Stephan Hennig wrote:
Am 23.02.2010 17:15, schrieb Taco Hoekwater:
Stephan Hennig wrote:
Are there any plans to replace libhnj by a pure Lua implementation? As
long as hyphenation is hard-coded in C, it remains a black box for the
LuaTeX user.
But it is a reasonably fast black box. :)
Sure. But I wonder what impact hyphenation has on runtime in comparison
to paragraph breaking.
Not a lot, that is true.
I haven't found much about libhnj's finite state hash approach, so I
have to ask here: Is that something else than an automaton that is the
result of pre-calculating all states the /list/ of tries can ever take
on for arbitrary (word) input?
That's pretty much what it does, yes. There is a severe lack of
documentation, though. I recall seeing a set of slides explaining
the algorithm, but I can't find it on the web any more.
To me, it sound like Lua's tables are just made for automata (but who am
I?). I don't want to push you re-writing existing code. Actually, I'm
considering doing that myself (for various reasons)). So, an honest
question: Is there anything that makes implementing the finite state
hash approach hard in Lua (other than performance)?
No, it should be reasonably straightforward. Ok, as I have some time to
spare today, I'll have a go at it (I got stuck in metapost anyway, and
can use a bit of distraction).
Best wishes,
Taco
