ARROW-14095: [C++] subtract(timestamp, duration) -> timestamp kernel #12137
Conversation
|
I would expect that for non-matching resolution in the input, we cast to the most detailed resolution. But, this doesn't necessarily need to have kernels for all combinations of resolutions. I think the casting step before calling the kernel ("DispatchBest") should take care of that (like The same is true for |
|
@jorisvandenbossche heterogeneous time resolutions are now supported. |
|
This is now a test only change? |
|
@jorisvandenbossche It would appear test + doc only yes. ARROW-14095 added the |
docs/source/cpp/compute.rst
Outdated
| | sign | Unary | Numeric | Int8/Float32/Float64 | \(2) | | ||
| +------------------+--------+----------------------------+----------------------------+-------+ | ||
| | subtract | Binary | Numeric/Date/Duration | Numeric/Date/Duration | \(1) | | ||
| | subtract | Binary | Numeric/Timestamp/Duration | Numeric/Timestamp/Duration | \(1) | |
There was a problem hiding this comment.
Temporal instead of Timestamp?
There was a problem hiding this comment.
Changed Numeric/Timestamp/Duration -> Numeric/Temporal
| auto seconds_3_tz = ArrayFromJSON(timestamp(TimeUnit::SECOND, "UTC"), R"([3, null])"); | ||
| auto milliseconds_1k_tz = | ||
| ArrayFromJSON(timestamp(TimeUnit::MILLI, "UTC"), R"([1000, null])"); | ||
| CheckScalarBinary(op, seconds_3, milliseconds_2k, milliseconds_1k); |
|
Benchmark runs are scheduled for baseline = d8f8c09 and contender = d747326. d747326 is a master commit associated with this PR. Results will be available as each benchmark for each run completes. |
4 participants
ARROW-14095