Data Contract CLI Files

[1.2.0] - 2026-09-08

This release adds support for the Open Data Contract Standard v3.2.0 (#1557).

Fixed

• Iceberg testing uses AWS SSO credentials for S3 Tables data files and supports logical table aliases.
• Iceberg catalog imports supply a catalog name, preserve qualified table identifiers, and correctly map temporal and binary fields.
• datacontract test checks vector element types when the data source reports them (#1585)

Added

• datacontract test resolves variables in property names, enum values, nested type options, library quality arguments, and service levels (#1583)
• Lint requires vector dimensions and rejects duplicate enum values regardless of their labels (#1586)
• Contracts declaring apiVersion: v3.2.0 lint and round-trip, including enum, map, vector, semanticType, synonyms, deprecated, context, variables in string values, and the new server types (#1558)
• datacontract export avro-idl writes map<...> and array<float> for map and vector properties; datacontract export sqlalchemy writes JSON and ARRAY(Float) (#1558)
• datacontract test resolves ${VAR} and ${VAR:-default} references in server fields and SQL quality queries from the environment; an unset variable without a default fails the run with its name, and export keeps the references (#1559)
• Server port may be a string such as ${DB_PORT}, including in Excel imports; config files accept ${VAR:-default} (#1559)
• enum on properties is read by datacontract test, the jsonschema, avro, avro-idl, protobuf, pydantic-model, dcs, great-expectations, sodacl and data-caterer exporters, and dbt test mapping, ahead of logicalTypeOptions.enum, the enum custom property and the invalidValues rule; HTML export lists the values with labels and descriptions (#1560)
• datacontract import from JSON Schema, Avro, Protobuf and DCS writes allowed values as enum entries instead of an invalidValues quality rule or custom properties (#1560)
• datacontract export pydantic-model types an enumerated string or integer property as typing.Literal (#1560)
• logicalType: map with a map block (key and value as full property definitions) is a first-class type: datacontract test checks the key and value types, including nested objects and maps, on DuckDB, Databricks, Snowflake, Trino, Kafka and the file sources (#1562)
• datacontract export writes native map types for snowflake, databricks, dataframe, duckdb (local, s3), clickhouse, trino, spark, iceberg, avro, avro-idl, protobuf, pydantic-model, go and dcs, JSON for postgres, mysql, sqlserver, oracle and bigquery, and additionalProperties for jsonschema; HTML shows the key and value types (#1562)
• datacontract import from sql, databricks/unity, spark, glue, iceberg, avro, parquet and dcs writes logicalType: map with the key and value instead of physicalType: map with custom properties; the mapKeyType, mapValueType, mapKeys and mapValues custom properties are still read but deprecated (#1562)
• logicalType: vector with logicalTypeOptions.dimensions is a first-class type: datacontract test accepts a native vector or an array of numbers and compares the dimensions when the column states them; export writes vector(n) for postgres, VECTOR(FLOAT, n) for snowflake, FLOAT[n] for duckdb, vector(n) for mysql, arrays of floats for databricks, dataframe, trino, clickhouse, bigquery, spark, iceberg, avro, avro-idl, protobuf, pydantic-model, go and dcs, and a fixed-length array of numbers for jsonschema (#1563)
• datacontract import from sql, postgres and snowflake reads vector(n), halfvec(n) and VECTOR(FLOAT, n) columns as logicalType: vector with their dimensions, and parquet reads a fixed-size list of floats the same way (#1563)
• datacontract export html renders semanticType, synonyms, deprecated and context on schema objects and properties, the contract-level context, vendor on custom properties, and customProperties and authoritativeDefinitions on SLA entries; export rdf writes enum entries, synonyms, map definitions, customProperties and other nested definitions as nodes of their own instead of dropping them or printing a model repr (#1561)
• datacontract changelog matches synonyms, enum entries, context statements and constraints, and the nested lists of SLA entries by their natural key, and relationships by id when present, instead of by position (#1561)
• datacontract test reads CSV files from local, s3, gcs and azure servers and Kafka JSON messages in the server's declared encoding, and uses the Athena workgroup (also DATACONTRACT_ATHENA_WORKGROUP), which makes stagingDir optional (#1564)
• The ODCS v3.2.0 server types hana, iceberg, exasol, teradata, ingres, vectorwise, versant and poet lint and export; test explains that it cannot connect to them yet, and the synonyms fastobjects and btrieve resolve to poet and zen (#1564)
• datacontract export sql --sql-server-type auto warns before falling back to the snowflake dialect for a server type without one (#1564)
• datacontract test supports the iceberg server type: tables are read from the REST catalog named by catalogUrl, namespace and warehouse with pyiceberg (credentials via DATACONTRACT_ICEBERG_CREDENTIAL or DATACONTRACT_ICEBERG_TOKEN, data files via the S3 options; DATACONTRACT_ICEBERG_CATALOG_TYPE selects a sql, glue or hive catalog instead of rest, DATACONTRACT_ICEBERG_S3_ENDPOINT points at an S3-compatible store, Amazon S3 Tables and the Glue REST endpoint are signed with SigV4 from the AWS credentials, DATACONTRACT_ICEBERG_PROPERTIES passes further catalog properties through), and datacontract import iceberg --catalog-url creates a contract from a catalog table with a ready-to-test server (#1565)

Changed

• datacontract breaking detects enum restrictions and vector shape changes (#1584)
• datacontract init and all importers write apiVersion: v3.2.0 (#1558)
• The bundled Data Contract Editor (datacontract edit) is updated to 0.1.13 (#1566)
• open-data-contract-standard dependency bumped to 3.2.x (#1558)
• Lint accepts stable ids with any character except whitespace and .#/\@!%&^, as in the released ODCS v3.2.0 schema