BUG: pandas doesn't install on python 3.10

[eyalk11]

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the master branch of pandas.

Reproducible Example

pip install pandas (installs 1.3.3)

Issue Description

I tried to install on 3.10.0rc2

ultiarray\scalartypes.obj
    scalartypes.c
    numpy\core\include\numpy/npy_3kcompat.h(227): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\common.h(239): warning C4244: 'return': conversion from 'npy_uintp' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(425): warning C4267: '=': conversion from 'size_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(426): warning C4267: '=': conversion from 'size_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(1899): warning C4244: 'function': conversion from 'npy_double' to 'npy_float', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(1902): warning C4244: 'function': conversion from 'npy_double' to 'npy_float', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2577): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2561): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2577): warning C4244: '=': conversion from 'Py_ssize_t' to 'int', possible loss of data
    numpy\core\src\multiarray\scalartypes.c.src(2967): error C2440: 'function': cannot convert from 'double' to 'PyObject *'
    numpy\core\src\multiarray\scalartypes.c.src(2967): warning C4024: '_Py_HashDouble': different types for formal and actual parameter 1
    numpy\core\src\multiarray\scalartypes.c.src(2967): error C2198: '_Py_HashDouble': too few arguments for call
    numpy\core\src\multiarray\scalartypes.c.src(2976): error C2440: 'function': cannot convert from 'double' to 'PyObject *'
    numpy\core\src\multiarray\scalartypes.c.src(2976): warning C4024: '_Py_HashDouble': different types for formal and actual parameter 1
    numpy\core\src\multiarray\scalartypes.c.src(2975): error C2198: '_Py_HashDouble': too few arguments for call
    numpy\core\src\multiarray\scalartypes.c.src(2982): error C2440: 'function': cannot convert from 'double' to 'PyObject *'
    numpy\core\src\multiarray\scalartypes.c.src(2982): warning C4024: '_Py_HashDouble': different types for formal and actual parameter 1
    numpy\core\src\multiarray\scalartypes.c.src(2981): error C2198: '_Py_HashDouble': too few arguments for call
    numpy\core\src\multiarray\scalartypes.c.src(2967): error C2440: 'function': cannot convert from 'npy_longdouble' to 'PyObject *'
    numpy\core\src\multiarray\scalartypes.c.src(2967): warning C4024: '_Py_HashDouble': different types for formal and actual parameter 1
    numpy\core\src\multiarray\scalartypes.c.src(2967): error C2198: '_Py_HashDouble': too few arguments for call
    numpy\core\src\multiarray\scalartypes.c.src(2976): error C2440: 'function': cannot convert from 'npy_longdouble' to 'PyObject *'
    numpy\core\src\multiarray\scalartypes.c.src(2976): warning C4024: '_Py_HashDouble': different types for formal and actual parameter 1
    numpy\core\src\multiarray\scalartypes.c.src(2975): error C2198: '_Py_HashDouble': too few arguments for call
    numpy\core\src\multiarray\scalartypes.c.src(2982): error C2440: 'function': cannot convert from 'npy_longdouble' to 'PyObject *'
    numpy\core\src\multiarray\scalartypes.c.src(2982): warning C4024: '_Py_HashDouble': different types for formal and actual parameter 1
    numpy\core\src\multiarray\scalartypes.c.src(2981): error C2198: '_Py_HashDouble': too few arguments for call
    numpy\core\src\multiarray\scalartypes.c.src(2997): error C2440: 'function': cannot convert from 'double' to 'PyObject *'
    numpy\core\src\multiarray\scalartypes.c.src(2997): warning C4024: '_Py_HashDouble': different types for formal and actual parameter 1
    numpy\core\src\multiarray\scalartypes.c.src(2997): error C2198: '_Py_HashDouble': too few arguments for call
    numpy\core\src\multiarray\scalartypes.c.src(3913): warning C4244: 'return': conversion from 'npy_intp' to 'int', possible loss of data

Expected Behavior

Works

Installed Versions

Not relevant

[eyalk11]

Notice that I do have numpy install, but the requirements require numpy that is not compatible.

[eyalk11]

OK, I tested it without the numpy requirement and it worked. You just need to alter it. My numpy is 1.21.2 .

[lithomas1]

Just to confirm, the wheels work correctly and the issue is that our pyproject.toml does not specify a compatible numpy?
Can you post the full logs from the failed install and post what version of numpy you had initially before you upgraded to 1.21.2?

Thank you.

[eyalk11]

Yep, it seems to be the case. I don't know what happend now, I tried to build it(after clean). It seems fine(though, didn't look deeper). Anyway, here is the log. Last time I tried setup.py install. Now setup.py build .

• I think I did setup install without removing deps. Of course I tried pip install first, that didn't work.
  build.log

[jreback]

@fangchenli @lithomas1 I think after we merge #43729 could attempt to backport here (but likely needs to be manual).

[craigboman]

This also fails for me, but I think this is a python 3.10 compatibility bug with the dependency numpy, rather than pandas. Looking at numpy/numpy#20045

[lithomas1]

OK, installing from sdist works now for me so the fix I put up seems to work. @craigboman The issue was that there's no wheels, so pip compiles from source(sdist). And when compiling from source, numpy got pinned to a version not compatible with Python 3.10. This should be fixed in our next release.

[simonjayhawkins]

This should be fixed in our next release.

@lithomas1 Anything more to do here or can we close this?

[simonjayhawkins]

closing. can re-open or open new issue if problems with the latest release.