kaushal-py / graceengine Goto Github PK

The external libraries that the project uses -
spacy
nltk
anytree
flask
flask-RESTful
flask-CORS

This is to the best of my knowledge. Though this should be tested on a fresh conda environment to see if it works before sending a PR.

@kaushal-py Just a few intuitions before getting into the topic. pip freeze is a command used to list out the particular version of a library which you have installed in your local machine. So that's why it shows the == sign as it shows the version which has been installed in your particular local machine.

Now when you write the requirements.txt file only write the files which are key to your particular project. For example, if I am working on a project where I did everything using TensorFlow and not used PyTorch but PyTorch is installed on my local machine and I include that into the requirements.txt it's unnecessary to download for a particular user.

So coming back to your project, always try to use >= sign when writing the requirements.txt as it will always install the latest version of a particular library. But do keep in mind that certain libraries after getting update may have changed earlier methods, generally doesn't happen like this as decrypted methods are still usable. I also see that you have included all the Flask dependencies which by default get installed when you install flask so this is why also we don't copy the entire pip freeze

I hope this helps. If you want just tell me the list of libraries which are absolutely necessary for your project I will make the changes in the requirements.txt accordingly.

Originally posted by @TanveshT in #25 (comment)

Description
First of all, great project and initiative.
While setting up the project, I had to update two of the package dependencies in requirements.txt in order to run the WSGI server successfully. I would have raised a PR for this, but I don't have the repo forked.

Platform
Ubuntu 18.04 LTS Bionic (WSL2)
Python3 3.6.9 + Pip3 9.0.1

Errors

# Werkzeug 0.14.1 causes this
from ..http import dump_csp_header
ImportError: cannot import name 'dump_csp_header' from 'werkzeug.http' (\fakepath\werkzeug\http.py)

Suggested fixes
Change the following in requirements.txt (these are minimum possible to make it work, I would recommend finding latest possible versions that keep backward compatibility with your code)
For MarkupSafe, its a bump to next stable version, it also fixes install on WSL2 (as 1.0 .whl is not available on PyPI)

Werkzeug==0.16.1
MarkupSafe==1.1.1