Tags give the ability to mark specific points in history as being important
  • v1.13.4
    Release v1.13.4
    • !137 Add pyproject.toml: This follows bob packages new standard
    • !138 Gracefully handle docker client creation error: Up until now, the client creation did not fail when Docker daemon was not running but the behaviour seems to have changed on macOS. In any case, ensuring that the client creation failure does follow the rest of the logic is the right thing to do. Fixes #107 Fixes beat.editor#282
  • v1.13.3
    Release v1.13.3
    • !136 Cleanup empty fields in v2 databases: While they are optional, beat/beat.editor> ensures that you do not get useless empty fields that may look like something is missing hence remove them from here to ensure testing goes well.
  • v1.13.2
    Release v1.13.2
    • !134 Fix protocole template creation when migrating a database: The wrong name was used to generate the template. Also improve handling of protocol template double creation. If the content is different warn the user and ask whether it should continue with the database creation.
    • !135 Fix generated protocol templates: The original work done for the protocol template was working however it has been discovered that the helper script did not correctly generate the protocol templates. The script has been fixed in !134 however the test data were not. This merge request fixes that. As a byproduct of this cleanup, it has been found that the large database was in fact wrongly written and thus has been fixed as well.
  • v1.13.1
    Release v1.13.1
    • !134 Fix protocole template creation when migrating a database: The wrong name was used to generate the template. Also improve handling of protocol template double creation. If the content is different warn the user and ask whether it should continue with the database creation.
  • v1.13.0
    Release v1.13.0
    • !133 Custom tmpfs through label: This merge request implements a new feature handling additional tmpfs mount points provided through image label. The goal here is to allow images with special needs to provide them through a well defined format. Fixes #106
  • v1.12.0
    Release v1.12.0
    • !132 Refactor database sharing: This refactoring moves the database sharing responsibility to the database declaration. The author can choose whether the dataset shall be made available directly to the algorithm container. As a consequence, the implementation done in beat.cmdline!103 will not work anymore. Fixes #105
  • v1.11.1
    Release v1.11.1
    • !131 Fix algorithm name pattern in experiment globals: It's missing support for - in the author part of algorithm author name. Fixes #104
  • v1.11.0
    Release v1.11.0
    • !128 Algorithm database sharing: Implement a way to mount the database folder on the algorithms containers. Fixes #103
    • !129 Sphinx 3 autodoc no longer accepts autodoc_default_flags and autodoc_default_options should be used
    • !130 Add missing global parameters to shared_datasets experiment: While missing parameters in globals will not affect the execution, it will be detected by beat/beat.editor>.
  • v1.10.7
    Release v1.10.7
    • !126 Pre commit cleanup: This merge request updates the pre-commit configuration and applies a full source cleanup.
    • !127 Update docker related information: Following the release of BEAT runtime images on Docker hub and beat/beat.env.examples>, update the documentation related to the use of Docker.
  • v1.10.6
    Release v1.10.6
    • !123 Ensure that writing to a non existing output shows an error: The test added by this merge request ensure that writing to an output that does not exists is reported. The error will be that "NoneType" as no property "write" which is expected following how outputs work. Part of #99
    • !124 Add test for setup in cxx algorithm: Currently only the processing part during an experiment run was tested. This new test ensures that the setup feature works properly.
    • !125 Add minimal rw tmpfs for /tmp and /run: At least in production, some issues happened that required to have /tmp writable. To preserve the read-only state of the container, use a small tmpfs that is enough for the container to run. /run is another that might be needed so make it part of the defaults used. Fixes #101
  • v1.10.5
    Release v1.10.5
    • !121 Fix database prototype: The database prototype setup method has a wrong signature and is used in beat/beat.editor> when creating a new Database asset. This merge request fixes it as well as add the call to the default implementation which is mandatory for everything to be setup properly. The setup method has been added to two Views in the test prefix so that its use is also tested. Fixes #97
    • !122 Add optionnal environment entry to database: This merge request add support for defining the database environment to use in the database declaration. This allows the BEAT infrastructure to no depend on list of database supported by a given environment and allows for more flexibility to test new databases locally especially when the use of Docker is required when running for example C++ algorithms. Part of #64
    • !123 Ensure that writing to a non existing output shows an error: The test added by this merge request ensure that writing to an output that does not exists is reported. The error will be that "NoneType" as no property "write" which is expected following how outputs work. Part of #99
  • v1.10.4
    Release v1.10.4
    • !120 Move to new images: This merge request moves all the test images to the new version created through beat/beat.env.builder> The use of these images means that the old image enumeration becomes effectively a backup solution. The test should now take less time as the label based search is faster. Using these new images also means that we can archive all the various beat.env.XXX repositories and only use beat/beat.env.builder>.
  • v1.10.3
    Release v1.10.3
    • !114 Improve execution handling by enforcing the use of the current user for the execution of the docker container where nothing specific was written in the configuration. This also allows to make the container itself read-only.
    • !115 Only enable GPU runtime if available so that environment can be re-used on both CPU and GPU only types of node.
    • !116 Don't use f string yet as the processing nodes are not on Python 3.6 yet.
    • !117 Add size constraint for description field to ensure that the field is not longer than what is allowed in beat/beat.web>
    • !118 Logger cleanup: update deprecated warn call to warning and remove unused logger.
    • !119 Add test assets with too long descriptions to allow test improvments also in beat/beat.cmdline>
  • v1.10.2
    Release v1.10.2
    • !113 Add missing variable initialisation for optional argument in worker script which makes it fail to start without the --docker option
  • v1.10.1
    Release v1.10.1
    • !111 improve complex setup handling with NFS mounts
    • !112 Improve disconnection handling on worker reload
  • v1.10.0
    Release v1.10.0
    • !106 Implement environment discovery using docker labels with the original code is kept as a fallback method in case a node only has images using the old build process
    • !107 Add logging information to lable based discover to match the describe based implementation.
    • !108 Handle docker exception following the raise on error property for the new environment discovery
    • !109 Improve handling of message handler shutdown in local executor for early errors
    • !110 Make cache access more specific: This improves the handling of the cache by only mounting what is necessary
  • v1.9.2
    Release v1.9.2
    • !105 Remove loops from representation in the schema. It' already covered by the blocks entry.
  • v1.9.1
    Release v1.9.1
    • !99 Dependencies cleanup: move matplotlib and pillow to the test dependencies as they are not needed for this package to work.
    • !100 Implement the use of the experiment execution related scripts from a different python environment.
    • !101 Move conda related helper methods out of test class: That will make them re-usable in other tests like in beat/beat.cmdline>
    • !102 Remove Biometrics evaluation from the definition of BEAT as it became more generic.
    • !103 Improve protocol template lookup for database prototype: Rather than blindingly selecting the first protocol template found, look for the first valid one.
    • !104 Remove "license_file" entry on conda recipe to fix builds.
    • !105 Remove loops from representation in toolchain schema as the blocks entry covers them as it already does for the blocks and analysers.
  • v1.9.0
    Release v1.9.0
    • !55 Add a plotterparameter schema file: Adds a schema file for plotterparameters and the associated infrastructure for linting plotterparameter instances
    • !66 Add BEAT framework classifier
    • !67 Protocol template: This merge request implements the support for protocol templates
    • !68 Fix simple/2 database code: The migration script didn't properly copy the code.
    • !70 Add db migration scripts to installed commands
    • !71 Implement support for protocol template prototype
    • !72 Add additional information to the asset storage classes
    • !73 Fix plotterparameter writing: The write method was wrong as well as there were missing pieces in the class for it to work properly
    • !74 Implement missing prototypes handling
    • !75 Add missing class variables to plotter storage
    • !76 Reorganize error related prefix content
    • !77 Add dummy documentation to valid assets in test prefix
    • !78 Implement duplicate key handling
    • !80 Fixed external and internal documentation links
    • !81 Move all API v1 algorithms in test prefix to their own folder
    • !82 Update algorithm for loop user new types
    • !83 Implement loop block output
    • !84 Loop synchronized read
    • !85 Improve loop parts name
    • !86 Fix two_loops experiment declaration
    • !87 Remove the possibility to parametrize an analyzer in V2 as it creates unwelcome side effects
    • !88 Improve two loops testing
    • !89 Improve algorithm setup handling in local executor
    • !90 Fix loop field names
    • !91 Improve loop algorithms crosschecking for experiment
    • !92 Make loop processor related fields clearer
    • !93 Fix global parameters handling with regards to the loop blocks
    • !94 Improve two_loops test experiment by using similar setup as the loop experiment for globals and processor algorithm specific parameters to ensure the tests cover also these cases
    • !95 Modify the loop test so that it triggers the beat.backend.python#26 issue
    • !96 Improve error handling on algorithm local execution so that it provides the error that actually happened
    • !97 Remove rogue print statement: Experiment code cleanup
    • !98 Update test loop evaluator write method signature following beat.backend.python!61 the method has now a third argument containing the output currently written to on the processor side
  • v1.8.0
    Release v1.8.0
    • !51 Add automatic test image pulling if not already available
    • !52 Algorithm schema update: The schema wasn't restrictive enough
    • !53 Zmq refactoring': Improve the reliability of the system at the cost of making it slightly more complex
    • !54 Re-enable nitpicky mode and Fix broken docs
    • !56 Make default dataformat search look for more types on algorithm prototype load as the prefix generated for tests and production are a bit different
    • !57 Fix docker-pull implementation
    • !58 Worker: decode recieved job data if python older than 3.6: Prior to 3.6 the json module handles only str type
    • !59 Move license to BSD 3 Clause
    • !60 Document new ZMQ architecture
    • !61 Decode algorithm prototype content if python is older than 3.6. Same as for !58
    • !62 Code cleanup: bring the beat.core code base to the standard now used for the beat packages
    • !63 Standardize simplejson: This patch set ensure that the simplejson module is used everywher.
    • !64 Factorize simplejson: This patch set make it so that all imports for simplejson are done as json so if there's a need to change it to something else it will be easier