-
Notifications
You must be signed in to change notification settings - Fork 908
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
fixed_point
+ cudf::binary_operation
API Changes
#7435
Conversation
Thanks for the quick turn around |
Well this PR is going to be modified based on the proposal laid out here: #7442. Post any comments if you have them. |
fixed_point
cudf::binary_operation
specified output_type
parameterfixed_point
+ cudf::binary_operation
API Changes
Codecov Report
@@ Coverage Diff @@
## branch-0.19 #7435 +/- ##
===============================================
+ Coverage 81.88% 82.30% +0.42%
===============================================
Files 101 101
Lines 16900 17273 +373
===============================================
+ Hits 13838 14216 +378
+ Misses 3062 3057 -5
Continue to review full report at Codecov.
|
Fixed! This was an awesome catch! Thank you! |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks good overall here. I think this cleans things up.
rerun tests |
@shwina @trxcllnt @brandon-b-miller @jrhemstad Can we get this going as my other PRs are blocked by this. |
@gpucibot merge |
@codereport is making changes to the way `DIV` will behave for fixed-point types #7435. This PR contains Java changes to support those changes. Note: This is a draft until #7435 is merged Authors: - Raza Jafri (@razajafri) Approvers: - MithunR (@mythrocks) - Jason Lowe (@jlowe) - Gera Shegalov (@gerashegalov) URL: #7527
This resolves rapidsai#7442 Recently while working with @razajafri on `fixed_point` binary ops, it became clear that the `cudf::binary_operation` is breaking the "easy to use, **hard to misuse**" # 1 design guideline. I knew about this but I slotted it as technical debt to be cleaned up later. Long story short, after discussions with both @razajafri, @jrhemstad and comments on the rapidsai#7442, we will implement the following: * [x] For `fixed_point` + `cudf::binary_operation` + `DIV` always **use** the `cudf::data_type output_type` parameter * [x] ~~For `fixed_point` + `cudf::binary_operation` + `TRUE_DIV`, require that the columns/scalars provided as arguments (`lhs` and `rhs`) will result in the specified `data_type`/`scale`~~ * [x] Provide a convenience function (something like `binary_operation_fixed_point_scale()`) that will compute the "expected" scale given two input columns/scalars and a `binary_operator` * [x] Remove `TRUE_DIV` * [x] Add unit tests for different output data_types * [x] Update Python/Cython **This will be a breaking change for all `fixed_point` + `cudf::binary_operation`.** Authors: - Conor Hoekstra (@codereport) Approvers: - Keith Kraus (@kkraus14) - Mike Wilson (@hyperbolic2346) URL: rapidsai#7435
This resolves #7442
Recently while working with @razajafri on
fixed_point
binary ops, it became clear that thecudf::binary_operation
is breaking the "easy to use, hard to misuse" # 1 design guideline. I knew about this but I slotted it as technical debt to be cleaned up later. Long story short, after discussions with both @razajafri, @jrhemstad and comments on the #7442, we will implement the following:fixed_point
+cudf::binary_operation
+DIV
always use thecudf::data_type output_type
parameterForfixed_point
+cudf::binary_operation
+TRUE_DIV
, require that the columns/scalars provided as arguments (lhs
andrhs
) will result in the specifieddata_type
/scale
binary_operation_fixed_point_scale()
) that will compute the "expected" scale given two input columns/scalars and abinary_operator
TRUE_DIV
This will be a breaking change for all
fixed_point
+cudf::binary_operation
.