Repository navigation
[Python] Expose RecordBatchFileReader::CountRows in Python #49305
Description
Describe the enhancement requested
Problem / Motivation
In PyArrow, pyarrow.ipc.RecordBatchFileReader the only way I found to gather the total number of rows of contained in a Feather file is to do something along the lines of:
num_rows = sum(reader.get_batch(i).num_rows for i in range(reader.num_record_batches))This is not very efficient when it seems you can directly count the rows using the metadata (as in RecordBatchFileReader::CountRows) (if I understand the code correctly?)
The current way of doing is intractable when reading from remote file systems.
Proposed solution
Expose RecordBatchFileReader::CountRows in Python.
References
arrow/cpp/src/arrow/ipc/reader.h
Line 204 in 76f7815
Thank you for reading my suggestion and all the amazing work!!
Component(s)
Python
Activity
AlenkaF commentedon Mar 27, 2026
Thank you for opening the issue @adrien-grl !
From a quick look of the code I would agree. This would be nice to have in PyArrow and looks like a straightforward binding to add with all the info in the above description already figuring out important steps.
Would you be interested in submitting a PR?
GujaLomsadze commentedon Jul 26, 2026
Hi @AlenkaF, @adrien-grl.
Seems like this issue has been stale for quite a bit. I would love to take this.
apacheGH-49305: [Python] Drop redundant closed-reader check in count_…
pitrou commentedon Jul 30, 2026
Issue resolved by pull request 50646
#50646
apacheGH-49305: [Python] Expose RecordBatchFileReader.count_rows (apa…