Capture unique statement count before doing any transformations

[nwinter]

Proposed in this thread. This would let each parser do a sane job of counting how many real statements are used, instead of getting all crazy with trying to reverse-engineer what was once a meaningful statement from the transpiled output.

[Gazzonyx]

Originally I hadn't considered branches, but I think the most common case is probably an if/else or if/else if/else where the block that contains the most statements is probably the branch doing the work.

Of course, with loops where n is relatively large and you're expecting the common case to be a no-op and the exception being where you do work, that algorithm falls on its head. Also, I have to consider that this is geared a lot towards those who are newer to programming and their code very well might not reflect what my gut says is the common case.

But I think it's still possibly more workable than trying to reverse engineer translated code if you accept a wider margin of error. Anything else that I can think of involves injecting code into the emitted javascript to increment counters.

<thinkingOutLoud>
OTOH, that approach would allow for all languages to be supported if they can be translated to JS, but I suspect there would still be edge cases that are different for each language that would have to be dealt with on a per language basis and then we're back at square 1.
</thinkingOutLoud>

[nwinter]

We can just count all unique statements; it's really not important whether they get run or not. The incentives are still there to minimize the amount of code you write.