rdblue commented on a change in pull request #3369:
URL: https://github.com/apache/iceberg/pull/3369#discussion_r735759839
##########
File path:
spark/v3.2/spark/src/main/java/org/apache/iceberg/spark/source/SparkTable.java
##########
@@ -212,33 +220,44 @@ private String getRowLevelOperationMode(String operation)
{
@Override
public boolean canDeleteWhere(Filter[] filters) {
- if (table().specs().size() > 1) {
- // cannot guarantee a metadata delete will be successful if we have
multiple specs
- return false;
- }
-
- Set<Integer> identitySourceIds = table().spec().identitySourceIds();
- Schema schema = table().schema();
+ Expression deleteExpr = Expressions.alwaysTrue();
for (Filter filter : filters) {
- // return false if the filter requires rewrite or if we cannot translate
the filter
- if (requiresRewrite(filter, schema, identitySourceIds) ||
SparkFilters.convert(filter) == null) {
+ Expression expr = SparkFilters.convert(filter);
+ if (expr != null) {
+ deleteExpr = Expressions.and(deleteExpr, expr);
+ } else {
return false;
}
}
- return true;
+ return deleteExpr == Expressions.alwaysTrue() ||
canDeleteUsingMetadata(deleteExpr);
}
- private boolean requiresRewrite(Filter filter, Schema schema, Set<Integer>
identitySourceIds) {
- // TODO: handle dots correctly via v2references
- // TODO: detect more cases that don't require rewrites
- Set<String> filterRefs = Sets.newHashSet(filter.references());
- return filterRefs.stream().anyMatch(ref -> {
- Types.NestedField field = schema.findField(ref);
- ValidationException.check(field != null, "Cannot find field %s in
schema", ref);
- return !identitySourceIds.contains(field.fieldId());
- });
+ // a metadata delete is possible iff matching files can be deleted entirely
+ private boolean canDeleteUsingMetadata(Expression deleteExpr) {
Review comment:
Looks like this pretty much just runs the delete logic to see if it will
succeed.
I was thinking about something slightly different, which is to check whether
the strict projection and the inclusive projection of the delete filter are
equivalent. If so, then we know that the delete aligns with partitioning and
will succeed. I'm okay with this if we don't think it's going to be too
expensive during planning though.
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
