The second of four PRs, as part of dataset configuration file validation. To be merged second. Updates schema files with:
required fields
recommended fields
ifabsent (default value) fields
min/max values (for numerical schema fields)
new types (e.g., FloatFormattedString)
new schema classes based on annotation.sources refactor
removing no longer used schema classes
some deprecated notes
Updates were based on template.yaml. Also improves / fixes schema.py schema generation. Metadocs/ folder is re-make-d.
Files that were primarily changed (and are not code-gen'd files):
schema/v1.1.0/schema.py
schema/v1.1.0/metadata.yaml
schema/v1.1.0/common.yaml
Note that any_of is not currently working for LinkML Pydantic code generation. So some schema updates are commented out and waiting on it. Specifically, we cannot use the newly created types (like FloatFormattedString) accept both formatted strings and floats/ints for fields, so we plan on parsing the formatted string before running validation as a temporary workaround for now.
Note that the Pydantic models were also generated with a temporary-hotfix that hasn't been yet implemented by linkml (see https://github.com/linkml/linkml/pull/2201). If the docs are regenerated without this hotfix, validation will be incorrect, but only more lax (required multivalued fields that are not provided in config files will not be correctly reported as errors). In the near future, I can look into creating a Python wrapper class / using jinja2 templates to extend the current LinkML release to provide a consistent fix.
The second of four PRs, as part of dataset configuration file validation. To be merged second. Updates schema files with:
requiredfieldsrecommendedfieldsifabsent(default value) fieldsmin/maxvalues (for numerical schema fields)FloatFormattedString)Updates were based on
template.yaml. Also improves / fixesschema.pyschema generation. Metadocs/ folder is re-make-d.Files that were primarily changed (and are not code-gen'd files):
schema/v1.1.0/schema.pyschema/v1.1.0/metadata.yamlschema/v1.1.0/common.yamlNote that
any_ofis not currently working for LinkML Pydantic code generation. So some schema updates are commented out and waiting on it. Specifically, we cannot use the newly created types (likeFloatFormattedString) accept both formatted strings and floats/ints for fields, so we plan on parsing the formatted string before running validation as a temporary workaround for now.Note that the Pydantic models were also generated with a temporary-hotfix that hasn't been yet implemented by linkml (see https://github.com/linkml/linkml/pull/2201). If the docs are regenerated without this hotfix, validation will be incorrect, but only more lax (required multivalued fields that are not provided in config files will not be correctly reported as errors). In the near future, I can look into creating a Python wrapper class / using jinja2 templates to extend the current LinkML release to provide a consistent fix.