MAINT: simplify complex math function handling in npymath #22161
Conversation
|
We have our own function declarations in |
|
this needs a rebase now that gh-22090 is merged |
a569eb7 to
c6366c1
Compare
|
Tried to fix the issue with the mandatory functions by creating the header. That is slightly annoying, since MSVC does not have proper complex support, so the header has to define the complex type. gh-5687 may mean we have to put the |
|
It's probably a pragmatic choice to add all those complex functions to a single file. It does couple them together though. I'll note that I implemented this as separate checks for Meson already: https://github.com/numpy/numpy/compare/main...rgommers:numpy:meson?expand=1#diff-1b8787c0de2167aaf65f45eadb93cec2810f91b92c42e422b1e8d07bc152a92fR65-R79 |
64b4397 to
05e213e
Compare
| @@ -102,12 +102,14 @@ def set_sig(sig): | |||
| args = args.rpartition(")")[0] | |||
| funcname = prefix.rpartition(" ")[-1] | |||
| args = [arg.strip() for arg in args.split(",")] | |||
| FUNC_CALL_ARGS[funcname] = ", ".join("(%s) 0" % arg for arg in args) | |||
| # We use {0} because 0 alone cannot be cast to complex on MSVC in C: | |||
| FUNC_CALL_ARGS[funcname] = ", ".join("(%s){0}" % arg for arg in args) | |||
There was a problem hiding this comment.
Lets see if this works. I had to add the struct here because (_Dcomplex)0 is not a valid cast in MSVC complex.h...
|
@rgommers should we just roll with this? The deletion seems nice (whether we need to roll back or not eventually). |
|
xref #20880 |
Continuation of #22090: now for complex functions