Unify dpnp public API exports by refactoring __init__.py
#2666
Conversation
vlad-perevezentsev
requested review from
antonwolfy and
ndgrigorian
as code owners
|
View rendered docs @ https://intelpython.github.io/dpnp/pull/2666/index.html |
|
Array API standard conformance tests for dpnp=0.20.0dev0=py313h509198e_69 ran successfully. |
| * Redesigned `dpnp.modf` function to be a part of `ufunc` and `vm` pybind11 extensions [#2654](https://github.com/IntelPython/dpnp/pull/2654) | ||
| * Refactored `dpnp.fft` and `dpnp.random` submodules by removing wildcard imports and defining explicit public exports [#2649](https://github.com/IntelPython/dpnp/pull/2649) | ||
| * Added support for the `out` keyword to accept a tuple, bringing ufunc signatures into alignment with those in NumPy [#2664](https://github.com/IntelPython/dpnp/pull/2664) | ||
| * Unified `dpnp` public API exports by consolidating function exports in `__init__.py` and removing wildcard imports [#2665](https://github.com/IntelPython/dpnp/pull/2665) [#2666](https://github.com/IntelPython/dpnp/pull/2666) |
There was a problem hiding this comment.
Can we remove "--disable=c-extension-no-member" from .pre-commit-config.yaml now?
There was a problem hiding this comment.
And E501 skipping in .flake8
| from .dpnp_iface_types import ( | ||
| e, | ||
| euler_gamma, | ||
| inf, | ||
| nan, | ||
| newaxis, | ||
| pi, | ||
| ) |
There was a problem hiding this comment.
It has to be a part of the below Routines
| ) | ||
|
|
||
| # ----------------------------------------------------------------------------- | ||
| # Constants (borrowed from NumPy) |
There was a problem hiding this comment.
| # Constants (borrowed from NumPy) | |
| # Constants |
| # The order of these declarations are borrowed from the NumPy document: | ||
| # https://numpy.org/doc/stable/reference/arrays.scalars.html |
There was a problem hiding this comment.
It does not really make any sense, because there is a single from .dpnp_iface_types import below which is connected with that.
There was a problem hiding this comment.
The order is defined by the python sort linter and not by the pointing numpy's page.
| # ----------------------------------------------------------------------------- | ||
| # Type-related helper functions | ||
| # ----------------------------------------------------------------------------- | ||
| from .dpnp_iface_types import ( |
There was a problem hiding this comment.
That is a part of Data type routines from https://numpy.org/doc/stable/reference/routines.dtype.html.
Why isn't it moved below to the routines?
| iinfo, | ||
| isdtype, | ||
| issubdtype, | ||
| is_type_supported, |
There was a problem hiding this comment.
That needs to be moved away. It's a kind of deprecated API which we should not expose.
Probably it would be better to move that to dpnp/random/dpnp_iface_random.py (where it is only used) and to make it an internal function.
| # ----------------------------------------------------------------------------- | ||
| from .dpnp_iface_types import ( | ||
| common_type, | ||
| finfo, |
There was a problem hiding this comment.
Do we need to move dtype here instead of the above line#97 ?
| isdtype, | ||
| issubdtype, | ||
| is_type_supported, | ||
| ) |
There was a problem hiding this comment.
It seems result_type has be a part of that block, if we are following the routines page.
2 participants
This PR removes the remaining wildcard imports and
__all__declarations from internal modules and consolidates the corresponding exports indpnp/__init__.pyas part of the ongoing public API unification.This work follows #2665.