I have troubles getting BigQuery to load timestamps from avro-files correctly.
The avro-files have date columns stored as long, with logical type timestamp-micros. As per documentation, BigQuery should store this as timestamp data type. I have also tried timestamp-millis for logical type.
The data is stored in avro like this:
{'id': '<masked>', '<masked>': '<masked>', 'tm': 1553990400000, '<masked>': <masked>, '<masked>': <masked>, 'created': 1597056958864}
The fields tm and created are longs, 2019-03-31T00:00:00Z and 2020-08-10T11:50:58.986816592Z respectively.
The schema for the avro is
{"type":"record","name":"SomeMessage","namespace":"com.df",
"fields":
[{"name":"id","type":"string"},
{"name":"<masked>","type":"string"},
{"name":"tm","type":"long","logicalType":"timestamp-micros"},
{"name":"<masked>","type":"int"},
{"name":"<masked>","type":"float"},
{"name":"created","type":"long","logicalType":"timestamp-micros"}]}";
When imported to BigQuery through bq load, records ends up like these:
<masked> <masked> tm <masked> <masked> created
________________________________________________________________________________________________________
<masked> | <masked> | 1970-01-18 23:39:50.400 UTC | <masked> | <masked> | 1970-01-19 11:37:36.958864 UTC
________________________________________________________________________________________________________
The import command used is:
bq load --source_format=AVRO --use_avro_logical_types some_dataset.some_table "gs://some-bucket/some.avro"
The timestamps in BigQuery are nowhere near the actual values provided in avro.
Anyone have any ideas on how to do this properly?
I figured out that the avro schema is actually wrong. The timestamp fields should be like this:
{"name":"created","type":{"type":"long", "logicalType":"timestamp-millis"}}