[
https://issues.apache.org/jira/browse/BEAM-6611?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16780110#comment-16780110
]
Pablo Estrada commented on BEAM-6611:
-------------------------------------
Hello [~ttanay]
glad you're interested. A couple good pointers:
- This transform would be in the Python SDK, and it would appear inside
`WriteToBigQuery`. Note that `WriteToBigQuery` has two methods (Streaming
inserts or batch loads). Currently, Batch loads only work in batch
[[link|https://github.com/apache/beam/blob/master/sdks/python/apache_beam/io/gcp/bigquery.py#L831-L867]]
- A similar transform exists in the Java SDK. It's applied called
BatchLoads.java, specifically the
[expandTriggered|https://github.com/apache/beam/blob/master/sdks/java/io/google-cloud-platform/src/main/java/org/apache/beam/sdk/io/gcp/bigquery/BatchLoads.java#L247]
transform.
> A Python Sink for BigQuery with File Loads in Streaming
> -------------------------------------------------------
>
> Key: BEAM-6611
> URL: https://issues.apache.org/jira/browse/BEAM-6611
> Project: Beam
> Issue Type: Improvement
> Components: sdk-py-core
> Reporter: Pablo Estrada
> Assignee: Pablo Estrada
> Priority: Major
> Labels: gsoc, gsoc2019, mentor, triaged
>
> The Java SDK supports a bunch of methods for writing data into BigQuery,
> while the Python SDK supports the following:
> - Streaming inserts for streaming pipelines [As seen in [bigquery.py and
> BigQueryWriteFn|https://github.com/apache/beam/blob/master/sdks/python/apache_beam/io/gcp/bigquery.py#L649-L813]]
> - File loads for batch pipelines [As implemented in [PR
> 7655|https://github.com/apache/beam/pull/7655]]
> Qucik and dirty early design doc: https://s.apache.org/beam-bqfl-py-streaming
> The Java SDK also supports File Loads for Streaming pipelines [see BatchLoads
> application|https://github.com/apache/beam/blob/master/sdks/java/io/google-cloud-platform/src/main/java/org/apache/beam/sdk/io/gcp/bigquery/BigQueryIO.java#L1709-L1776].
> File loads have the advantage of being much cheaper than streaming inserts
> (although they also are slower for the records to show up in the table).
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
