Create /tests folder, remove canvas polyfill

[lindapaiste]

Fixes #97
Implements @gohai's recommendations in #82 (comment)

Changes:

• Move all test-related code to a new folder /tests.
  • Move the /coverage folder.
  • Move the jest configuration file (jest.config.js).
    • Specify the location in package.json since it is no longer at the root.
    • Adjust the paths in the config using <rootDir> syntax.
  • Use jest only on the /unit folder.
  • Add an /integration folder for tests requiring a browser. These are not run at the moment and will need to be set up in a future PR.
• Uninstall canvas package - fixes Error installing dependencies #97
  • Remove the polyfill from the setupTests.js file.
  • Move files which involve reading an image into the /integration folder.
• Delete Babel config files and put the config in package.json.

[shiffman]

This looks great, excited to merge! @gohai would have time to review as a second pair of eyes?

[shiffman]

@ziyuan-linn and I are reviewing this together and the tests run on my machine but not his... I'm going to merge this but we will continue to investigate.

I'll note that I had to run npm rebuild @tensorflow/tfjs-node --build-addon-from-source for the tests to work.

[gohai]

I am getting an error doing yarn install (not sure if this the same issue @ziyuan-linn had?):

error .../ml5-ng/node_modules/@tensorflow/tfjs-node: Command failed.

From the detailed error message, I am getting the sense that it might be related to me not having Xcode installed? (see also) (I do have the command-line build tools, though)

  File "/Users/gohai/.nvm/versions/node/v18.19.0/lib/node_modules/npm/node_modules/node-gyp/gyp/pylib/gyp/xcode_emulation.py", line 1737, in _GetXcodeEnv
    "XCODE_VERSION_ACTUAL": XcodeVersion()[0],
                            ^^^^^^^^^^^^^^
  File "/Users/gohai/.nvm/versions/node/v18.19.0/lib/node_modules/npm/node_modules/node-gyp/gyp/pylib/gyp/xcode_emulation.py", line 1507, in XcodeVersion
    version = CLTVersion()  # macOS Catalina returns 11.0.0.0.1.1567737322
              ^^^^^^^^^^^^
  File "/Users/gohai/.nvm/versions/node/v18.19.0/lib/node_modules/npm/node_modules/node-gyp/gyp/pylib/gyp/xcode_emulation.py", line 1545, in CLTVersion
    return re.search(regex, output).groupdict()["version"]
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'groupdict'

@lindapaiste Since probably a lot of people will run into this issue: what's your thought of moving @tensorflow/tfjs-node into optionalDependencies? When I do this, I am still seeing that installing tfjs-node failed (the first time running yarn install):

warning Error running install script for optional dependency: ".../ml5-ng/node_modules/@tensorflow/tfjs-node: Command failed.

.. but the installation process continues with:

$ patch-package
patch-package 8.0.0
Applying patches...
@tensorflow-models/face-landmarks-detection@1.0.5 ✔
✨  Done in 46.85s.

[shiffman]

@gohai just curious does running this command fix the error?

npm rebuild @tensorflow/tfjs-node --build-addon-from-source

[ziyuan-linn]

@gohai I'm not sure if this is the same issue, but removing any spaces in the directory names containing the project fixed the issue for me.

[gohai]

I filed an issue with tfjs here. I got the same error when I ran npm rebuild @tensorflow/tfjs-node --build-addon-from-source, and the path also doesn't have any spaces. So temperamental!

In the Readme there is a mention that Python 2.7 is needed - hoping it's not that, which would make it very difficult for people to get going..

[lindapaiste]

We could consider removing @tensorflow/tfjs-node from the dev dependencies and running the tests using the main/browser version. We will get warnings in the console about how we should be using the Node version, but I get a lot of those still even when I try to use the Node version.

@gohai I definitely appreciate you calling out the difficult dependencies. This is another situation where I've previously installed that package on my computer so it installs easily for me and I forget that there was some complicated setup required at some point in the past.

[gohai]

@lindapaiste If using the main/browser version is an option, then I think we should go with that. (If there are some additional, more intricate setup steps needed, perhaps we could put it into a tests/Readme.md?)