Hi all,

   I came across the GitHub issue #17743 and wanted to bring it to the list
since it seems a useful spec-level improvement.

  The manifest_list today summaries counts and row counts per manifest but
not bytes.

  file_size_in_bytes already exists in every data file entry inside the
manifest body. Rolling it up the manifest_level is the natural next step,
it’s hasn’t been done. Getting total live bytes for a table today requires
opening and decompressing manifest body.

 Proposal :
add three optional long fields to manifest list entires

added_files_size_in_bytes
existing_files_size_in_bytes
deleted_files_size_in_bytes

Write path cost is zero, the use case could be table size monitoring
without manifest scans.

also v4 Is the right moment to close this gap and give planners reliable
byte level summaries at the manifest level.

Happy to contribute the spec and the implementation if there is consensus
to move forward.

Thanks
Hemanth Boyina

Reply via email to