From 108f08611f2bdf3a4b96694b4d175411fa6ea18e Mon Sep 17 00:00:00 2001 From: Ben McDonald <46734217+bmcdonald3@users.noreply.github.com> Date: Tue, 27 Aug 2024 18:41:00 -0700 Subject: [PATCH] Closes #3695: Fix Parquet Fixed-Length Code Path for Incorrect File Size Assumption (#3696) * Fix Parquet Fixed-Length Code Path for Incorrect File Size Assumption The current implementation of the fixed-length code path for reading Parquet files incorrectly assumes that all files within a dataset have the same number of strings. This can lead to incorrect results or unexpected behavior when dealing with datasets where files have varying numbers of strings. * Remove accidental + --- src/ParquetMsg.chpl | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/src/ParquetMsg.chpl b/src/ParquetMsg.chpl index 6943dee93d..160160bedc 100644 --- a/src/ParquetMsg.chpl +++ b/src/ParquetMsg.chpl @@ -1029,7 +1029,8 @@ module ParquetMsg { byteSizes = calcStrSizesAndOffset(entrySeg.a, filenames, sizes, dsetname); } else { entrySeg.a = fixedLen; - byteSizes = fixedLen*len; + for i in sizes.domain do + byteSizes[i] = fixedLen*sizes[i]; } entrySeg.a = (+ scan entrySeg.a) - entrySeg.a;