Fw: [cdt-dev] Questions about indexing

["Andrew Niefer" <aniefer@xxxxxxxxxx>]

It seems I forgot to switch the dev
list to my IBM address, so my first message didn't go through (see below).

I also didn't mention below, its not
quite clear yet how the search engine is going to get from AST node to
ICElement (whether it builds its own ICElements or tries to find the correct
one from the C Model).  The AST nodes contain all the information
you want, so a visitor to that tree could be enough for code assist.

----- Forwarded by Andrew
Niefer/Ottawa/IBM on 07/07/2003 10:20 AM -----

Andrew Niefer/Ottawa/IBM

07/07/2003 10:02 AM

To	cdt-dev@xxxxxxxxxxx
cc	cdt-dev@xxxxxxxxxxx, cdt-dev-admin@xxxxxxxxxxx
Subject	Re: [cdt-dev] Questions about indexingLink

Hello, 

The outline of the searching goes like
1: get from the indexer a list of files to look in, 2: use the parser to
search those files, and 3: report any matches.

1) Starting in SearchEngine.search,
 get the index manager call it with a new search job.  This part
is still under development, but the ultimate result will be a collection
of paths in which we will find references to the item in question.  (Bogdan
knows the details here more than I)
2) With this collection of paths,  make
a new MatchLocator and call locateMatches.  MatchLocator implements
the parser interface ISourceElementRequestor.  It starts a parse for
each file returned from the indexer, and it uses the ICSearchPattern as
it receives AST nodes to determine if it finds a match.
3) Matches are reported to ICSearchResultCollector.
 This class is still being worked on, but the idea is to create an
ICElement  using the information from the AST node and then hook that
up to an IMarker and put it in the search pane.

In general, I think if the code assist
wanted to use the search engine, it could implement ICSearchPattern and
ICSearchResultCollector to decide for itself what is a match and then what
to do with that match.

-Andrew

"Thomas Fletcher" <thomasf@xxxxxxx>
Sent by: cdt-dev-admin@xxxxxxxxxxx

07/06/2003 02:30 PM

Please respond to
cdt-dev@xxxxxxxxxxx

To	<cdt-dev@xxxxxxxxxxx>
cc
Subject	[cdt-dev] Questions about indexing

So with the recent flurry of activity on indexing
I thought .. well now
might
be the time to start looking at improving the current content assist.  So
I
started looking into the indexing ... and I got confused.

There is the original IndexModel defined in
index/org.eclipse.cdt.core.index.
Using this very simple indexing model I would do:

    IndexModel model = IndexModel.getDefault();
    ITagEntry [] tags = model.query(project, "*string-reg-exp*",
false,
false);

Which was mega simple, but also very far from being complete (even though
the
tags theoretically included lots of additional information about variables
and types etc).

Looking into the search stuff under development I found the new indexing
stuff
    CModelManager modelManager = CModelManager.getDefault();
    IndexManager indexManager = modelManager.getIndexManager();

    indexManager.performConcurrentJob(new PatternSearchJob(),

ICSearchConstants.WAIT_UNTIL_READY_TO_SEARCH,
                    
                     
                   monitor);

I looked at PatternSearchJob() and it is pretty well an empty shell at
the
moment since
it is under development, but even then I don't see what the interaction
is
between this
class and the actual "searching for something" since there is
also a
mysterious MatchLocator/
ICSearchPattern whose roles have to do with reporting the match back, but
seem to be
distantly located from the the indexManager and more related to the
parser/model.

My question is, can someone outline the basic flow of how I would be able
to
use the
facilities of the new indexer to programmatically generate an array of
entries which might
match a given _expression_.  As a starter I'll provide my current guess
(and
its problems?):

- Create a new search engine using SearchEngine()
- Create a search pattern using SearchEngine.createSearchPattern()
- Create a collector which will grab the results and stuff them somewhere
- Limit the scope appropriately and execute the search engine with the
above
  pieces (incidentally can I limit the scope to variables local to
a
function? Can
I limit the scope to global variables? ... didn't see this in the
definition).

My question on this, is that the results returned back by the collector
are
mostly
file oriented, though there is an ICElement included.  Lets say that
I
submit a
query and get back a function and a variable: foofunc and foovar.
- I can get the element name: Great!
- Can I get the return value/parameters for a function or the type
information
for a variable?  Is is possible to go back to the model easily given
an
ICElement
to extract this information?

I'll wait until I hear back on this until I decided to go further and
replace/enhance the
editor code completion facility.

Thanks,
  Thomas

_______________________________________________
cdt-dev mailing list
cdt-dev@xxxxxxxxxxx
http://dev.eclipse.org/mailman/listinfo/cdt-dev