v2.54.0

Added:

• METS Server: providing concurrent additional access to the METS file for parallel processing, #966
• Web API: Cache jobs in the processing server with an optional callback once processed, #1069
• Web API: Lock pages output file groups of a workspace to prevent simultaneous non-additive access to workspaces, #1069
• Web API: Support job dependency for caching complete fully-deterministic workflows, #1069
• Web API: Processing server will start all ready requests, not just the first one, #1069
• Web API: Workers will create on demand, retry attempts configurable via OCRD_NETWORK_WORKER_QUEUE_CONNECT_ATTEMPTS, #1093
• ocrd_utils.config to collect all configuration based on environment variables in one place, #1081
• Processor CLI: Warn if a given page ID cannot be found in METS, #1088, #1089

Changed:

• Processors now have worker and server subcommands, with separate --help, for starting processing worker/processor server, #1087
• Move tf_disable_interactive_logs (to silence keras/tensorflow print statements) to ocrd_utils.logging and do not call on module-level, #1090, #1091
• 🔥 We do now properly distinguish between original and local-file FLocat, original URL will not be changed for downloads anymore, #323, #1079
• 🔥 logging has been streamlined to be better usable as a library, #1080

v2.53.0

Fixed:

• WorkspaceValidator: make the check for consistency of pc:Page[@pcGtsId] and mets:file[@ID] optional with the mets_fileid_page_pcgtsid skip flag, #1066
• ocrd resmgr download: use basedir as an arg and not a kwarg, #1078

Changed:

• WorkspaceValidator: Download files temporarily/on-demand, #1066
• ocrd-* --version now prints only the version of the processor without noise or core version, #1068

Added

• Environement variables to control optional retries and timeouts for downloading files:
  • OCRD_DOWNLOAD_RETRIES: Number of times to retry failed attempts for downloads of workspace files. #1073
  • OCRD_DOWNLOAD_TIMEOUT: Timeout in seconds for connecting or reading (comma-separated) when downloading. #1073
• Environment variables used throughout core are now documented in README and ocrd --help, #1073
• Web API: --create-queue option to on-demand create RabbitMQ for processing workers, #1075
• Web API: --queue-connec-attempts to retry connection to RabbitMQ in case server is not yet running, #1075

v2.52.0

Added:

• make deps-cuda: Makefile target to set up a working CUDA installation, both for native and Dockerfile.cuda, #1055
• Implementation of the Standalone Processor Server module, #1030
• ocrd_utils.guess_media_type to consistently try to determine media type from a file name, #1045

Changed:

• Refactoring the Network CLI, all network module CLI are in ocrd_network now, #1030
• The Processing Server uses ocrd-all-tool.json file, removing local processor install dependencies, #1030
• Overall improvement and refactoring of the ocrd_network package, #1030
• Optionally skip deployment of mongodb and rabbitmq to make external usage/management possible, #1048
• page_from_file now also accepts a (str) file path in addition to OcrdFile, #1045
• packaging: install/uninstall in correct build order, use python -m build instead of python setup.py sdist bdist_wheel, #1051

Removed:

• Obsolete travis CI configuration removed, #1056
• Support for end-of-life python versions 3.5 and 3.6, #1057

Fixed:

• Makefile FIND_VERSION macro: use grep -E instead of grep -P for macos compatibility, #1060
• ocrd resmgr: detect HTTP errors as such and don't try to continue if HTTP >= 400, #1062
• PageValidator: Ensure TextLine has coordinates when checking for Baseline containment, #1049

v2.51.0

Changed:

• core cuda Docker: CUDA base image working again, based on ocrd/core not nvidia/cuda in a separate Dockerfile.cuda, #1041
• core-cuda Docker: adopt #1008 (venv under /usr/local, as in ocrd_all, instead of dist-packages), #1041
• core-cuda Docker: use conda (micromamba) for CUDA toolkit, and nvidia-pyindex for CUDA libs – instead of nvidia/cuda base image, #1041
• more robust workaround for shapely#1598, #1041

Removed:

• Revert #882 (fastentrypoints) as it enforces deps versions at runtime
• Drop ocrd_utils.package_resources and use pkg_resources.* directly, #1041
• ocrd resmgr: Drop redundant (processor-provided) entries in the central resource_list.yml.

v2.50.0

Added:

• 🔥 ocrd_network: Components related to OCR-D Web API, #974

Changed:

• bashlib: support file input from multiple file groups, #1027, #1031

Fixed:

• Don't output default docstrings for bashlib processors, #1026

v2.49.0

Changed:

• 🔥 (for now: also) publish Docker images to ghcr.io, not docker.io, #997
• ocrd resmgr: eynollah models now provided by eynollah itself, qurator-spk/eynollah#91

v2.48.1

Changed:

• make docker-cuda: Support CUDA 11.3 not 11.2, #1020

v2.48.0

Changed:

• 🔥 ocrd.run_processor / ocrd.processor.get_processor: rm unnecessary ocrd_tool kwarg #998, #1009
• chdir into workspace directory for both cached and uncached get_processor, #972, 987
• 🔥 new CUDA base image 20.04, support CUDA runtime 11/12, not 10, #1014

Fixed:

• make install: do not update opencv-python-headless or numpy for python <= 3.6, #1014

v2.47.4

Changed:

• resmgr: ocrd-typegroups-classifier resources now listed decentrally, #1011, OCR-D/ocrd_typegroups_classifier#15

v2.47.3

Fixed:

• Docker: reintroduce python3-pip because it is needed in the standalone core image, #1004