src-d/vecino

Failed to install dependencies

Closed this issue · 9 comments

bzz commented

pip3 install -r requirements.txt results in error in CLI

Logs

Obtaining bblfsh from git+git://github.com/bblfsh/client-python.git@v0.0.3#egg=bblfsh (from -r requirements.txt (line 1))
  Updating ./.venv/src/bblfsh clone (to v0.0.3)
Obtaining ast2vec from git+git://github.com/src-d/ast2vec.git@develop#egg=ast2vec (from -r requirements.txt (line 2))
  Updating ./.venv/src/ast2vec clone (to develop)
Collecting wmd>=1.2.4 (from -r requirements.txt (line 3))
  Using cached wmd-1.2.5.tar.gz
    Complete output from command python setup.py egg_info:
    Traceback (most recent call last):
      File "<string>", line 1, in <module>
      File "/private/var/folders/rx/z9zyr71d70x92zwbn3rrjx4c0000gn/T/pip-build-8k90vv9s/wmd/setup.py", line 3, in <module>
        import numpy
    ModuleNotFoundError: No module named 'numpy'

    ----------------------------------------
Command "python setup.py egg_info" failed with error code 1 in /private/var/folders/rx/z9zyr71d70x92zwbn3rrjx4c0000gn/T/pip-build-8k90vv9s/wmd/

Steps to reproduce

 git clone https://github.com/src-d/vecino.git; cd vecino; git checkout develop
 virtualenv -p python3 .venv
 source .venv/bin/activate
 pip3 install -r requirements.txt
 echo $?

Expected result: 0

Actual result: 1

Update

Even after manually installing numpy it fails without -e git://github.com/src-d/modelforge.git@develop#egg=modelforge

bzz commented

Tried adding numpy>=1.12,<2.0 to requirements.txt but that did seem to be an issue.

Can't reproduce, I get a different problem:

Collecting modelforge>=1.0.0 (from ast2vec->-r requirements.txt (line 2))
  Could not find a version that satisfies the requirement modelforge>=1.0.0 (from ast2vec->-r requirements.txt (line 2)) (from versions: )
No matching distribution found for modelforge>=1.0.0 (from ast2vec->-r requirements.txt (line 2))

Which is fixed by adding -e git://github.com/src-d/modelforge.git@develop#egg=modelforge at the beginning of requirements.txt

bzz commented

@fineguy thx for double-checking! Just to make sure - did you used virtualenv as well?

@bzz Hmm, interesting...
The output in my first reply is from docker on science-3. But if I try to do it locally on my Mac I get the same output as you.
It seems there's some issues with wmd package, cause I can't install it separately either. The only workaround I know of is to install numpy beforehand.

bzz commented

Thanks for verifying the issue!

Strange enough, manually pip3 install numpy before doing pip3 -r is a workaround, but putting numpy>=1.12,<2.0 to the first line of requirements.txt is not :/

Indeed, it might be an issue in https://github.com/src-d/wmd-relax now, as soon as it's verified, let's wait for @vmarkovtsev to triage.

bzz commented

Indeed, after manually installed numpy it still fails untill

-e git://github.com/src-d/modelforge.git@develop#egg=modelforge is added to requirments.txt

Manually pip3 install numpy before doing pip3 -r is the supposed workaround. Whatever I do, this will still always fail. I will add a message about numpy but I guess few people will read it. The reason behind this is the way pip install works.

Modelforge shall be installed separately for now. When we have GCS buckets, I will be able to release the packages on pypi and this problem will go away.

Anyway, if in trouble, please see .travis.yml as we always pass CI here.

Finally, vecino was not updated for the most recent changes in ast2vec and will fail as it failed on the engineering demo. I am having meetings today, so will not be able to fix it until this evening.

Oh, and yes: since we do not have GCS buckets, no way you can easily run vecino at all. Please come to me and I tell you the name of the secret bucket.

This is completely resolved. numpy is still needed.