Add scan min/max support for chrono types to libcudf reduction-scan (not groupby scan)

[davidwendt]

Reference #8709

Enable chrono types (timestamp, duration) support in cudf::scan with inclusive scan-type. The only requirement for min/max scan is for the type to be relationally-comparable. The code already supported string type for min/max so the change here only required generalizing that approach.

New gtests were added to scan_tests.cpp for all chrono types. Also, the to_strings specialization in the column test utilities was enhanced to show timestamp-type-specific details which is helpful when using cudf::test::print on timestamp types.

Note that this does not include groupby scan support but only reduction's scan.

[codecov]

Codecov Report

Merging #9518 (1401417) into branch-21.12 (ab4bfaa) will decrease coverage by 0.12%.
The diff coverage is n/a.

@@               Coverage Diff                @@
##           branch-21.12    #9518      +/-   ##
================================================
- Coverage         10.79%   10.66%   -0.13%     
================================================
  Files               116      117       +1     
  Lines             18869    19733     +864     
================================================
+ Hits               2036     2104      +68     
- Misses            16833    17629     +796     

Impacted Files	Coverage Δ
python/dask_cudf/dask_cudf/sorting.py	92.90% <0.00%> (-1.21%)	⬇️
python/cudf/cudf/io/csv.py	0.00% <0.00%> (ø)
python/cudf/cudf/io/hdf.py	0.00% <0.00%> (ø)
python/cudf/cudf/io/orc.py	0.00% <0.00%> (ø)
python/cudf/cudf/__init__.py	0.00% <0.00%> (ø)
python/cudf/cudf/_version.py	0.00% <0.00%> (ø)
python/cudf/cudf/core/abc.py	0.00% <0.00%> (ø)
python/cudf/cudf/api/types.py	0.00% <0.00%> (ø)
python/cudf/cudf/io/dlpack.py	0.00% <0.00%> (ø)
python/cudf/cudf/core/frame.py	0.00% <0.00%> (ø)
... and 66 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update e632f97...1401417. Read the comment docs.

[davidwendt]

Looks like we should be able to support the following for scan (inclusive):

type(s)	MIN	MAX	SUM	PRODUCT
INT8, INT16, INT32, INT64, UINT8, UINT16, UINT32, UINT64, FLOAT32, FLOAT64, BOOL8	x	x	x	x
TIMESTAMP: DAYS, SECONDS, MILLISECONDS, MICROSECONDS, NANOSECONDS	x	x
DURATION: DAYS, SECONDS, MILLISECONDS, MICROSECONDS, NANOSECONDS	x	x	x
STRING	x	x
DECIMAL32, DECIMAL64	x	x	x

All other types (DICTIONARY32, LIST, STRUCT) are not supported. The std::is_arithmetic() covers the first and last row because the type-dispatcher is called using the dispatch-storage-type and there is a special if-statement throwing an error for fixed-point and PRODUCT. The STRING row is already supported. This PR adds support for TIMESTAMP and DURATION for MIN and MAX only. That leaves one element not accounted for: DURATION and SUM.

I could include support for DURATION/SUM in this PR (including the gtests). I did not realize we were just missing one case. But I'm inclined to adding into a follow up PR instead to keep this one focused on min/max support for which the target issue is requesting.

[jrhemstad]

But I'm inclined to adding into a follow up PR instead

I'm fine with that. I would really like to see this code moved away from using a handrolled type/operator support matrix to using something like std::is_invocable to programmatically query what operations are supported on what types so we can avoid this issue where we don't support certain operations on certain types because the handrolled is_supported check was incorrect.

[davidwendt]

But I'm inclined to adding into a follow up PR instead

I'm fine with that. I would really like to see this code moved away from using a handrolled type/operator support matrix to using something like std::is_invocable to programmatically query what operations are supported on what types so we can avoid this issue where we don't support certain operations on certain types because the handrolled is_supported check was incorrect.

This one is particularly offensive because the is_supported logic is coded into the gtests too.
https://github.com/rapidsai/cudf/blob/branch-21.12/cpp/tests/reductions/scan_tests.cpp#L148-L167
I tried fixing the tests first but gave up after a few hours. I need to go back and re-watch the is_invocable talk.

[ttnghia]

So this is only reduction scan, right? And for groupby scan it is not yet supported? If so then we should clarify that in the PR title/description.

[davidwendt]

@gpucibot merge