for Scripted about enhanced-resolve

[kdvolder]

It does a bit of caching. It caches every io call for a short amount of time (currently fixed to 4 seconds), ...

Thanks for correcting my misunderstanding. This is good to know. Also a short-lived 4 second cache, I think that's fine. My assumption is really that changes to installed node modules are relatively rare. So this is not likely to cause any issue.

Actually, I'd be inclined to increase the lifetime of the cache... maybe with an explicit 'clear button' so the client has some way of explicitly flushing the cache at specific times / events.

Thanks for your comments. You've convinced me that it will be best for scripted to ditch its 'home baked' node resolution and use enhanced-resolve. It seems we will be able to work with you and contribute patches to get stuff/hooks we need :-)

[sokra]

Here is some documentation and the (wip) next version: https://github.com/webpack/enhanced-resolve/tree/wip-0.4

Checkout this file for your filesystem issue: https://github.com/webpack/enhanced-resolve/blob/wip-0.4/lib/cachedFsResolve.js

It also allows to make own caching logic.

It now supports query strings in requests. You'll need this if you want to support require.js (and I want to add this feature to webpack).

And one bugfix which caused in a rar double error case the callback to be invoked two times.

[kdvolder]

Thanks Tobias, that's great. Especially the documentation is very useful to
help me use your library.

Note that most of the 'concerns' I expressed about the library are not
really urgent. Currently I think the library does pretty much what we want.
I was just looking ahead and anticipating a need in the future to use
different filesystem implementation and/or JSON parsers etc.

None of these are really things we need right this moment.

Kris

On Wed, Oct 31, 2012 at 5:26 AM, Tobias Koppers [email protected]:

Here is some documentation and the (wip) next version:
https://github.com/webpack/enhanced-resolve/tree/wip-0.4

Checkout this file for your filesystem issue:
https://github.com/webpack/enhanced-resolve/blob/wip-0.4/lib/cachedFsResolve.js

It also allows to make own caching logic.

It now supports query strings in requests. You'll need this if you want to
support require.js (and I want to add this feature to webpack).

And one bugfix which caused in a rar double error case the callback to be
invoked two times.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/2#issuecomment-9942731.

[sokra]

So I've added the completion feature: (beta!)

You can use the resolve.complete method and pass a incomplete require string. The string must contain exactly one *, which indicates the insert postion (cursor position). The function returns an array of possible strings the user can input (or choose from).

resolve.complete.sync(/* context = */ __dirname, "enhanced-re*", /* optional options = */ {});
// returns
[
 {
  insert: "solve",
  seqment: "enhanced-resolve",
  part: "enhanced-resolve",
  result: "enhanced-resolve"
 }, {
  insert: "quire",
  seqment: "enhanced-require",
  part: "enhanced-require",
  result: "enhanced-require"
 }
]

"enhanced-resolve/l*/resolve" ->
[
 {
  insert: "ib",
  seqment: "lib",
  part: "enhanced-resolve/lib/resolve",
  result: "enhanced-resolve/lib/resolve"
 }
]

"enhanced-resolve/lib*/resolve" ->
[
 {
  insert: "",
  seqment: "lib",
  part: "enhanced-resolve/lib/resolve",
  result: "enhanced-resolve/lib/resolve"
 }
] // already valid entry returns with insert == ""

"enhanced-resolve/liba*/resolve" ->
// invalid entry in at current position returns empty array
[]

"enhanced-resolve/liba/resolve*" ->
throws error "liba" not found

Here is a simple playground executable to play with the code completion:

https://gist.github.com/4021254

[kdvolder]

Looks very useful! I've added a link to a relevant Jira issue in our bugtracker to point here:
https://issuetracker.springsource.com/browse/SCRIPTED-27

Tobias, currently I have two separate resolver algorithms in scripted. The node/commonjs resolver is just a wrapper around 'enhanced-resolve'.

But the amd resolver is still 'home-baked'. I'm not quite sure if it is possible to configure enhanced-resolve so it can assume the role of an amd resolver. If possible, I think that would be preferable though. Do you have some ideas on that?

The reason why I'm uncertain about this is that it looks to me the amd algos are a bit different than what happens in node. Rather than looking through and trying a number of different places (several dirs, extensions, package.json etc)... it looks like AMD loaders like requirejs and curl tend to just take a module name, some config data and from that derive a single path where the module must be loaded from. So it seems a lot less 'searchy' to me.

I'm really still learning how all this stuff works, so my ideas on this are surely naive and may even be totally incorrect.

So, Im very interested in hearing your thoughts on using enhanced-resolve for amd-resolution.

Kris

[sokra]

https://github.com/webpack/enhanced-resolve/wiki/Common-Options

[sokra]

A cool demo for require completion: http://webpack.github.com/enhanced-resolve-completion-demo/
Type in the input, up and down arrow, enter, click...