[RFC]: move @stdlib/math/base/ops/* packages to @stdlib/number/* and @stdlib/complex/*

[kgryte]

Description

This RFC proposes moving packages found in @stdlib/math/base/ops to respective sub-namespaces in @stdlib/number/* and @stdlib/complex/*. For example,

• @stdlib/math/base/ops/add => @stdlib/number/float64/base/add
• @stdlib/math/base/ops/addf => @stdlib/number/float32/base/add
• ...

Currently, we rely on prefixes and suffixes to distinguish packages in ops; however, this practice breaks down when wanting to add C variants for various integer dtypes. To keep things cleaner, this RFC advocates for migrating to dedicated dtype sub-namespaces for arithmetic, binary, and logical operations, where the latter were already planned for dtype sub-namespaces. Moving their arithmetic counterparts seems better for the sake of consistency than following the math/base/special precedent which stems from C99 and is primarily oriented to floating-point numbers.

Related Issues

• [RFC]: refactor @stdlib/complex into sub-namespaces #2260

Questions

No.

Other

Note that this RFC does not propose changing math/base/special. This applies only to ops which are more universal for each numeric dtype, and where we do actually want dedicated C interfaces for each dtype (e.g., int8_t stdlib_int8_base_add( int8_t x, int8_t y )).

The migration path would be as follows:

• Copy dtype-specific packages to their respective sub-namespaces.
• Update all require paths using the dtype-specific packages.
• Remove all dtype-specific packages from @stdlib/math/base/ops namespace.

cc @Planeshifter

Checklist

• I have read and understood the Code of Conduct.
• Searched for existing issues and pull requests.
• The issue name begins with RFC:.

[yashathwani]

Please assign me this issue!

[gururaj1512]

@stdlib/math/base/ops/add => @stdlib/number/float64/base/add
@stdlib/math/base/ops/add3 => @stdlib/number/float64/base/add3
@stdlib/math/base/ops/add4 => @stdlib/number/float64/base/add4
@stdlib/math/base/ops/add5 => @stdlib/number/float64/base/add5
@stdlib/math/base/ops/addf => @stdlib/number/float32/base/add

@stdlib/math/base/ops/cdiv => @stdlib/complex/float64/base/cdiv
@stdlib/math/base/ops/cneg => @stdlib/complex/float64/base/cneg
@stdlib/math/base/ops/cnegf => @stdlib/complex/float32/base/cneg
@stdlib/math/base/ops/csub => @stdlib/complex/float64/base/csub
@stdlib/math/base/ops/csubf => @stdlib/complex/float32/base/csub

@stdlib/math/base/ops/div => @stdlib/number/float64/base/div
@stdlib/math/base/ops/divf => @stdlib/number/float32/base/div
@stdlib/math/base/ops/mul => @stdlib/number/float64/base/mul
@stdlib/math/base/ops/mulf => @stdlib/number/float32/base/mul
@stdlib/math/base/ops/sub => @stdlib/number/float64/base/sub
@stdlib/math/base/ops/subf => @stdlib/number/float32/base/sub

@stdlib/math/base/ops/imul => @stdlib/number/int32/base/imul
@stdlib/math/base/ops/imuldw => @stdlib/number/int32/base/imuldw
@stdlib/math/base/ops/umul => @stdlib/number/uint32/base/umul
@stdlib/math/base/ops/umuldw => @stdlib/number/uint32/base/umuldw

@kgryte, can you confirm if these mappings are correct or if any adjustments are needed