Linguist: Dynamically suppress files/folders

[frans-k]

I think files in node_modules/ should be suppressed in diffs, or at least the option for it should exist.

I recently added a module to package.json and ended up with 300 new files and 46k additions in node_modules/, and then made a pull request. That comparison took a while to load.

Related but not exactly the same: https://github.com/github/linguist/issues/137

[wouterds]

That's something you would set in a .gitignore.
Why on earth would you want to commit your node_modules?

Here are 3 reasons why to put it in a .gitignore

1. It makes the repo much bigger
2. It's vendor code that can be downloaded by running a single command
3. It's garbage code that messes with Githubs linguist

Just put _node_modules_ in the main .gitignore

[frans-k]

Yeah, that was my own first thought as well, but I'm not making that choice by myself for the project in question.

Choice based on: http://www.futurealoof.com/posts/nodemodules-in-git.html.

TL;DR: Node doesn't have a Gemfile.lock, so

• dependencies of dependencies and
• dependencies with unspecified versions

could differ between deploys and developer machines.

As for your number 3: if it's vendor code, shouldn't it _not_ mess with linguist?

[wouterds]

Hmm.. I understand.
You have a point there.

And yes it shouldn't mess with it but it always does (for me).
I've been looking for a solution for a while now to exclude specific files or folders from linguist.

[frans-k]

Ah, what a pity. I'm guessing you haven't missed the discussion in #137?

As far as I can tell node_modules/ should be added permanently as a vendor folder though.

[tnm]

Is there any time you wouldn't make node_modules/? ignored. Seems like that could just be set in our in vendored file, and we'll always ignore site-wide.

[tnm]

Closing this issue since we're working on improving detection site-wide instead of just doing per-repo restrictions, but please feel free to open an PR for ignoring this directory.

[frans-k]

node_modules/ is already listed in vendor.yml, so I'm assuming that already works.

[Frogging101]

I do not understand why you (GitHub) are so opposed to per-repo configuration.

[perlun]

@Frogging101, probably because it makes the parsing of the language data a lot more complex if each project can set its own world-view on these things?

[milani]

Also Pods folder is excluded in vendor.yml, it doesn't work. It counts ruby as a language in my repo, and also it counts Pods in calculating line additions, shows Pods folder in diff window.

[pchaigno]

@milani Please open a new issue, it's not the same problem.

[carnesen]

@tnm Re "Is there any time you wouldn't make node_modules ignored?" Here's a use case. When you do require("module-x") in Node.js it searches upwards from the file containing the statement until it finds a subdirectory named node_modules/module-x. So if I have a bunch of interdependent source-code-controlled modules in a single repository I can make them accessible to each other by simply putting them into a directory named node_modules. After git clone and npm install, our repo looks like this:

• monorepo
  
  
  
  • node_modules (Shared 3rd party dependencies. gitignored)
  
  
  • src
  
  
  • node_modules (our JavaScript modules. Committed to git)
    
    
    
    
    
    • module-x
    
    
    
    • node_modules (module_x-specific 3rd party dependencies. gitignored)
    
    
    
    • module-y
    
    
    
    • node_modules (module_y-specific 3rd party dependencies. gitignored)
    
    
    
  
  

[dbkaplun]

+1. I think a good solution would be to add a URL param like ?exclude="node_modules/*". This would match the interface for the existing ?w= param which toggles whitespace diffing.