ENH: Annotate the arithmetic operations of `ndarray` and `generic` (#17273)

* ENH: Added annotations for arithmetic-based magic methods

* TST: Added arithmetic tests

* TST: Moved a number of tests to `arithmetic.py`

* ENH: Ensure that objects annotated as `number` support arithmetic operations

* MAINT: Arithmetic operations on 0d arrays return scalars

* MAINT: Clarify the type of generics returned by `ufunc.__call__`

* TST: Added more arithmetic tests

* MAINT: Use `_CharLike` when both `str` and `bytes` are accepted

* MAINT: Change the `timedelta64` baseclass to `generic`

* MAINT: Add aliases for common scalar unions

* MAINT: Update the defition of `_NumberLike`

* MAINT: Replace `_NumberLike` with `_ComplexLike` in the `complexfloating` annotations

* MAINT: Move the callback protocols to their own module

* MAINT: Make `typing._callback` available at runtime

* DOC: Provide further clarification about callback protocols

* MAINT: Replace `_callback` with `_callable`

Addresses https://github.com/numpy/numpy/pull/17273#discussion_r485821346

The use of `__call__`-defining protocols is not limited to callbacks. The module name name & docstring now reflects this.

* MAINT: Removed `__add__` from `str_` and `bytes_`

Most `np.bytes_` / `np.str_` methods return their builtin `bytes` / `str` counterpart.
This includes addition.

* MAINT: Fix the return type of boolean division

Addresses https://github.com/numpy/numpy/pull/17273#discussion_r486271220

Dividing a `np.bool_` by an integer (or vice versa) always returns `float64`

* MAINT: Renamed all `_<x>Arithmetic` protocols to `_<x>Op

Addresses https://github.com/numpy/numpy/pull/17273#discussion_r486272745

* TST: Add tests for boolean division

* ENH: Make `np.number` generic w.r.t. its precision

* ENH,WIP: Add a mypy plugin for casting `np.number` instances to appropiate subclasses

* Revert "ENH,WIP: Add a mypy plugin for casting `np.number` instances to appropiate subclasses"

This reverts commit c526fb619d20902bfd77709c8983c7a7d5477c95.

* Revert "ENH: Make `np.number` generic w.r.t. its precision"

This reverts commit dbf20183cf7ff71e379cd1a165d07e1a1d643135.

* MAINT: Narow the definition of `_ComplexLike`

Addresses https://github.com/numpy/numpy/pull/17273#discussion_r490440238

* MAINT: Refined the return type of `unint + int` ops

`unsignedinteger + signedinteger` generally returns a `signedinteger` subclass.
The exception to this is `uint64 + signedinteger`, which returns `float64`.

Addresses https://github.com/numpy/numpy/pull/17273#discussion_r490442023

* MAINT: Use `_IntLike` and `_FloatLike` in the definition of `_ComplexLike`

0 files changed, 0 insertions, 0 deletions