Github user paul-rogers commented on the issue:
https://github.com/apache/drill/pull/932
Added a fix for repeated columns that have a low cardinality. If, say, one
row in ten has an array entry, then average cardinality (values per row) is
0.1. This was represented by an int, rounded to 0 and caused a zero-length
vector to be allocated. Drill then tried to double the length, which resulted
in 0, which was doubled again, and so on forever.
The fix has three parts:
* The "record batch sizer" uses floats to allow fractional cardinality.
* The vector initializer now works with fractional cardinality.
* If all else fails, if a fixed-width vector is asked to double from zero,
it sizes the vector to 256 bytes.
---
