Feature Request: Report reason for match failure

[startupandrew]

It would be very useful if rather than a simple true / false, there was some sort of info returned about why the code did not match the structure. This could be a helpful text string, or perhaps something more formal.

[swestwood]

Definitely, this was something I thought about as I wrote the first version of StructuredJS given how obviously useful it would be. The main issue is that the parse structure does not lend itself to digestible reasons why the code did not match, and even on a higher level, there is a lot of ambiguity regarding the reason why they don't match.

For example, when trying to match "var foo = 3; bar += 11;" "var $a = _; $a += 10" an intelligent reason might be, "The code uses bar in place of foo, then assigns it to 11 instead of 10." But also reasonable might be "The code does not increment foo by 11." And also reasonable might be "The code does not declare bar." So you can see why it would be hard for StructuredJS to report a reason that actually makes sense just based on the current work it does to just figure out if it matches. To speculate, it would be interesting (and much more complex than the current system) to try to incorporate some notion of edit distance so that it could detect user code that was "close" in terms of the parse tree similarity, and then attempt to find the smallest number of modifications to the code that it would take to make it match the structure. But, this is far beyond the scope of StructuredJS at the moment. :)

But, I just put in a feature that might help with some use cases -- check out variable callbacks in the README and demo page (http://khan.github.io/structuredjs/index.html). These allow you to specify callbacks that place additional restrictions on variables or sets of variables, and return messages about why the variable failed your restrictions. We're combining the feature with a sequence of calls to Structured.match to give better suggestions of why the code does not match the requirements, as well as to define more versatile requirements (ie, to say that a value must be between 0 and 100).

Anyway, hope that helps!

Sophia

[startupandrew]

Thank you very much for the response!

I understand the issue with ambiguity and I know it's a difficult problem. I suspect some reasonable guesses could be made. Perhaps I should try my hand at implementing this and send a pull request rather than just throwing my wish list up here : )

The callbacks look like a great new feature! I think this will help in many cases. Thank you for pointing this out to me.