emkornfield commented on code in PR #242:
URL: https://github.com/apache/parquet-format/pull/242#discussion_r1603959474
##########
src/main/thrift/parquet.thrift:
##########
@@ -1165,6 +1317,62 @@ struct FileMetaData {
9: optional binary footer_signing_key_metadata
}
+/** Metadata for a column in this file. */
+struct FileColumnMetadataV3 {
+ /** All column chunks in this file (one per row group) **/
+ 1: required list<ColumnChunkV3> columns
Review Comment:
So I am not advocating full shredding (e.g. a sub parquet schema for each
one of these elements), I agree that would be overdone and most likely more of
a perfomance hindrance.
I do think moving away from using thrifts lists as much as possible will
help significantly in overhead reduction. The issue here is that you cannot
selectively decode part of lists in an easy manner. I think a design goal here
is we should be trying to make deserialization cost O(columns scanned) vs
O(number of column in schema)
--
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]
