Added:

• ocrd network client check-status has a --verbose flag for more detailed job status, #1348

Changed:

• Support multiple output file groups for processors, #1344
  • OcrdPageResult: replace by proxy class OcrdPageResultVariadicListWrapper with list semantics and variadic constructor (with the original class now under SingleOcrdPageResult)
  • Processor.process_page_file: handle results from process_page_pcgts
    as lists:
    • split output_file_grp with commas (just as input_file_grp)
    • iterate over output file groups and OcrdPageResult
    • log error if there are more results than output file groups (that will get lost)
    • raise FileExistsError (in order to skip actual computation) iff output file exists for all output file groups
    • make output files (and file IDs), and save images etc for each output independently
• PAGE API: get_AllRegions available for all region types, not just PAGE root, #1344
• ocrd_network: Update RabbitMQ from 3.12 to (latest) 4.2, #1348
• ocrd_network: Fix and improve logging for network integration tests, #1348

Fixed:

• 🔥 do not log RabbitMQ credentials, #1346, #1348

Added:

• test combinations of OCRD_* config variables and multi-output, #1344

Fixed:

• Include ocrd-command and ocrd-merge in the ocrd-all-tool.json, #1347

Merged PR:

• add ocrd-command and ocrd-merge to distributed ocrd-all-tool.json by @bertsky in #1347

Added:

• ocrd-command processor to run arbitrary PAGE transformation CLIs, #1343
• various parameter presets for ocrd-command, #1343
• ocrd-merge processor to join multiple PAGE inputs by concatenation, #1343
• test coverage for ocrd-filter, ocrd-command, and ocrd-merge, #1343
• Resource Manager Server as ocrd_network analogon of ocrd.cli.resmgr, #1309
  • ocrd network resmgr-server for triggering Resource Manager Server (RMS) in the background
  • Processing Server also deploys RMS on each processing host

Fixed:

• Page.get_ReadingOrderGroups: sort by index, use OrderedDict as result
• OcrdAgent.notes: convert to dict to accommodate pydantic 2 with older lxml
• ocrd.resource_manager: ensure necessary + reduce unnecessary updates of user database
• ocrd.resource_manager: deduplicate entries (newer wins) before updating user database
• ocrd resmgr download: extract archives independent of whether they are URLs or local paths
• ocrd resmgr download: if --overwrite, ensure the old res gets removed
• ocrd resmgr download: default to data location instead of first in list of allowed
• ocrd_utils.list_all_resources: filter module non-resource files w/ more anti-patterns
• ocrd_utils.list_all_resources: no subpaths except for cwd location, OCR-D/spec#263, #1315
• ocrd_utils.list_all_resources: filter resources via media (MIME) type, if specified, #1315

Merged PR

• drop ocrd-distributed resource_list.yml for good by @bertsky in #1322
• resmgr download: implement git clone by @bertsky in #1340
• Continuation of #1309: Implementation of the resource manager server (issue #1294) by @MehmedGIT in #1319
• add builtin processors ocrd-command and ocrd-merge by @bertsky in #1343

Full Changelog: v3.7.0...v3.8.0

Changed:

• 🔥 upgrade and adapt ocrd_network to pydantic v2, #1342

Removed:

• 🔥 drop bashlib processors, retain only the Python API, #1339

Changed:

• ocrd network client workflow run: Allow passing workflow as ocrd process tasks as alternative to a workflow file, #1264, #1335
• Processing Server: Store page ranges as comma-separated page IDs instead of relying on .. numeric range operator, #1333, #1334
• ocrd network: More verbose error messages instead of plain assertions, #1337

Removed:

• 🔥 Remove Processor Server in favor of Processing Server + Processing Workers only, #1338, OCR-D/spec#265

Merged PR:

• run_workflow: pass comma-separated page ids instead of interval range by @bertsky in #1334
• run_workflow client: allow workflow param as in ocrd process by @bertsky in #1335
• Improve ocrd-network-client error handling by @joschrew in #1337
• Remove the processor server by @joschrew in #1338

Fixed:

• Pin shapely runtime dependency and numpy build dependency for Python <= 3.8, #1332

Added:

• Extended syntax for filtering pages with --page-id and OcrdMets.find_files, #1328, #1328
  • negation operator ~, e.g. ~PHYS_0001..PHYS_0004 to not match the first 4 pages
  • filtering by attributes of mets:div for both physical (prefix physical:) and logical (prefix logical:) elements, e.g. ~logical:type:binding to skip all pages that are book bindings

Fixed:

• OCRD_MISSING_INPUT config variable only checked existence of mets:file not the actual file on disk even when set to ABORT, #1330
• Fix processor CLI subcommand worker|server broken due to changed Enum handling in click 8.2, #1331
• Typo: proper log message if media type could not be determined, #1330, bertsky/workflow-configuration#35
• Processor.zip_input_files: always sort pages in physical order, #1330

Added:

• make test-cuda-{torch,tf2,tf1}: ensure CUDA is available and shapely#1598 does not trigger, #1326
• ocrd ocrd-tool dump-module-dirs to dump moduledir of every tool in an ocrd-tool.json, #1326

Fixed:

• ocrd workspace find --undo-download: Unset local_filename only after unlinking, #1324, #1325
• METS Server: do not convert None to the string "None", #1324, #1325

Changed:

• Ensure logging files and directories are writeable for non-root users, #1214