-
Type:
Sub-task
-
Resolution: Fixed
-
Priority:
Minor
-
Affects Version/s: 4.4.0
-
Component/s: SQL
Most decoding errors thrown by the vectorized Parquet reader while reading a batch, e.g. the ParquetDecodingExceptions from VectorizedRleValuesReader (SPARK-59832) and from the PLAIN, DELTA_LENGTH_BYTE_ARRAY and DELTA_BYTE_ARRAY readers (SPARK-59812), do not say which column is corrupt. The error only names the file, which makes a corrupt page hard to locate in a wide table.
Errors thrown while initializing a page are already wrapped with the page and the column by VectorizedColumnReader (readPageV1/readPageV2/initDataReader). Decoding errors thrown later from VectorizedColumnReader.readBatch could be wrapped the same way, e.g. by catching ParquetDecodingException there and adding the column descriptor.
This came up in the review of SPARK-59832 (https://github.com/apache/spark/pull/59117).
- is related to
-
SPARK-60316 Validate the dictionary page against its size when reading Parquet
-
- Open
-
-
SPARK-60317 Report out-of-range dictionary ids in the vectorized Parquet reader
-
- Open
-
- links to