Fix exclusive scan when including nulls and improve testing #8478
Conversation
Codecov Report
@@ Coverage Diff @@
## branch-21.08 #8478 +/- ##
===============================================
Coverage ? 83.00%
===============================================
Files ? 109
Lines ? 18215
Branches ? 0
===============================================
Hits ? 15119
Misses ? 3096
Partials ? 0 Continue to review full report at Codecov.
|
harrism
added
bug
cpp/include/cudf_test/type_lists.hpp
Outdated
| @@ -85,15 +86,21 @@ typename std::enable_if<cudf::is_fixed_width<TypeParam>() && | |||
| make_type_param_vector(std::initializer_list<T> const& init_list) | |||
| { | |||
| thrust::host_vector<TypeParam> vec(init_list.size()); | |||
| std::transform(std::cbegin(init_list), std::cend(init_list), std::begin(vec), [](auto const& e) { | |||
| if (std::is_unsigned<TypeParam>::value) | |||
| std::transform(std::cbegin(init_list), std::cend(init_list), std::begin(vec), [](T const& e) { | |||
There was a problem hiding this comment.
This is interesting. Why did we need the T to be explicit?
There was a problem hiding this comment.
We don't. Fixed, thanks. Also changed from if constexpr {return} else {return} to if constexpr {return} return to workaround known CUDA compiler issue (spurious warning I just noticed).
|
@gpucibot merge |
Fixes #8462 by generalizing the existing
mask_inclusive_scanused ininclusive_scanso it can be applied the same way inexclusive_scan. #8462 demonstrated holes in test coverage, so this PR also reorganizesscan_teststest infrastructure to enable a single set of tests to be applied to all supported data types and parameters, correctly expecting exceptions in unsupported cases (e.g. no product scans on fixed-point, only min/max inclusive scans on strings).