Add test around define([], function(a, b, c) {})

[thecodedrift]

requirejs/requirejs#669

According to the spec, if dependencies are listed as an empty array, this is to be treated differently than if the dependencies argument is omitted completely.

[jrburke]

I'm open to a test being added for this. I do not have the time at the moment, but feel free to add it yourself. I plan on fixing the issue very soon, for the next requirejs release, and all tests are opt in anyway by implementations.

[thecodedrift]

Quick guidance: Since this was in the spec, but a missed test, do we force-opt people in, or do we give it a new test class?

basic_empty?

[jrburke]

I think any new tests should always be opt-in. The spec is all opt-in. Some specialty loaders do not support the full breadth for implementation size/specific use case goals.

basic_empty_deps?

[thecodedrift]

Sounds good. I'll give it a go since I'm in AMD test writing mode.

[unscriptable]

I disagree. This test case should not be optional. This case and its
behavior has been described in the spec for quite some time. To guarantee
consistency across AMD tools, we need to enforce correct behavior. Don't
we?

On Tue, Mar 26, 2013 at 7:42 PM, Jakob Heuser [email protected]:

Sounds good. I'll give it a go since I'm in AMD test writing mode.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/14#issuecomment-15496057
.

[jrburke]

I have never forced any AMD implementation to pass a particular set of tests, each implementer is responsible for configuring which tests to run for their implementation.

What these tests should do is give someone who wants to decide between an AMD loader, some info on which one they should use. I think it will reflect poorly on an implementation if it cannot pass many tests though. Particularly the "basic" set.

[unscriptable]

On Wed, Mar 27, 2013 at 3:32 PM, James Burke [email protected]:

I have never forced any AMD implementation to pass a particular set of
tests, each implementer is responsible for configuring which tests to run
for their implementation.

If we're not even enforcing a basic set of tests, then there is no
standard, effectively.

What these tests should do is give someone who wants to decide between an
AMD loader, some info on which one they should use. I think it will reflect
poorly on an implementation if it cannot pass many tests though.
Particularly the "basic" set.

These tests are for implementors. End users don't even look at these
tests. Most don't know they exist. If end users try a few AMD tools and
get inconsistent results, they'll look for other solutions.

-- John

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/14#issuecomment-15548001
.

[jrburke]

@unscriptable For me, the main point is that we have tests that can be run. It is not like we go around policing libraries that claim they are AMD, or have any official certification process. I certainly do not have the time for that. I think the market eventually weeds them out.

I also do not want to hold off merging in tests to wait for implementers to make sure they are conformant. It is harder for the implementer to pull a branch of a pull request, test and then merge in. To me, it has always been good to merge the tests to master first, and tell implementers to then add it if they want, and if there is a severe problem with one of the tests, then we fix it and merge back in and repeat. I also do not want good tests that people could use now to wait in some queue while the slowest person in the group finally gets to testing.

I do not think we are disagreeing on core principles, just on the specific ordering of steps and number of steps to get to that end. I think how it has been done to date requires the least amount of work. If you want to suggest a different way and are going to commit the time to ensure the new way is consistently followed going forward, I am fine with that. Best to discuss on the list though.

[thecodedrift]

@jrburke @unscriptable I think the biggest difference here is that these tests address something in the spec we were never testing (versus new functionality).

If we want this to be opt-in, we should reword the spec emphasis mine:

The dependencies argument is optional. If omitted, it should may default to ["require", "exports", "module"]. However, if the factory function's arity (length property) is less than 3, then the loader may choose to only call the factory with the number of arguments corresponding to the function's arity or length.

I'll pick up this thread and put it on the list (if nobody else does) once I'm out of my last meeting for the day and not typing & running.

[jrburke]

OK, if this is just about this particular test, then fine, auto add it to all the implementations. We're spending way too much time on this.