ReadFromBigQuery doesn't support DATETIME type

[damccorm]

ReadFromBigQuery with Avro export by documentation should produce a python native object i.e. datetime.datetime. 

> the BigQuery types for DATE, DATETIME, TIME, and TIMESTAMP will be exported as strings. This behavior is consistent with BigQuerySource. When using Avro exports, these fields will be exported as native Python types (datetime.date, datetime.datetime, datetime.datetime, and datetime.datetime respectively)

 

However, in practice this doesn't happen as BigQuery doesn't have a type that maps to Avro logical type.

 

> Note: There is no logical type that directly corresponds to DATETIME, and BigQuery doesn't support any direct conversion from an Avro type into a DATETIME field.

 

I've also checked manually the AVRO export file that get exported to GCS. The logical type in Avro schema for BQ is datetime.

{"name":"datetime_col","type":["null",{"type":"string","logicalType":"datetime"}]}

 

Avro spec doesn't support datetime logical type[3] nor FastAvro library[4] that is used underneath to read the avro files.

 

Resources:

[1] https://beam.apache.org/releases/pydoc/2.29.0/apache_beam.io.gcp.bigquery.html#apache_beam.io.gcp.bigquery.ReadFromBigQuery

[2] https://cloud.google.com/bigquery/docs/loading-data-cloud-storage-avro#logical_types

[3] https://avro.apache.org/docs/1.8.0/spec.html#Logical****Types|https://avro.apache.org/docs/1.8.0/spec.html#Logical****Types]

[4] https://fastavro.readthedocs.io/en/latest/logical_types.html

 

Imported from Jira BEAM-12375. Original Jira may contain additional context.
Reported by: sayatez.

[lostluck]

This is definitely IO related, but it's not Go, given it's linking to python. Re-tagging.