New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Failed to build 1.4.0 from tarball #247
Comments
I am not sure if the logs generated while logging bug is same. Hence attaching the verbose log. |
I've have the same problem trying to build pmdarima on CentOS from the version v1.4.0 sdist on Pypi. The issue is that the sdist on Pypi doesn't include pmdarima/model_selection/tests - I verified this by simply downloading the sdist via The problem seems related to #201 - it added model_selection/tests to pmdarima/setup.py. When I download the v1.3.0 sdist, that code isn't there and I can build just fine. When I clone the repo and run I was able to get the required files included in the sdist by adding |
Thanks for sleuthing that out @televi we will get this patched and released with 1.5.0 at the end of this week |
Thanks guys, I'll look forward. |
Just merged a fix for this. If you want to test it immediately, you can try cloning the repo and running: # if you don't already have Cython and Numpy, you'll need them:
pip install "cython>=0.29" "numpy>=1.16"
# make the sdist
python setup.py sdist
# install from sdist (version is managed by tag in CI/CD)
pip install dist/pmdarima-0.0.0.tar.gz This has been tested and shown to work on Mac and Linux, but we have an open task to get a CI/CD check for sdist installation into our Windows CI pipeline. The 1.5.0 release is going to happen tomorrow, so hopefully everything is resolved by then. |
Hello @tgsmith61591 thanks for pre release help. Build and install was succes but got error while importing module this time. Error screenshot. I will wait till release. |
Cool, thanks for the follow-up @RSwarnkar. The reason for that is you probably tried the import inside of the cloned
But at least we're past the breaking install :) Looks like 1.5.0 should clear up your issues. |
I pulled the 1.5.1 sdist and built a wheel from it so this problem is fixed for me as well. One thing that I do see (should probably be another issue and an enhancement): there isn't a pyproject.toml file included, so if I try to build pmdarima in a fresh venv that has just pip, setuptools and wheel, it will fail hard. I think this can be trivially fixed by addding a pyproject.toml file with requires = [
"setuptools",
"wheel",
"Cpython",
"numpy",
"scipy",
] I'll give that a try but would you be willing to implement something like that? |
@televi what environment are you in? Our
You are correct that if you build from |
@tgsmith61591 We're not allowed to use vendored wheels so we have to build everything from source (when it comes to FOSS at least). FYI, the environment I use to build is just a venv, e.g.
If a package (like pmdarima but there are some others too) doesn't include in its sdist what is required to build (via pyproject.toml, for example), then the above simply fails. What we're doing is relatively complex so we have hundreds of FOSS packages we build and then keep a copy of the resulting build artifact (for things like routine security scanning or proving to a customer that we can reproduce our software even if the FOSS we use were to be removed or somehow compromised). Having things not build using the above means we have to do a bunch of exception processing to get it built. It's not earth ending - just kind of annoying when I trip over it 😃 I know scipy (for example) has converted to using pyproject.toml so that may be a reasonable reference for pmdarima? |
I gotcha @televi I think this is something we could accommodate. Bummer that you can't use binary dists but I understand how that goes :) We'll try to get this into the next release |
Hello, This time installation was successful but while import I am getting I guess this is a bug with Android platform. Android does not supports multiprocessing APIs. Thank you for fixing the build anyways. |
In this thread it looks like there could be a way around it: I'll read a bit more. I get the impression rewriting a parallel task with the threading library would be highly non-trivial |
What we may be able to do however is lazily load joblib, and if you are on a platform that doesn't support it, we could just enforce that you must use |
Describe the bug
Getting
ValueError
while installing thepmdarima
using pip3.Failed building wheel for pmdarima
To Reproduce
Steps to reproduce the behavior:
Tried installing pmdarima in termux app running on Android 9, Redmi Note 7 pro, MIUI 10.2.
Versions
Linux-4.14.81-perf+-aarch64-with-libc Python 3.7.5 (default, Oct 23 2019, 08:30:10) [Clang 8.0.7 (https://android.googlesource.com/toolchain/clang b55f2d4ebfd35bf6
NumPy 1.17.3
SciPy 1.3.1
Expected behavior
Actual behavior
Additional context
The text was updated successfully, but these errors were encountered: