On Fri, Aug 6, 2010 at 10:42, Lex Trotman elextr@gmail.com wrote:
On 6 August 2010 17:46, John Yeung gallium.arsenide@gmail.com wrote:
On Thu, Aug 5, 2010 at 6:54 PM, Jiří Techet techet@gmail.com wrote:
By coincidence, I have been playing with TextMate under MacOS and it has an extremely elegant system of indent detection, see:
It does seem elegant. I actually have wondered why there aren't more editors that do this, and I had always thought maybe it was just not as fast to use regexes. But I could be wrong, and even if regexes are indeed slower, they might certainly be Fast Enough (tm) on today's computers.
Glib regex uses PCRE which seems to have reasonable performance (From Boost Regex performance comparisons with other regex engines) where times are sub .001 sec for files about the length of Geany source files. I'd say that was Fast Enough (C).
That sounds really good. Of course the exact numbers depend on for how long the regex is a candidate to match a line - the results will be better if it fails on the first character in every line than on the last character. But since we will be matching against a single line, this should be a non-issue.
Some observations from playing with TextMate:
- increaseNextLinePattern is not normally used, see
http://ticket.macromates.com/show?ticket_id=425D3D1C
so we could drop it.
I disagree. The bracketless if-statement (or other single-line "blocks") is extremely common, and SciTE has relatively ugly code just to handle that case.
- decreaseIndentPattern appears to be matched against the current
line after every single keypress (e.g. once you type ':' in "elsif:" in python, the line containing it unindents). Apparently regex matching is fast enough to do this. The other patterns are matched only when enter is pressed.
Well, Python uses elif, not elsif. But aside from that, Python is a very special case. It would be more natural for Python to not have any automatic decreases, since your elif might apply to not the latest if but rather a previous one (more than one level deep). In Python, it really makes sense for the programmer to explicitly decide the amount of de-indentation,
I agree with that, Geany does my Python fine because it doesn't get in the way, but some may like additional "help". So long as I can turn it off per language too.
and Python programmers are used to making
this decision for every single block. It would be very weird to have elif be an oddball that doesn't fall in line with everything else in the language.
For the record, Python should be a very easy language to get right with the autoindentation, and it is kind of surprising how many editors still manage to mess it up. If I recall correctly, last time I used Geany (several versions ago), it handled Python quite well (either perfectly, or only quite unlikely situations could mess it up).
I wonder how much illegal/unethical it is to copy configuration options from a commercial editor...
In my opinion, it is not unethical at all. I would not want to be a part of copying or stealing someone else's not-open-source source code, but I have no problem with the idea of reimplementing a published spec. I don't know if the law agrees with me.
Too complicated since it is likely that it depends on which country you are in as to legality. But I would observe that there is likely to be only one optimal regex for each language for each setting so even from first principles you will likely get the same answer ;-)
Well, after some more exploration this looks quite good. The bundles SVN for TextMate is here:
http://svn.textmate.org/trunk/
and the license says this:
http://svn.textmate.org/trunk/LICENSE
Under tags you can find settings for all the files:
http://svn.textmate.org/tags/release_1.1b5/
including indentation settings. So I think we could use it, at least as a base for further customization if we are not satisfied with the settings.
Cheers,
Jiri
Cheers Lex
John _______________________________________________ Geany mailing list Geany@uvena.de http://lists.uvena.de/cgi-bin/mailman/listinfo/geany
Geany mailing list Geany@uvena.de http://lists.uvena.de/cgi-bin/mailman/listinfo/geany