require should treat "." and ".." like any other directory names #1178

Activity

1. silverwind

   added

   moduleIssues and PRs related to the module subsystem.Issues and PRs related to the module subsystem.

   discussIssues opened for discussion and feedback.Issues opened for discussion and feedback.

   on Mar 17, 2015
2. silverwind commented on Mar 17, 2015

   

   silverwind

   on Mar 17, 2015

   Contributor

   More actions

   I too once expected require('.') to work, but turns out you need to do require('./'). Is there a specific reason we need that trailing slash inside therequire path?

   Here's the current error:

   > require('.')
   Error: Cannot find module '.'
       at Function.Module._resolveFilename (module.js:320:15)
       at Function.Module._load (module.js:262:25)
       at Module.require (module.js:349:17)
       at require (module.js:368:17)
       at repl:1:1
       at REPLServer.defaultEval (repl.js:116:27)
       at bound (domain.js:254:14)
       at REPLServer.runBound [as eval] (domain.js:267:12)
       at REPLServer.<anonymous> (repl.js:269:12)
       at emitOne (events.js:77:13)
   
3. rvagg commented on Mar 18, 2015

   

   rvagg

   on Mar 18, 2015

   Member

   More actions

   I can't think of a good reason so I'm +1 on this unless someone comes up with something, perhaps @isaacs can chime in?
4. targos commented on Mar 18, 2015

   

   targos

   on Mar 18, 2015

   Member

   More actions

   @smcmurray require('..') does already load the correct file
5. silverwind

   added a commit that references this issue on Mar 20, 2015

   module: allow require('.')

   ...

   6fc5e95
6. silverwind commented on Mar 20, 2015

   

   silverwind

   on Mar 20, 2015

   Contributor

   More actions

   Fixed by 6fc5e95
7. silverwind

   closed this as completedon Mar 20, 2015
8. rlidwka commented on Apr 7, 2015

   

   rlidwka

   on Apr 7, 2015

   Contributor

   More actions

   @silverwind ,

   I know it's nitpicking, but... ..test is still treated incorrectly (as a local file, but it should be a module).

   Since we're bikeshedding the hell out of it anyway, my suggestion for iojs@2.0 is this:

   // old method
   /*function isRelative(p) {
     start = p.substring(0, 2);
     if (start !== '.' && start !== './' && start !== '..') {
       return false;
     } else {
       return true;
     }
   }*/
   
   // proposed method
   function isRelative(p) {
     if (p[0] !== '.') return false;
     if (p.length === 1) return true; // "."
   
     if (p[1] !== '.') return p[1] === '/'; // "./"
     if (p.length === 2) return true; // ".."
   
     return p[2] === '/';
   }
   
   /*
    * Tests
    */
   var assert = require('assert');
   
   assert(isRelative('.'));
   assert(isRelative('./foo'));
   assert(isRelative('..'));
   assert(isRelative('../foo'));
   
   assert(!isRelative('.foo'));
   assert(!isRelative('..foo'));
9. silverwind commented on Apr 7, 2015

   

   silverwind

   on Apr 7, 2015

   Contributor

   More actions

   Good suggestion. I (and whoever wrote those module parts) didn't even think about module names starting with dots. I wonder if these are valid npm module names (e.g. they can be published).
10. rlidwka commented on Apr 7, 2015

    

    rlidwka

    on Apr 7, 2015

    Contributor

    More actions

    They are not valid module names, npm forbids anything starting with a dot, see validate-npm-package.

    So it's purely for consistency reasons. Right now you can create file named "..foo.js" and require('..foo') it. It looks like a module, but it's not. I'd like to discourage that.