@jkuczm No, I didn't. I decided to ask a question here and mail support, get a clearer picture of the state of tex stuff before doing annything
@halirutan Yes, that seems tempting
I'd rather have something that by default does nothing and asks for manual conversion, that can grow in time, very easily customizeable, that the one-fits-all that doesn't seem to work now
Because I don't actually need/want much now from the conversion
I don't mind the content looking different in MMA, I don't mind coding latex. I just want to avoid duplicating content, and I want to work dynamically in mma
@jkuczm If it is too hackly, I'd rather use an alternative approach.
I just hoped one could also apply a function to the full cell expression and not just the boxes
@halirutan Graphics that are inline or somewhere weird, it should fail or ask, unless the user has decided it should do something differently. I don't mind always putting graphics in its special style cell
in which case, perhaps create a figure environment, allow for captions and ref somehow, convert it to a file with some criteria or manually specified in the cell, not sure
but, given the notebook, it should be transparent what the latex code will be, or at least easy to tweak
Transparent, custimizeable, not-smart, learns with you
@Rojo I mean, basically you only need a translator for boxed data. I kind of did something similar when I was extracting the usage messages of all built-in functions and converted them to MathML
I think I found an easy solution. Although my question was how to extract a simple 1d string, I show how to transform usages into nice and simple html. The rules for this can be adapted so that each box-structure is converted into whatever representation is wanted.
The basic trick is the followi...
@Rojo, @halirutan Guys, just a quick note: when I was working on the code formatter, I initially went down the same road (just translator from boxes), but then it turned out that boxes are a bit too low-level. It may or may not be relevant in your context, because the problem is somewhat different. But you might also benefit from introducing some intermediate inert symbolic representation, like I did for formatter.
@LeonidShifrin About the box structures: It really hard to circumvent mistakes you made without your experience. I wouldn't know a better way than just trying for a night and see how it goes.
@LeonidShifrin I'm currently working on a small side project called the new linksnooper
@Rojo Yep. This was the end of 2012, M9 with my first production installment (RLink) was just out, and I had to "come out" as well, could not hide my affiliation any longer :)
@halirutan Well, yes. But I didn't actually read anything (except WReach's great answer on lazy lists - but I'm quite far from there by now), was just designing from scratch. It turned out though, that my design was amazingly similar to the design of similar structures in some other languages.
@halirutan Actually, the hardest part here is very Mathematica-specific, and has to do with speed, of course.
I was reading some answer of, I guess, Daniel Lichtblau about compiling quicksort and come near the internal implementation. I tried myself and couldn't find a place for optimization but the speed was not even close..
@halirutan I can't really beat Sort with no args. But as soon as you provide a testing function, I can get more or less even. Besides, my implementation has other virtues.
@halirutan @SimonWoods I sent my RAM stick to the seller and they sent back a new one which is now working perfectly :) I'm assuming that I was unlucky enough to get a faulty stick :)
