Joris Van den Bossche created ARROW-6642: --------------------------------------------
Summary: [Python] chained access of ParquetDataset's metadata segfaults Key: ARROW-6642 URL: https://issues.apache.org/jira/browse/ARROW-6642 Project: Apache Arrow Issue Type: Bug Components: Python Reporter: Joris Van den Bossche Creating and reading a parquet dataset: {code} table = pa.table({'a': [1, 2, 3]}) import pyarrow.parquet as pq pq.write_table(table, '__test_statistics_segfault.parquet') dataset = pq.ParquetDataset('__test_statistics_segfault.parquet') dataset_piece = dataset.pieces[0] {code} If you access the metadata and a column's statistics in steps, this works fine: {code} meta = dataset_piece.get_metadata() row = meta.row_group(0) col = row.column(0) {code} but doing it chained in one step, this segfaults: {code} dataset_piece.get_metadata().row_group(0).column(0) {code} {{dataset_piece.get_metadata().row_group(0)}} still works, but additionally with {{.column(0)}} then it segfaults. -- This message was sent by Atlassian Jira (v8.3.4#803005)