Fixes nonlocal reducers in which the first list is empty.

[jpivarski]

This error was reported in #1371, and the fix is in the kernel:

   int64_t j = 0;
   int64_t k = 0;
-  int64_t maxdistinct = -1;
+  int64_t maxdistinct = -2;
   int64_t lasti = -1;
   for (int64_t i = 0;  i < lendistincts;  i++) {
     if (maxdistinct < distincts[i]) {

The initial value of maxdistinct is supposed to be different from all possible values so that the first item is identified as a distinct list, but empty lists are represented in the distincts array as -1. Setting the initial value to a smaller number allows it to trigger on the first list, even if that list is empty.

This bug goes way back: in the first implementation of this kernel (Feb 15 2020)

https://github.com/scikit-hep/awkward-1.0/blob/b4ea752b641e1162e6e3ac9b2778ac544a52d5ff/src/cpu-kernels/reducers.cpp#L462-L466

and the studies that preceded it

https://github.com/scikit-hep/awkward-1.0/blob/fd94fd6978e06795f2a36aefa8b1bc2a7772387e/studies/reducers.py#L1014-L1018

Thanks, @masonproffitt!

[codecov]

Codecov Report

Merging #1378 (3a0045f) into main (b2fd2be) will decrease coverage by 0.97%.
The diff coverage is 52.14%.

Impacted Files	Coverage Δ
src/awkward/_v2/_connect/cling.py	0.00% <0.00%> (ø)
src/awkward/_v2/_connect/cuda/__init__.py	0.00% <0.00%> (ø)
src/awkward/_v2/_lookup.py	97.50% <0.00%> (ø)
src/awkward/_v2/_prettyprint.py	66.09% <0.00%> (+2.29%)	⬆️
src/awkward/_v2/_typetracer.py	69.14% <0.00%> (ø)
src/awkward/_v2/identifier.py	55.69% <0.00%> (ø)
src/awkward/_v2/operations/convert/ak_from_jax.py	75.00% <0.00%> (ø)
src/awkward/_v2/operations/convert/ak_to_jax.py	75.00% <0.00%> (ø)
src/awkward/_v2/operations/io/ak_from_parquet.py	75.00% <0.00%> (ø)
src/awkward/_v2/operations/io/ak_to_parquet.py	75.00% <0.00%> (ø)
... and 145 more