Just curious

[josdejong]

Is there some special reason why you use bracket notation this["re"] instead of dot notationthis.re to access properties? I find the latter easier to read.

[infusion]

Yep, I like the latter also better, but I set these on all the attributes I want to keep as they are, when they run through closure compiler in advanced mode. I started to use the advanced mode for all my projects and so I rename the attributes after finishing a project to the bracket version. This is only a little tradeoff compared to the things closure compiler brings in advanced mode, like function inlining, and code rewriting in a more optical way.

[josdejong]

Ah, does closure compiler mangle the property names too?! hm, that's not nice. I didn't know that

[infusion]

Yep, closure compiler tries to minimize them (1-2 chars) if they are not getting exported. By using this explicit syntax, you can mark them as an api. It also tries to remove objects at all. For example the P object, you don't like - if the compiler sees this object has constantly 2 attributes, it will make two variables of it (I'm not sure if in the case of P this really happens, but it tries to flatten internal objects)

[josdejong]

Well, it makes sense of course - mangling internal variable names. But it's not handy debuging in this case, you want to see re and im.

As for P being replaced by two variables - that actually sounds like a good idea - it's just internals and will probably improve performance :D

[infusion]

I think not performance, as the lookup mechanism of v8 is quite advanced with it's hidden classes. Closure compiler tries to optimize for gzip, meaning building a lot of similar chunks of code, which at the end benefits performance as performance in the web means loading time.

And for debugging: You shouldn't debug your minimized code in the best case ;P