TypeError: Cannot call method 'error' of undefined in 1.6.3, works fine in 1.6.2

[eranation]

compiling with 1.6.3 breaks with the error as in the title, but the same code compiles in 1.6.2, I think it has something to do with 'own' in for loops. but not sure yet.

I reproduce it but my code base is too big (and a proprietary project) so I need to work on narrowing it down

But i found someone who managed to reproduce it that also included the code:
https://groups.google.com/forum/#!topic/coffeescript/qiUWkPBJ3gM

I will update this issue soon once I get a small enough piece of code to reproduce as well.

EDIT: of course I didn't even bother to check and see that the sample in the link is a js file (which explains the "And what's your coffee ?" comment by @Nami-Doc ) sorry about that! :)

[vendethiel]

the included sample is .js

[michaelficarra]

The error is being caused on this line: https://github.com/jashkenas/coffee-script/blob/0c9f0fd09975185a2cd14f1a9539d94fc7505613/src/nodes.coffee#L1840

[michaelficarra]

Should lines 1840 and 1841 be @pattern.error and @own.error, respectively?

edit: never mind, that doesn't make any sense.

[vendethiel]

1841 is mine (b3ffd25) and I remember that @own.error obviously did not work (same error as here)

[michaelficarra]

Yeah, because it's a bool and doesn't have an error method. We just need to find out why an instance of Value doesn't have an error method.

[marchaefner]

@michaelficarra: You're off by one line. The stacktrace points at the change by @Nami-Doc.

It's easily reproducible:

$ coffee -e 'x for own x in l'
TypeError: Cannot call method 'error' of undefined
...

And it works as expected if an index is used:

$ coffee -e 'x for own x,i in l'
[stdin]:1:13: error: cannot use own with for-in
x for own x,i in l
            ^

I think it would be best to just use @error.

[vendethiel]

@error didn't work either, didn't looked further why.

[marchaefner]

@error didn't work either, didn't looked further why.

Ah yes, For gets its locationData only after instantiation.

@name.error would then be the next best thing. @name is always defined if own is used (and it works - i tested it this time).

[michaelficarra]

The error is on in, so that keyword should be highlighted.

[sferik]

I'm seeing this regression as well. I can confirm that downgrading to version 1.6.2 fixes the issue.

[gottfrois]

+1 for the error. Had to downgrade to 1.6.2

[vendethiel]

I have a commit fixing that (870a940) but I'd rather not have a solution so hacky.

[nathanpalmer]

+1 for this error. Just downgraded to 1.6.2. We have "for own property in result" in our code that is triggering it.

[marchaefner]

A less hacky (but also less user-friendly) solution would be to just not parse the offending code (see 91b2cf6) which would look like this:

$ bin/coffee -bce 'x for own x in l'
[stdin]:1:13: error: unexpected FORIN
x for own x in l
            ^^

(Not really proposing this, just iterating over options ;)

[vendethiel]

@nathanpalmer The solution is not to downgrade but to fix the code

@marchaefner This seems like a good solution, since it actually shrinks down the code. Something really needs to be done about the grammar, but that's another thing on its own.

[michaelficarra]

@marchaefner @Nami-Doc: Agreed. That's what CoffeeScriptRedux does.

[jashkenas]

@Nami-Doc -- want to send over your fix as a PR?

[vendethiel]

I'd feel ashamed knowing a code that bad is in the code of coffeescript. I started working on this because I thought new Literal('in') would work.
I'd rather ask @marchaefner, if you're fine with not parsing own-in

[vendethiel]

Oh yeah, while I'm meddling with grammar ...

# rule : 
'Assignable = TERMINATOR Expression'
# supposedly allowing that :
a =
5
# but also allows this
a = ; 5

but it doesn't seem to make sense to me

[michaelficarra]

Ha. Nice. Please open a new issue for that.