upload_dataset, never public; symbols now live at bfabric.operations.dataset._column_types.

Fixed

• User.is_employee returns False when empdegree is missing or None, instead of raising ValueError.

[1.18.0] - 2026-04-20

Added

• EntityReader.query_one: returns at most one matching entity (or None).

• User.is_employee property: True iff the user’s empdegree parses to a value > 0.

• BfabricTokenValidationFailedError in bfabric.errors: raised when token validation fails.

• CLI commands decorated with use_client now accept --config-env and --config-file flags.

Changed

• EntityReader.query accepts an expected_type keyword for typed narrowing of results.

• User.find_by_login no longer triggers a FindMixin deprecation warning.

Fixed

• FindMixin.find_all now correctly handles string IDs.

[1.17.0] - 2026-03-12

Added

• Improved token authentication

  • bfabric.experimental.webapp_integration_settings

  • Bfabric.connect_token, Bfabric.connect_token_async: this one respects the list of allowed bfabric instances

  • bfabric.rest.token_data.validate_token

  • Bfabric.from_token_data: creates a new Bfabric instance from token data.

• ResultContainer.to_polars now has a flatten parameter to flatten struct columns into individual columns.

Deprecated

• Bfabric.from_config

• Bfabric.connect_webapp

Changed

• WorkunitExecutionDefinition no longer validates that dataset or resources are provided.

• setup_script_logging() now supports a BFABRICPY_LOG_LEVEL environment variable (DEBUG, INFO, WARNING, ERROR, CRITICAL, OFF/0). The previous BFABRICPY_DEBUG environment variable has been removed; use BFABRICPY_LOG_LEVEL=DEBUG instead.

Fixed

• References correctly handles references with extra fields like _position.

• ResultContainer.to_polars sets the schema length to None to fix bugs in some cases with more than 100 items.

[1.16.2] - 2026-03-02

Fixed

• BfabricAuth does not prevent users with username shorter than 3 characters from logging in.

Changed

• BfabricAPIEngineType is a StrEnum now, rather than str, Enum subclass.

[1.16.1] - 2025-12-15

Fixed

• Dataset correctly handles None values in items.

[1.16.0] - 2025-12-15

Added

• bfabric.typing with specific types for the API request and response objects.

• Executable.decoded_str, Executable.decoded_bytes

• Workunit.workunit_parameters to list workunit context parameters

• ExternalJob.client_entity

Changed

• Type hints have been narrowed in the public interfaces of the following classes:

  • Bfabric

  • ResultContainer

  • References

  • Entity

  • EntityReader

  • EngineSuds

  • EngineZeep

• Parameter.value returns "" for values not specified with required = false. This solves a few rare compatibility bugs and I do not think having the ability to distinguish between None and "" makes much sense at this level.

Removed

• Legacy workaround for B-Fabric with non-list technology field is removed. Assumed to not be breaking since for the current B-Fabric instances, this change should be transparent.

Deprecated

• Executable.decoded: misleading name

[1.15.1] - 2025-12-09

Fixed

• Allow http://localhost for EntityURIs.

[1.15.0] - 2025-12-09

Added

• TokenData.web_service_user which indicates whether the user has permission to use webservices API.

[1.14.1] - 2025-12-02

Added

• EntityReader.read_ methods accept a expected_type parameter to specify the expected type of the entity (default Entity).

Fixed

• A few imports within the package were accidentally broken in 1.14.0 which is resolved now. basedpyright has been added to identify issues like this earlier.

[1.14.0] - 2025-12-02

• The minimal Python version has been updated to 3.11.

• bfabric.entities

  • The read-only API in bfabric.entities has been redesigned to support multiple B-Fabric instances cleanly.

  • In particular, when handling configuration files it should now be obvious how to specify entities by their URI rather than having to specify instance, classname, and id separately.

  • The main new concept is the EntityUri which standardizes the way entities are specified by their URI.

  • EntityReader is a new class which implements the resolution, currently for only one B-Fabric instance, but extensible to support multiple B-Fabric instances in the future.

  • We also remove the reliance on custom entity classes providing generic functionality to read entities and their references. This is available through the EntityReader class which can be accessed through Bfabric.reader.

• Versioning:

  • The bfabric Python package versioning starting with 1.14.0 does not track the B-Fabric server version anymore.

  • This will allow us to adapt to implement a more semantically meaningful versioning, starting with 1.14.0.

  • Applications, like bfabric-scripts can instead track the particular version of bfabric against which they are tested/developed either in their version (planned for bfabric-scripts) or through documentation.

Added

• Entity.uri property to obtain the URI of an entity.

• Entity.refs resolves all references of entities generically, writes the data into data_dict and is compatible with pre-loaded references from B-Fabric.

• bfabric.entities.core.uri.EntityUri to identify entities regardless of B-Fabric instance.

• bfabric.entities.core.uri.EntityUriComponents to access individual components (bfabric_instance, entity_type, entity_id) from a URI.

• bfabric.entities.core.uri.GroupedUris helper for looping over lists of EntityUris.- bfabric.entities.core.entity_reader which allows reading entities by URI, ID, and general queries.

• bfabric.entities.cache which supersedes bfabric.experimental.cache.

• HasOne and HasMany should be able to use resolved entities when loading with fulldetails=True.

• Entity.custom_attributes for reading custom attribute values.

• bfabric.experimental.update_custom_attributes for updating custom attributes correctly.

Changed

• Minimal Python version is now 3.11.

• BFabricClientConfig does not allow passing None for values anymore.

• HasMany retains response order, instead of redundant sorting by ID.

• bfabric.entities allows loading entity references without custom definitions in Python.

• Internal

  • Entity has a new constructor parameter bfabric_instance which in the future will become mandatory.

  • BfabricClientConfig.base_url always ends with exactly one / now.

  • bfabric.entities do not define custom constructors anymore, simplifying future changes (and removing tiny inconsistencies).

  • TokenData retrieval uses async httpx internally, but provides a sync interface for compatibility.

Removed

• bfabric.experimental.cache is removed in favor of bfabric.entities.cache.

• Future deprecations:

  • Entity.find, Entity.find_all, Entity.find_by will be removed in favor of EntityReader methods.

  • Passing entity class to cache_context will be removed in favor of their name (this is because not all entities have custom classes).

[1.13.36] - 2025-10-27

Changed

• This is the last release supporting Python 3.9. From the next release onwards, Python 3.11 or higher will be required.

• Flask is not a dependency anymore, as the dependency will be moved to bfabric-scripts.

• Version upper bounds for dependencies have been defined.

• More robust URL handling for token authentication.

Fixed

• A test-only incompatibility with Polars was fixed.

[1.13.35] - 2025-09-22

Changed

• Dataset column type detection now supports case-insensitive entity matching (e.g., “resource” is detected as “Resource” type automatically).

[1.13.34] - 2025-09-19

Added

• bfabric.experimental.cache which implements re-entrant lookup caching for entities (when retrieved by ID).

• Fields filename, storage_relative_path, storage_absolute_path have been added to Resource. This should be used in the future, to ensure path handling is performed consistently

Removed

• bfabric.experimental.entity_lookup_cache has been removed in favor of the new (experimental) API.

[1.13.33] - 2025-08-26

Added

• Add new field WorkunitDefinition.registration.user_id.

• Add created/modified fields to Workunit entity.

Fixed

• Legacy wrapper creator also writes the inputdataset name.

[1.13.32] - 2025-08-26

Added

• New entities: Workflow, WorkflowStep, WorkflowTemplate, WorkflowTemplateStep

• New relationship: HasUser to resolve user references (since these are modeled differently from other relationships).

Fixed

• There was an issue in import_entity.py which made loading entities with additional uppercase characters in the type fail.

• Fix compatibility in legacy wrapper creator for dataset-flow workunits.

[1.13.31] - 2025-09-20

Changed

• The Bfabric.connect_webapp will use the caller field from the token data to set the base_url of the client.

  • Existing code should not break, but will emit a deprecation warning as we plan to remove the old parameters.

  • Theoretically, it changes the semantics, but in practice it should yield the same result for all known use cases.

Added

• TokenData will include the caller url now.

• Entity User has been added.

Fixed

• Legacy wrapper creator creates "" rather than None for empty parameter values.

[1.13.30] - 2025-08-19

Fixed

• Fix legacy wrapper creator for orders without projects.

[1.13.29] - 2025-07-04

Changed

• bfabric.rest.token_data.get_token_data now only requires the base_url instead of the whole config.

[1.13.28] - 2025-06-27

Breaking

• base_url is now a mandatory parameter in the configuration.

Added

• It is now possible to configure ~/.bfabricpy.yml without a default environment. In that case it will always be necessary to specify the requested config environment to be used.

• bfabric.entities.Executable has parameters relationship now

Removed

• Some old submitter related functionality is deleted.

Changed

• Columns of tables named after B-Fabric entities, containing only integers, will be set as the specified type when saving to B-Fabric (in experimental.upload_dataset).

[1.13.27] - 2025-05-21

Added

• Attribute Bfabric.config_data to obtain a ConfigData object directly.

• TokenData.load_entity convenience method to load an entity from the token data.

• Entities Instrument, Plate, Run were added (but with no extra functionality).

• Workunit.{application_parameters, submitter_parameters} to access parameter values.

Changed

• Submitter parameters will not be written any longer to WorkunitExecution parameters.

Fixed

• Compatibility with upcoming change that Application can have multiple technology values.

Deprecated

• Workunit.parameter_values will be removed in favor of Workunit.application_parameters and Workunit.submitter_parameters in a future version.

[1.13.26] - 2025-04-26

This release introduces an environment variable BFABRICPY_CONFIG_OVERRIDE to configure the Bfabric client completely, along with a new method for creating an instance of the Bfabric client, Bfabric.connect(). This will allow us to propagate any configuration to subprocesses reliably. BFABRICPY_CONFIG_ENV remains available with the same semantics, but lower priority than BFABRICPY_CONFIG_OVERRIDE.

This also simplifies the logic that was present in Bfabric.from_config which is why this introduced with a new API to prevent configuration mix-ups.

Added

• New environment variable BFABRICPY_CONFIG_OVERRIDE to configure the Bfabric client completely

• New method Bfabric.connect() for creating an instance of the Bfabric client

Changed

• Renamed Bfabric.from_token to Bfabric.connect_webapp() (along with some changes, no known users of this API yet)

• Disallowed default as an environment config name

• bfabric.cli_integration.utils.use_client uses Bfabric.connect() instead of Bfabric.from_config()

Deprecated

• Bfabric.from_config is now deprecated in favor of Bfabric.connect()

[1.13.25] - 2025-04-22

Breaking

• Bfabric.from_token returns the TokenData in addition to the client instance. While this is breaking, I’m not aware of any existing users of this API and I noticed that this information is going to be needed in this context often.

Added

• bfabric.experimental.upload_dataset.warn_on_trailing_spaces function used by bfabric-scripts to validate

[1.13.24] - 2025-04-08

Removed

• cyclopts is not a dependency of bfabric anymore, but rather of bfabric-scripts and bfabric-app-runner.

Changed

• Bfabric client now does not take engine argument anymore, but rather this information comes from ClientConfig. For compatibility reasons, the engine argument is still accepted in Bfabric.from_config and Bfabric.from_token.

[1.13.23] - 2025-03-25

Fixed

• Unsuccessful deletions are detected by checking the B-Fabric response.

• Handle problematic characters in Workunit.store_output_folder.

• BfabricRequestError did not properly subclass RuntimeError.

Changed

• WorkunitDefinition uses PathSafeStr to normalize app and workunit names.

• Internal: ResultContainer has no optional constructor arguments anymore to avoid confusion.

Added

• Generic functionality in bfabric.utils.path_safe_name to validate names for use in paths.

• bfabric.entities.Dataset.{write_parquet, get_parquet} methods for writing parquet

[1.13.22] - 2025-02-19

Fixed

• Correctly read datasets, if columns were swapped in B-Fabric.

Changed

• Dataset.types was renamed to Dataset.column_types, since there is only 1 usage and it is recent, this is not considered breaking.

[1.13.21] - 2025-02-19

Added

• Entity.load_yaml and Entity.dump_yaml

• Bfabric.from_token to create a Bfabric instance from a token

• bfabric.rest.token_data to get token data from the REST API, low-level functionality

Changed

• Internally, the user password is now in a pydantic.SecretStr until we construct the API call. This should prevent some logging related accidents.

[1.13.20] - 2025-02-10

Breaking

• the old bfabric.cli_formatting and bfabric.bfabric2 modules have been deleted

Changed

• move use_client to bfabric.utils.cli_integration: this will allow reuse between bfabric-scripts and bfabric-app-runner

• move functionality from bfabric.cli_formatting to bfabric.utils.cli_integration

[1.13.19] - 2025-02-06

Fixed

• Config: Log messages of app runner are shown by default again.

[1.13.18] - 2025-01-28

Changed

• bfabric_upload_resource.py does not print a list anymore, but rather only the dict of the uploaded resource.

Fixed

• Some scripts were not Python 3.9 compatible, which is restored now. However, we can consider increasing the Python requirement soon.

[1.13.17] - 2025-01-23

Added

• bfabric-cli workunit export-definition to export workunit_definition.yml files

• Executable.storage relationship

Fixed

• Workunit.parameters and Workunit.resources are optional

• bfabric-cli had unmet dependencies, that were not caught by the tests either

[1.13.16] - 2025-01-22

Added

• Add missing Entity.__contains__ implementation to check if a key is present in an entity.

• polars_utils.py which contains functionality to normalize relational fields in tables

• Add bfabric-cli executable inspect command to inspect executables registered in B-Fabric.

• Add bfabric-cli executable upload command to upload executables to B-Fabric.

Fixed

• Order.project is optional

[1.13.15] - 2025-01-15

Added

• Entity Sample

• Relationship Resource.sample.

• Extract logic for container resolution into HasContainerMixin which is right now shared between Sample and Workunit.

Fixed

• Fix bug in bfabric_save_fasta.py.

[1.13.14] - 2025-01-08

Changed

• Slurm submitter prints more diagnostics about PATH variable etc.

Fixed

• Some commands in bfabric-cli are broken because __future__.annotations is imported and this breaks cyclopts.

[1.13.13] - 2024-12-18

Changed

• Move bfabric-cli read to bfabric-cli api read.

• (internal) use_client decorator is introduced to simplify and standardize the client usage in the new CLI code.

Added

• Functionality to log by specifying the workunit instead of external job, which is used in some legacy scripts.

• Add logging commands to bfabric-cli api log.

• Add bfabric-cli api save.

• Submitter script prints the id and hostname of apps, which in practice can be very useful.

Fixed

• A bug in bfabric_save_workflowstep.py that crashed the script.

[1.13.12] - 2024-12-17

Changed

• The submitter ensures that workunits always get set to processing.

Fixed

• The script bfabric_setResourceStatus_available.py and other uses of report_resource, correctly search files which have a relative path starting with / as in the case of the legacy wrapper creator.

[1.13.11] - 2024-12-13

Fixed

• The script bfabric_setWorkunitStatus.py did always set the status to available, instead of the specified status.

Changed

• bfabric_setResourceStatus_available.py calls the report_resource function, in general this functionality has been refactored.

[1.13.10] - 2024-12-13

Fixed

• If bfabricpy.yml contains a root-level key which is not a dictionary, a correct error message is shown instead of raising an exception.

• A bug introduced while refactoring in slurm.py which passed Path objects to subprocess instead of strings.

• Submitter

  • does not use unset JOB_ID environment variable anymore.

  • does not set unused STAMP environment variable anymore.

• Fix bfabric_save_workflowstep.py bugs from refactoring.

Added

• Experimental bfabric-cli interface. Please do not use it in production yet as it will need a lot of refinement.

[1.13.9] - 2024-12-10

From this release onwards, the experimental app runner is not part of the main bfabric package and instead a separate Python package with its individual changelog.

Added

• Relationship: ExternalJob.executable

• (experimental) EntityLookupCache that allows to cache entity lookups in a script to avoid redundant requests.

• Specific use case script: bfabric_save_resource_description.py (the functionality will be available in a future CLI).

Fixed

• Entity.find_all returns no values when an empty list is passed as an argument.

Changed

• Except for macOS x86_64 (which we assume is Rosetta emulation nowadays), we use the faster polars instead of polars-lts-cpu.

• BfabricRequestError is now a RuntimeError subclass.

• Add py.typed marker.

Removed

• bfabric_legacy.py has been removed.

• math_helper.py has been removed.

[1.13.8] - 2024-10-03

This release contains mainly internal changes and ongoing development on the experimental app interface functionality.

Added

• Entities can be compared and sorted by ID now.

• Show Python version in version info string.

• Caching for bfabric_list_not_existing_storage_directories.py.

• (experimental) add initial code for a resource based application dispatch

• (experimental) new app_runner cli that integrates all commands into a single interface

Fixed

• bfabric_read.py is a bit more robust if “name” misses and tabular output is requested.

Changed

• bfabric.scripts has been moved into a namespace package bfabric_scripts so we can later split it off.

• (internal) migrate to src layout

• (experimental) the former process step of the app runner has been split into a process and collect step where, the collect step is responsible for generating the output.yml file that will then be used to register the results.

• (experimental) app runner apps by default reuse the default resource

[1.13.7] - 2024-09-17

Fixed

• bfabric_save_csv2dataset.py considers all rows for schema inference, only considering the first 100 rows did cause problems with some files previously.

Added

• Script logging configuration

  • If BFABRICPY_DEBUG environment variable is set, log messages in scripts will be set to debug mode.

  • Log messages from __main__ will also be shown by default.

• bfabric.entities.MultiplexKit to extract multiplex kit information.

• bfabric.entities.Workunit.store_output_folder implements the old rule, but more deterministically and reusable.

• (Experimental) bfabric.experimental.app_interface core functionality is implemented

  • This can be used as a building block to standardize the input preparation for applications.

Changed

• Correctly support optional workunit parameters.

Removed

• bfabric.entities.Resource association application has been removed as it does not exist

[1.13.6] - 2024-08-29

Added

• Entity.find_by has new parameter max_results.

Changed

• bfabric_read.py prints non-output information exclusively through logger and does not restrict entity types anymore.

• bfabric_delete.py accepts multiple ids at once and does not restrict entity types anymore.

• Both engines raise a BfabricRequestError if the endpoint was not found.

Removed

• bfabric.endpoints list is deleted, since it was out of date and generally not so useful.

[1.13.5] - 2024-08-13

Added

• The Bfabric instance is now pickleable.

• Entities mapping:

  • Add Entity.id and Entity.web_url properties.

  • Add Entity.__getitem__ and Entity.get to access fields from the data dictionary directly.

  • Add Entity.find_by to find entities by a query.

  • More types and relationships

  • Relationships defer imports to descriptor call, i.e. circular relationships are possible now.

  • HasOne and HasMany allow defining optional=True to indicate fields which can be missing under some circumstances.

• Add nodelist column and application name to bfabric_list_not_available_proteomics_workunits.py output.

Changed

• Entity.find_all supports more than 100 IDs now by using the experimental MultiQuery API.

[1.13.4] - 2024-08-05

Added

• Add Workunit, Parameter, and Resource entities.

• Add concept of has_many and has_one relationships to entities.

• bfabric_slurm_queue_status.py to quickly check slurm queue status.

• Bfabric.save provides method which can be set to checkandinsert for specific use cases.

Changed

• Most messages are now logged to debug level.

• The old verbose version information is now always logged, to INFO level, since it could entail useful information for error reporting.

[1.13.3] - 2024-07-18

Added

• Flask

  • New endpoint GET /config/remote_base_url for testing

Changed

• Flask

  • Simplify logging by using loguru only.

  • Simplified setup logic since the production use case should use a WSGI server.

Fixed

• bfabric_save_csv2dataset.py had an undeclared dependency on numpy and a few bugs which was improved.

[1.13.2] - 2024-07-11

Added

• Add bfabric.entities.Dataset to easily read datasets.

• Pydantic-based configuration parsing

  • The config format did not change.

  • The code is easier to maintain now.

  • Additionally, there is a lot more validation of the configuration file now, that should catch errors early.

• Make host and port configurable in bfabric_flask.py (currently only dev mode).

[1.13.1] - 2024-07-02

Changed

• bfabric_save_csv2dataset will raise an error if problematic characters are found in any of the cells

• Correctly define bfabric_setWorkunitStatus_available.py, and processing and failed variants.

Added

• Add loguru for future logging refactoring.

• Easily runnable tests with nox and standardized formatting using pre-commit.

Removed

• Pandas is no longer a dependency, and has been replaced by polars.

[1.13.0] - 2024-05-24

This is a major release refactoring bfabricPy’s API.

Changed

• The Bfabric class operations now return ResultContainer objects.

  • These provide a list-like interface to access individual items or iterate over them.

  • Individual items are a dictionary, potentially nested, and not specific to suds/zeep anymore.

  • Convenience conversions, e.g. to a polars DataFrame, can be provided there.

• Configuration is now defined in ~/.bfabricpy.yml and supports multiple configurations, which can be selected by the BFABRICPY_CONFIG_ENV environment variable. Please consult the README for an example configuration.

• Use pyproject.toml for package configuration.

• Scripts have been refactored on a case-by-case basis.

Added

• Zeep can be used instead of suds for SOAP communication.

• Bfabric can be instantiated without authentication, that can be provided later. This is useful in a server setup.

• Pagination support in Bfabric, specify the number of max_results and a potential offset. Pages handling is abstracted away.

• Detect errors in responses, e.g. invalid login.

Removed

• Several old scripts have been moved into a deprecated_scripts folder.

• Wrapper creator related code is currently not updated but has been extracted into a dedicated folder wrapper_creator as well.