Gen3 Validator

Gen3 Validator is a Python toolkit designed to make working with Gen3 metadata schemas and data validation straightforward for developers.

Installation

pip install gen3_validator
pip show gen3_validator

Docs

• example usage

Quickstart

import gen3_validator

resolver = gen3_validator.ResolveSchema(schema_path = "../tests/schema/gen3_test_schema.json")
resolver.resolve_schema()
schema = resolver.schema_resolved

data = [
    {
        "baseline_timepoint": True, # variable not in data dictionary
        "freeze_thaw_cycles": "10", # should be an integer
        "sample_collection_method": "2fddbe7d09",
        "sample_id": "d4f31f7bb6",
        "sample_in_preservation": "snap Frozen",
        "sample_in_storage": "yes",
        "sample_provider": "USYD",
        "sample_source": "UBERON:3781554",
        "sample_storage_method": "not stored",
        "sample_type": "59a8fd8005",
        "storage_location": "UMELB",
        "subjects": {
            "submitter_id": "subject_e5616257f8"
        },
        "submitter_id": "sample_efdbe56d20",
        "type": "sample"
    },
    {
        "baseline_timepoint": True, 
        "freeze_thaw_cycles": 76,
        "sample_collection_method": "e2a6403b51",
        "sample_id": 3324635, # should be a string
        "sample_in_preservation": "not allowed to collect",
        "sample_in_storage": "unknown",
        "sample_provider": "USYD",
        "sample_source": "UBERON:9332357",
        "sample_storage_method": "frozen, liquid nitrogen",
        "sample_type": "8fd28ec2f3",
        "storage_location": "Baker",
        "subjects": {
            "submitter_id": "subject_071bc3e81a"
        },
        "submitter_id": "sample_f7645c1221",
        "type": "sample"
    }
]
results = gen3_validator.validate.validate_list_dict(data, schema)

print(results)

Example output:

[
    {
        'node': 'sample',
        'index': 0,
        'validation_result': 'FAIL',
        'invalid_key': 'freeze_thaw_cycles',
        'schema_path': 'properties.freeze_thaw_cycles.type',
        'validator': 'type',
        'validator_value': 'integer',
        'validation_error': "'10' is not of type 'integer'"
    },
    {
        'node': 'sample',
        'index': 0,
        'validation_result': 'FAIL',
        'invalid_key': 'root',
        'schema_path': 'additionalProperties',
        'validator': 'additionalProperties',
        'validator_value': False,
        'validation_error': "Additional properties are not allowed ('baseline_timepoint', 'subjects' were unexpected)"
    },
    {
        'node': 'sample',
        'index': 1,
        'validation_result': 'FAIL',
        'invalid_key': 'sample_id',
        'schema_path': 'properties.sample_id.type',
        'validator': 'type',
        'validator_value': 'string',
        'validation_error': "3324635 is not of type 'string'"
    },
    {
        'node': 'sample',
        'index': 1,
        'validation_result': 'FAIL',
        'invalid_key': 'root',
        'schema_path': 'additionalProperties',
        'validator': 'additionalProperties',
        'validator_value': False,
        'validation_error': "Additional properties are not allowed ('baseline_timepoint', 'subjects' were unexpected)"
    }
]

---

Bulk Folder Validation

Instead of assembling a list of objects by hand, you can point Gen3 Validator at a folder that holds one JSON file per node plus a data import order file, and it will validate every node in import order, reusing the same schema validation shown above. It also checks the links between nodes (reference integrity) by default — see below.

Expected folder layout

my_submission/
├── DataImportOrder.txt      # node names, one per line, in the order to process them
├── project.json             # a single JSON object  (one record)
├── subject.json             # a JSON array of record objects
├── sample.json
└── ... one <node>.json per node

• Each <node>.json is either a JSON array of records or a single object (e.g. project.json); single objects are treated as a one-record list.
• Every record must carry a "type" field equal to its node name (same requirement as the in-memory API).
• DataImportOrder.txt lists node names, one per line. A numbered format (1<TAB>project, 2<TAB>subject, ...) is also accepted.

Python API

import gen3_validator

# Convenience: resolve the schema from a path and validate the folder in one call.
results = gen3_validator.validate_data_folder_from_schema(
    folder_path="path/to/my_submission",
    schema_path="path/to/gen3_schema.json",
)

# Or, if you already have a resolved schema, reuse it:
resolver = gen3_validator.ResolveSchema(schema_path="path/to/gen3_schema.json")
resolver.resolve_schema()
results = gen3_validator.validate_data_folder("path/to/my_submission", resolver.schema_resolved)

The output is a single flat list of failures, ordered by import order. Each row is the same shape as the in-memory validator's output plus a source_file field naming the file the record came from:

[
    {
        'node': 'project',
        'index': 0,                       # index of the record within its node file
        'validation_result': 'FAIL',
        'invalid_key': 'root',
        'schema_path': 'additionalProperties',
        'validator': 'additionalProperties',
        'validator_value': False,
        'validation_error': "Additional properties are not allowed ('data_release', 'data_release_date' were unexpected)",
        'source_file': 'project.json'
    },
    ...
]

A node listed in the import order with no matching file is skipped with a warning, and a *.json file not listed in the import order is ignored with a warning. If a node file cannot be loaded or validated (invalid JSON, a record missing "type", etc.) a single row with 'validation_result': 'ERROR' is emitted for that node and processing continues — one bad file never aborts the run.

Link / reference integrity

By default the folder validator also checks that the links between nodes resolve: for every record, each link reference must point to a record that actually exists in the target node's file. In Gen3 a child links up to a parent via a property named after the parent (e.g. a sample links to a clinical_descriptor via "clinical_descriptors": {"submitter_id": "..."}), and the reference may be a single object or an array of them. Most nodes are referenced by submitter_id; project is referenced by code.

A dangling reference is reported as a row with validator: "link":

{
    'node': 'sample',
    'index': 0,                       # index of the record within its node file
    'validation_result': 'FAIL',
    'invalid_key': 'clinical_descriptors',   # the link property
    'schema_path': 'links',
    'validator': 'link',
    'validator_value': 'clinical_descriptor',  # the target node
    'validation_error': "Link 'clinical_descriptors' references clinical_descriptor "
                        "'clinical_descriptor_MISSING' (by submitter_id) but no matching record "
                        "exists in clinical_descriptor.json",
    'source_file': 'sample.json'
}

Rules:

• A link whose target node is absent from the folder (e.g. project links to program, but there is no program.json) is skipped with a warning — there is nothing to validate against.
• A link into a node that is present but empty is reported as a failure.
• Disable link checking with the check_links=False parameter (Python API) or the --no-link-check flag (CLI) to validate schemas only.

results = gen3_validator.validate_data_folder_from_schema(
    folder_path="path/to/my_submission",
    schema_path="path/to/gen3_schema.json",
    check_links=False,   # schema validation only
)

Command line

Installing the package also exposes a gen3-validate command:

gen3-validate path/to/my_submission -s path/to/gen3_schema.json

Flag	Description
-s, --schema	Path to the Gen3 JSON schema (required).
--order-file	Import order filename within the folder (default DataImportOrder.txt).
-o, --output	Write the JSON report to a file instead of stdout.
--no-link-check	Disable cross-node reference integrity checks (validate schemas only).
-v, --verbose	Verbose (INFO-level) logging.

The report is printed as JSON to stdout (or written with -o). The exit code is 0 when the folder is clean, 1 when any record is a FAIL or ERROR, and 2 for input errors (e.g. a missing import order file). This makes it convenient as a pass/fail gate in scripts and CI.

---

Dev Setup

1. Make sure you have poetry installed.
2. Clone the repository.
3. Run the following command to activate the virtual environment.

eval $(poetry env activate)

4. Run the following command to install the dependencies.

poetry install

5. Run the following command to run the tests.

pytest -vv tests/

---

License

See the license page for more information.