sql/pgwire: use OIDs when encoding some datum types #34818
Conversation
|
This change is |
There was a problem hiding this comment.
Reviewed 8 of 8 files at r1.
Reviewable status:complete! 1 of 0 LGTMs obtained (waiting on @bobvawter)
There was a problem hiding this comment.
Reviewable status:
pkg/sql/pgwire/command_result.go, line 81 at r1 (raw file):
formatCodes []pgwirebase.FormatCode oids []oid.Oid
Document this. I'm assuming it's a map of column index to Oid like formatCodes?
pkg/sql/pgwire/types.go, line 248 at r1 (raw file):
b.putInt32(8) b.putInt64(int64(*v)) default:
Do you need to detect the Oid == 0 case in order for the doc comment to be correct?
Teach pgwire how to encode int and float with the various widths. Do this by saving the oids during SetColumns. Teach cmd/generate-binary to also record oids and use those in tests. Fix varbits to expect the same OID as postgres produces. Sadly, our int2 and int4 types don't yet propagate all the way down and so we still encode them as an int8. This commit is a precursor to supporting that. Release note: None
There was a problem hiding this comment.
Reviewable status:
pkg/sql/pgwire/command_result.go, line 81 at r1 (raw file):
Previously, bobvawter (Bob Vawter) wrote…
Document this. I'm assuming it's a map of column index to
OidlikeformatCodes?
Done.
pkg/sql/pgwire/types.go, line 248 at r1 (raw file):
Previously, bobvawter (Bob Vawter) wrote…
Do you need to detect the
Oid == 0case in order for the doc comment to be correct?
No, the 0 case is for types that have a 1:1 mapping. The oid must be specified otherwise. I've updated the comment to reflect this.
|
bors r+ |
34818: sql/pgwire: use OIDs when encoding some datum types r=mjibson a=mjibson Teach pgwire how to encode int and float with the various widths. Do this by saving the oids during SetColumns. Teach cmd/generate-binary to also record oids and use those in tests. Fix varbits to expect the same OID as postgres produces. Sadly, our int2 and int4 types don't yet propagate all the way down and so we still encode them as an int8. This commit is a precursor to supporting that. Release note: None 34830: c-deps: bump RocksDB to pick up ingest-external-file deadlock r=tbg a=petermattis Fixes #34212 Release note (bug fix): Fix a deadlock that could occur during IMPORT and RESTORE, causing all writes on a node to be stopped. Co-authored-by: Matt Jibson <[email protected]> Co-authored-by: Peter Mattis <[email protected]>
Build succeeded |
Teach pgwire how to encode int and float with the various widths. Do
this by saving the oids during SetColumns.
Teach cmd/generate-binary to also record oids and use those in tests.
Fix varbits to expect the same OID as postgres produces.
Sadly, our int2 and int4 types don't yet propagate all the way down
and so we still encode them as an int8. This commit is a precursor to
supporting that.
Release note: None