-
Notifications
You must be signed in to change notification settings - Fork 27
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Look for node_modules in any recursive parent directory #44
Conversation
It turns out that when using Do you think you could write a few tests for this, @ariofrio? That would be awesome! I think a few unit tests for |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Needs test as described in my previous comment.
Thanks for working on this, @raphinesse! I'm pretty swamped right now; hopefully I can find some time for in a week or so to write a few tests for this. |
@ariofrio FYI we're in the process of preparing the next major release. If we want this change in it, we should start work on it rather sooner than later. I'm sorry to say I can give no specific time frame though. |
Working on this now. Not sure what a unit test for I have in mind an "integration" test that creates this directory structure:
And then tries to run pathToInstalledPackage to find What do you think? Update: Committed a version of this. Verified locally that the new tests fail on master. The first test succeeds because it tests finding the package in the current directory, which already worked. |
The tests look good to me on a first glance 👍 I never expected One last thing: it would be nice to add a test that shows that globally installed modules will be taken into account too. Not sure how to best test this, but maybe you have an idea? |
Or it might have been the linting errors that show up in the CI logs 😅 |
Huh, is taking into account globally installed modules expected behavior? Seems like it would be error-prone, since running the same command on a different environment without that globally installed module would fail. Not sure if resolve provides that functionality, anyway, reading the docs. |
While I agree that taking global modules into account might be problematic, I would have thought Any way: all the more reason to assert the presence/absence of this behavior in the current implementation. |
Good point about needing to test it either way. Research and links about globally-installed modulesAt least on my environment, requiring a module that's installed globally doesn't work. It seems that it's not supposed to work, unless NODE_PATH is set or the package is installed to a fixed set of three GLOBAL_FOLDERS (on Ubuntu using the nodesource distribution, globally installed modules don't go into any of them, but into /usr/lib/node_modules). A grep through the source code seems to indicate that So to achieve full fidelity, we'd probably need to pass NODE_PATH and those three directories to resolve. Even then, in common cases globally installed modules won't be accessible. Certainly doable. Looking for use-cases to work off of, it seems RisingStack and other use NODE_PATH to avoid relative paths in require(), but they don't use it with whole packages. Bottom line, globally installed modules probably shouldn't work on node, but it depends on whether the node environment is setup to install global packages to one of GLOBAL_FOLDERS. I don't know if we can assume that modern distributions avoid that, though this thread suggests we can. Supporting NODE_PATH would make cordova-fetch fully compatible with npm resolution algorithm AFAIK, and the test for it seems straightforward. |
Added NODE_PATH support and tests. This is the code that generates those GLOBAL_FOLDERS. It's straightforward to replicate this functionality in cordova-fetch, but the way it finds the node_prefix has changed because the node binary moved, and I don't know if it could in the future. Unfortunately, All of these options have drawbacks, and supporting GLOBAL_FOLDERS would be "mostly for historic reasons". @raphinesse, up to you whether to support them and which option has the best trade-offs for the cordova project. As far as testing global modules, I only see two ways of testing it:
|
Great research @ariofrio! 🏅 I think supporting local lookup plus |
Alright! I just ran the automated tests of |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
All good IMHO. Commits should be squashed on merge.
So after really putting this to work to do dependency prefetching for the |
@@ -22,7 +22,7 @@ var events = require('cordova-common').events; | |||
var path = require('path'); | |||
var fs = require('fs-extra'); | |||
var CordovaError = require('cordova-common').CordovaError; | |||
const { getInstalledPath } = require('get-installed-path'); | |||
const resolve = Q.denodeify(require('resolve')); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not an issue but I know there was an initiative to remove the Q dependency and use Native promises. But this is no problem for now.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Everything looks good IMO.
Great! Thanks @raphinesse and @erisu for taking the time to work with me on this! |
Thanks for your contribution @ariofrio! |
Resolves #43.
This needs tests at the moment.Tests added.