Stripping whitespace

[fatso83]

I did a test run on an existing codebase and it failed with some cryptic error messages about expecting a string, but getting NoneType. After some debugging I found out that the file names contained a space at the end (as in <!-- @import "sub1/sub2/myfile.kit " -->.

This works in CodeKit, and I also fixed it by simply doing .strip() on the file name strings, but I am unsure if this should be implemented. Is it in the specs, and if not, should we implement the spec or the actual reference implementation (CodeKit)?

Another matter is the case of better error messages. As it is, it is hard for the end user to get to the cause of the error, since there is very little contextual information (such as what filename causes the error).

[gjo]

Does it works in CodeKit?

• <!-- @import " sub1/sub2/myfile.kit" --> (space at start)
• <!-- @import "sub1 /sub2/myfile.kit" --> (space around directory separator)

To tell the truth, I don't have CodeKit :-)

[fatso83]

Actually both works! I would not expect the last one to work, but it did. I am not totally sure whether this is a bug or a feature of CodeKit ... What if you do have a space in the directory name? Simply stripping it away would make it fail, so the code would need to test for both locations (with and without space).

It is not uncommon to have code in something like "My Documents/my-project" and a path like "../module 1/foo.kit" would fail if simply stripping the whitespace.

[gjo]

First, I think it is just a CodeKit's Bug. We don't have to implement this behavior. Who does need this feature? What should be corrected is a "*.kit" file and is not a compiler.

However, for the people that wants to emulate CodeKit completely, I had started "issue2" branch for adding optional behavior (not default).
I need more test cases to emulate CodeKit's behavior.

[fatso83]

OK, I can probably supply you with a couple of test cases from our setup. Have to free up some time first, but will get back to you.

[fatso83]

A optional behaviour branch seems like a good idea. Just found some other quirks that CodeKit seems to be having with newlines, but will need to produce some test cases for that too before it's worth doing anything about. I can just create a pull request with the failing test cases in the issue2 branch, right? Or do you prefer to have them pushed to the development branch?

[gjo]

I think, we may not haste.
To emulate CodeKit's silly behaviors, we need many test cases.
We should give priority to a test case over a code.