lostluck commented on a change in pull request #12553:
URL: https://github.com/apache/beam/pull/12553#discussion_r488240030
##########
File path: model/pipeline/src/main/proto/schema.proto
##########
@@ -32,16 +32,27 @@ option java_package = "org.apache.beam.model.pipeline.v1";
option java_outer_classname = "SchemaApi";
message Schema {
+ // List of fields for this schema. Two fields may not share a name.
repeated Field fields = 1;
+ // REQUIRED. An RFC 4122 UUID.
string id = 2;
repeated Option options = 3;
}
message Field {
+ // REQUIRED. Name of this field within the schema.
string name = 1;
+ // OPTIONAL. Human readable description of this field, such as the query
that generated it.
string description = 2;
FieldType type = 3;
+
int32 id = 4;
+ // OPTIONAL. The position of this field's data in the encoding.
+ // Either no fields in a given row are have encoding position populated,
+ // or all of them are. Used to support backwards compatibility with schema
+ // changes.
+ // Two schemas, one with encoding positions and one without, are equal
+ // if the fields when sorted by encoding position would otherwise be equal.
Review comment:
Ack. There's nothing wrong with only using strict equivalence, rather
than assignable equivalence. We do need to document which we're using at some
point or all we'll be using is strict equivalence forever. Without a definition
of assignable, we can't really use assignable equivalence at all.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
