Repository navigation
Allow file reader to read without setting ReaderOptions::projection #514
Description
Currently ReaderOptions::projection (https://github.com/apache/iceberg-cpp/blob/main/src/iceberg/file_reader.h#L111) is required to be set before opening a parquet/avro reader, otherwise an error will be returned complaining about this. It would be good to allow reading all columns from the file if this field is unset.
Activity
dheeraj12347 commentedon Jan 21, 2026
Hi @wgtmac, I’d like to work on this enhancement.
My plan is:
– Update the Parquet/Avro file reader so that when ReaderOptions::projection is unset, it defaults to reading all columns from the file schema instead of returning an error.
– Add tests to verify reading works both with and without an explicit projection.
Does this approach align with what you have in mind, especially for any edge cases?
wgtmac commentedon Jan 21, 2026
This sounds perfect. Thanks for picking it up!
apacheGH-514: [Parquet] Infer schema when projection is null in Parqu…
apacheGH-514: [Parquet] Infer schema when projection is null in Parqu…
apacheGH-514: [Parquet] Infer schema when projection is null in Parqu…
apacheGH-514: [Parquet] Infer schema when projection is null in Parqu…
apacheGH-514: [Parquet] Infer schema when projection is null in Parqu…
apacheGH-514: [Parquet] Infer schema when projection is null in Parqu…
baha-bouali commentedon Oct 18, 2026
I would love to work on this please
wgtmac commentedon Oct 19, 2026
@baha-bouali Thanks for your interest! There was an initial attempt which turns out to be a large and complicated change. Feel free to create a PR if you have nice and clean approach.
feat(avro,parquet): read the full file schema when no projection is set
wgtmac commentedon Nov 3, 2026
Close this as explained in #1055 (comment)