Support no-copy conversion of row major ndarray

[songjhaha]

Here is the rule of conversion of ndarray:

• x is a numpy.ndarray

  • if x is col-major(Fortran order in memory), from_ndarray(x) returns an Array in julia with same shape.
  • if x is row-major(C order in memory), from_ndarray(x) returns a Transpose or PermutedDimsArray in julia with same shape.
  • py_coerce(T, x) will first convert x with conver(T, from_ndarray(x)), if T is Array but from_ndarray(x) returns a Transpose or PermutedDimsArray, we still need to copy data into a new Array.

• y is a julia's StridedArray(or Transpose\PermutedDimsArray )

  • if y is a StridedArray, py_cast(Py, y) returns a col-major ndarray.
  • if y is a Transpose, py_cast(Py, y) returns a row-major ndarray.
  • if y is a PermutedDimsArray, py_cast(Py, y) first converts y.parent to ndarray then transposes it with same permutation.

[codecov-commenter]

Codecov Report

Merging #36 (a4df9f3) into main (7e1b5c6) will increase coverage by 0.58%. The diff coverage is 97.05%.

@@            Coverage Diff             @@
##             main      #36      +/-   ##
==========================================
+ Coverage   61.33%   61.92%   +0.58%     
==========================================
  Files          11       11              
  Lines         869      885      +16     
==========================================
+ Hits          533      548      +15     
- Misses        336      337       +1     

Impacted Files	Coverage Δ
TyPython/src/CPython.jl	67.74% <ø> (ø)
TyPython/src/CPython.Julia.jl	95.06% <95.23%> (-0.26%)	:arrow_down:
TyPython/src/CPython.NumPy.jl	92.77% <100.00%> (+0.87%)	:arrow_up:
TyPython/src/CPython.ORM.jl	97.92% <100.00%> (-0.11%)	:arrow_down:

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[songjhaha]

Fix https://github.com/Suzhou-Tongyuan/jnumpy/issues/35 by the way

[thautwarm]

Is it fine to merge now? @songjhaha

[songjhaha]

Is it fine to merge now? @songjhaha

fine