I'm not sure. It's hard to detect which DLL is failing to load causing the import error. It might be possible to write some kind of "diagnostic" function to try to determine whether this particular issue is causing the problem or something else (though to do this may require an external tool like https://github.com/lucasg/Dependencies). See also ARROW-7078
On Tue, Oct 6, 2020 at 10:01 AM John Muehlhausen <[email protected]> wrote: > > Thanks, we will try to push conda. Is there any way to upgrade the import > error to one that suggests installing the redistributable? > > On Tue, Oct 6, 2020 at 9:22 AM Wes McKinney <[email protected]> wrote: > > > Since we use C++11, this requirement would occur with any Python > > package that requires VS >= 2015 to build. > > > > This is one of many reasons we recommend using conda to organizations > > because things like the VS runtime are automatically handled. I'm not > > sure if there's a way to equivalently handle this with pip > > > > On Tue, Oct 6, 2020 at 9:16 AM John Muehlhausen <[email protected]> wrote: > > > > > > "pip install pyarrow > > > If you encounter any importing issues of the pip wheels on Windows, you > > may > > > need to install the Visual C++ Redistributable for Visual Studio 2015." > > > > > > http://arrow.apache.org/docs/python/install.html > > > > > > Just now wading into the use of pyarrow on Windows. Users are confused > > and > > > irritated that numpy and pandas work out of the box (pip, python 3.7.9, > > > Windows 10) but pyarrow requires this additional step. > > > > > > Numpy is backed by C code, correct? Is this pyarrow requirement due to > > the > > > use of C++? Should there be a more descriptive error than "ImportError: > > > DLL load failed: The specified module could not be found." ? > > > > > > Is there a way to install the redistributable with pip, in order to > > > maintain a consistent user experience? > > > > > > Thanks, > > > John > >
