[CEDET-devel] code completion fails

[Krishnan V <v.s...@gm...>]

This might start off a new thread. Sorry about that. I am copying your
reply from the forum
to GMail and replying.
On the topic, sad to know that CEDET cannot do this in its current version.
Regarding the other IDEs; Visual studio does it pretty well(much to my
disappointment!).
Eclipse is not that good for ITK. Using Eclipse, completion happens
only for some members.
Apart from VS/Eclipse, I have tried Codeblocks. I used a recent binary
version from SVN of
codeblocks on Windows, it did the job well.  Unfortunately, I built
the same from source on
GNU/Linux and the completion is not that good; some plugin problems.
If one wants to help out,
what does one need to "learn"?!

Krishnan V. writes:
> I am trying to use CEDET along with the medical image processing library
> called ITK. This is highly templated code.

[...]

>   typedef itk::ImageFileReader<ImageType> ReaderType;
>   ReaderType::Pointer writer = ReaderType::New();

CEDET currently cannot parse this. I have a local CEDET branch with a
greatly expanded template parser which at least understands the above
syntax, but even this will not give the correct completions, and I
currently don't have the time to dig deeper.

>   reader->SetFileName("image-1.png");
>   reader->Update();

I'm assuming you mean writer-> here?

Anyway, I'm afraid you currently won't get completions with CEDET with
this library. Note that there's also a general problem with templates,
especially when you're using stuff like static polymorphism, curiously
recurring template pattern, etc. The problem is simply that a template
is, well, a template, and CEDET sometimes cannot know which types you're
actually going use and which methods they are going to provide. I guess
that the 'concepts' which were originally planned for C++0x would have
helped a great deal with that problem, but you probably know how that
ended up... I'm currently working with the Eigen library, and it exactly
has that problem due to excessive usage of static polymorphism and
expression templates. I'd actually be interested if VC++ or Eclipse can
deal with that (and how they do it).

I'm not saying that ITK cannot be parsed. You could try to use the clang
back end (if you have a pretty recent clang available); if that one
doesn't work, then Semantic surely never will.

-David


Hi David,
thanks for your detailed reply.