sankarreddy-atlan opened a new issue, #3536: URL: https://github.com/apache/iceberg-rust/issues/3536
### Apache Iceberg Rust version main @ 26356f5 ### Describe the bug parquet-rs truncates string and binary column chunk statistics to 64 bytes by default and marks them inexact. `MinMaxColAggregator::update` skips inexact statistics. #3464 shows this drops the bounds of a file with one row group. With more than one row group, the bounds are wrong. The file bound comes only from the row groups with exact statistics, so it can exclude values in the file. A scan that prunes by file bounds then skips the file and drops rows. ### To Reproduce Write a `string` column with `max_row_group_row_count = 2` and values `["a" * 100, "c", "b", "z" * 100]`. The data file gets `lower_bound = "b"` and `upper_bound = "c"`. The filter `s < 'b'` skips the file, but the file contains `"aaa…"`. ### Expected behavior Each bound holds for all values in the file, as the spec requires. A truncated min is a prefix of the min, and a truncated max is incremented, so both are valid bounds. Java's `ParquetMetrics` merges row group min and max with no exactness check, and drops the bounds when a row group has no statistics. ### Willingness to contribute I can contribute a fix for this bug independently -- 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]
