Re: Hello... Tasklist developement question..

Tor Norbye wrote:

[snip]

> Welcome Back!!!  (I'm probably the wrong person to say welcome since
> I've been too busy working on Project Rave even in my spare time to be
> able to do any work on the tasklist. But one of these days....)

Welcome Trond!

[snip]
 
> But yes, the scanner could definitely use some work, especially the one
> which works only in comments; if I recall correctly, it's splitting
> each and every line up into a separate String. It would be ideal if you
> could for example just scan the source, and identify each comment range
> starting and ending index, and then do a regexp match (with a
> precompiled regular expression) on each range.

Killing unneeded strings will help a little bit. I feed
profiler with Scan Suggestions action and major performance 
bottleneck is document loading. Then throwing it away.

But without document, code must guess encoding and be aware
of guarded sections (these hopefully do not change
line numbers). It's used in search module (so it's faster)
but it sometimes hits in unrecognized binary files etc.

[snip]

>> What do you think (or should I just "go back to sleep"? ;)
> 
> I think it would be really useful if you could do work to get the
> "other" suggestion providers working, since Petr is focusing on the
> docscan module (and copyright? I see that it was split apart from the
> docscan module, but I don't know if that means copyright will no longer
> be supported or if it was for purity reasons only.), and Tim is

Yes, I do not plan to push copyright module to stable distro.
Hence the separation because I need functionality aligned
with modules layout.

[snip]

> In particular, pmd, javaparser, and javadoc. I see that the pmd guys
> are going to provide auto fixers for their rules soon (I already have
> some for pmd in the tasklist/pmd module). It would be cool to make sure
> these are integrated - and of course to include the latest version of
> pmd, since they have some cool new rules.  Also, the javaparser could
> use some new autofixers - and this is pretty fun programming I think:
> identify a compiler error javac produces, and see if you can come up
> with some heuristics for suggesting how the error could be fixed.
> Adding try/catch blocks is hard without an AST for the code, but other
> errors, such as figuring out that the user has copied a class file but
> forgotten to rename the constructor, is pretty straightforward.

[snip]

One advantage of pioneering in pmd, javaparser, and javadoc
modules is that you are not tied by strict quality criteria.
I can easily imagine corner case where text based heuristics
fails (immediately raised as critical dataloss bug).

In spite of that current users like PMD integration a lot.
It's my long term favourite.

To sum up, raw or perceived performance improvements are
welcome as well as pioneering in code analysis area (PMD).

  Cc.