ETL Output Dataclasses

[mawelborn]

An initial implementation of ETL Output Dataclasses for working with OCR text, tokens, and tables that fits nicely with Result File Dataclasses. Supports loading both result file v1 and v3 and ReadAPI v2 and v3 submission ETL Output via sync and async reader functions.

def read_url(url: str) -> object:
   return client.call(RetrieveStorageObject(url))

result = results.load(submission.result_file, reader=read_url)
etl_outputs = {
   document: etloutput.load(document.etl_output_url, reader=read_url, tables=True)
   for document in result.documents
}

extraction = result.predictions.document_extractions[0]
etl_output = etl_outputs[extraction.document]
token = etl_output.token_for(extraction)
table, cell = etl_output.table_cell_for(token)