[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
Re: [cdt-dev] Turning off ctags indexer
|
Does it mean that an index built from scratch using the fast indexer is
just as accurate as the one built with the full one? Would it make
sense to give users a command that rebuilds the index from scratch,
something like make clean?
-sergey
On 5/22/06, Doug Schaefer <DSchaefer@xxxxxxx> wrote:
Both indexers create an instance of the
DOM, which includes the AST and Bindings, and saves certain objects from the
DOM to a database, called the PDOM.
The Full indexer creates a full DOM for
each compilation unit (.c or .cpp). As such, it works much like a compiler. And
is pretty accurate, or as accurate as we could make the DOM with the resources
we had.
The Fast indexer cheats and uses the
objects saved in the PDOM from previous indexing so that it can skip over
previously indexed header files when they are #included. This is similar to how
precompiled headers work and we're getting similar performance gains from
it. However, this is less accurate than the full indexer since not all objects
in the DOM are being saved so there may be dangling references or
misinterpretation. Over time, I'll add in more objects as long as
performance doesn't degrade. But, since it is all based on heuristics,
there will always be some chance of getting wrong answers.
To: CDT
General developers list.
Subject: Re: [cdt-dev] Turning off
ctags indexer
What is the accuracy difference between the two
indexers? How does it manifest itself? In what situations?
-sergey
I've changed the little
comment after the indexer names. It's now
-
Full Indexer (slow but accurate)
-
Fast Indexer (faster but less accurate)
I have had requests to
keep the full indexer around due to its better accuracy so am hesitant to call
it "Legacy". But I would like a better name than "Fast" so
if anyone has suggestions…
I'm also wandering what
to call the indexer that runs in ctags mode (i.e ignores all include statements
all together), but that probably won't show up until 4.0 or maybe 3.1.1 and then
only if people are having a hard time with the "Fast" indexer.
BTW. The default indexer
is the Null one, i.e. no indexing at all. I have put a check box in the Indexer
preference page that will allow you to change all the projects in your
workspace to the selected default. That should make it easier to turn things
on.
To: CDT General developers list.
Subject:
Re:
[cdt-dev] Turning off ctags indexer
It also makes sense to rename Fast C/C++ Indexer to just C/C++ Indexer
and Full C/C++ Indexer to Legacy C/C++ Indexer, or some other bad name that
would discourage its use.
-sergey
+1 from me to "turning it
off". I'd rather pour any effort into
improving "our own" technology now that we are sufficiently down
the road to independance.
Thomas
> Sent: May 22, 2006 1:04 PM
> To: CDT General developers list.
> Subject: [cdt-dev] Turning off ctags indexer
>
> Hey gang,
>
>
>
> I am continuing to feel the affects of trying to replace five
> people with myself. To help make sure we reach the dates with
> good quality, I am going to reduce content. Since the PDOM
> Fast indexer is now pretty fast, I am going to turn off the
> ctags indexer. I have not had the time to port it over to the
> PDOM completely. Also looking at the info we get from ctags,
> I'm not sure how the features that used the index worked well at all.
>
>
>
> Also, in the future, I can see now how I can get the DOM
> parser to work like ctags and probably have it produce better
> results. The ctags indexer was really meant as a stop gap to
> make sure we could get some index info for large projects in
> a reasonable time. This may longer be necessary.
>
>
>
> As always, I'm open to feedback. In this case, however, the
> ctags indexer will only be brought back if someone steps up
> to work on it.
>
>
>
> Cheers,
>
_______________________________________________
_______________________________________________
cdt-dev mailing list
cdt-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/cdt-dev