Re: [Ctags-devel] ctags library

[Elliott H. <en...@je...>]

On 2006-09-20, at 06:40, Massimo Cora' wrote:
> ... In our editor we wish to have a real-time updating of
> symbols, without the need to flush the buffer to the disk, then  
> parse it
> with ctags, and then use the symbols. This is IMHO is a bad design.

http://software.jessies.org/Evergreen/ does "real time" updating by  
writing to disk. i felt the same as you initially, that it was a bad  
choice, but i'm less sure. we ended up only calling ctags on a  
"significant" change because the more often you call ctags, the more  
often it's given ill-formed source, and the more often your tags tree  
ends up mostly empty because it's had to skip a lot to recover. so,  
anecdotally, that's one reason why you might prefer to keep ctags as  
a separate process.

> I mean, why use harddisk if we can just skip it and use RAM directly?

because you're then immune to crashing bugs, hanging bugs, and  
leaking bugs in ctags?

i understand the way you feel, but i don't think it's as clear-cut a  
decision as you seem to believe.

personally, i'd be more interesting in seeing if there's anything to  
gain by making ctags use mmap(2) where available, like grep(1) does.

  --elliott