A metadata toolkit written in Python

Recap reads and converts schemas in dozens of formats including Parquet, Protocol Buffers, Avro, and JSON schema, BigQuery, Snowflake, and PostgreSQL.

• Read schemas from filesystems, object stores, and databases.
• Convert schemas between Parquet, Protocol Buffers, Avro, and JSON schema.
• Generate CREATE TABLE DDL from schemas for popular database SQL dialects.
• Infer schemas from unstructured data like CSV, TSV, and JSON.

• Any SQLAlchemy-compatible database
• Any fsspec-compatible filesystem
• Parquet, Protocol Buffers, Avro, and JSON schema
• CSV, TSV, and JSON files

pip install recap-core

Read schemas from objects:

s = from_proto(message)

Or files:

s = schema("s3://corp-logs/2022-03-01/0.json")

Or databases:

s = schema("snowflake://ycbjbzl-ib10693/TEST_DB/PUBLIC/311_service_requests")

And convert them to other formats:

to_json_schema(s)

{
  "type": "object",
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "properties": {
    "id": {
      "type": "integer"
    },
    "name": {
      "type": "string"
    }
  },
  "required": [
    "id"
  ]
}

Or even CREATE TABLE statements:

s = schema("/tmp/data/file.json")
to_ddl(s, "my_table", dialect="snowflake")

CREATE TABLE "my_table" (
  "col1" BIGINT,
  "col2" STRUCT<"col3" VARCHAR>
)

See the Quickstart page to get started.