Parsing browser.log is too slow. Having streamed collections is faster. As a base how the collections can be made in the compiler the browcol for tcollection objects for the text ide can be used. But i'm not that good in the classes and streaming world yet. In the compiler is at least a TCStream class that is vmt compatible with Classes.TStream

A way to be independent of the compiler is to create a browser database file that lazarus then can load. MSDev does it also that way. The IDE should also keep track of the line number changes made by insertion or deletion.

Peter

Yes, Browser support. This is a difficult issue; in principle you need the browser info of all units that are included in the current unit. The person who implemented browser support is Pierre Muller; you can reach him at

Michael.

Hi

Just my opinion. What about a "ctags" approach? Does anybosy know of similar tool for object pascal?

bye,

Stefan

Not with RTTI.

Should I be using something else to implement code completion?

Yes, the browser info. Maybe have a look at the implementation of browser info in the text mode IDE. The important files for browser implementation are:

ide/text/fpsymbol.pas
compiler/browcol.pas
compiler/browlog.pas

Afaik the disadvantage is that the current implementation needs parts of the compiler (symtable etc.).
The browser.log generated by the switch -bl isn't very helpful.

If you need more info, Pierre or Peter may help.

I think this is actually the only way you can do it, without including the compiler in the IDE, and I don't think that doing that is a good idea.

Michael.

I'm not familiar with this stuff ata all so I'm a bit over my head so bear with me.

You guys are saying that I should create some browser database that I could reference for the classes in the LCL and the current unit that's loaded and then run through that for the code completion?

I was planning on just compiling the unit with browser information then parsing the browser.log file for the information I need. However, that' going to be tough because I can find the ***FORM1***
(or whatever they typed and then pressed the period). Then I have to parse the source to find out what FORM1 is a declaration of!

That isn't terribly easy because there could be comments or whatever between the text FORM1 and the text TFORM1.

Is parsing the browser.log file the way to go? If not, please explain the other options a bit more. I'm not sure what you guys mean when you say I should build a browser database. During runtime? ????

Shane

>> 02/05/01 02:27PM >>>

I think this is actually the only way you can do it, without including the compiler in the IDE, and I don't think that doing that is a good idea.

Michael.

_________________________________________________________________ To unsubscribe: mail with
"unsubscribe" as the Subject
archives at http://www.miraclec.com/list_archives/lazarus

I'm setting up LAzarus for Code Completion using the RTTI. However, when I call

PT := GetTypeData(Form1.ClassInfo)

Pt only reflects the published properties and methods. Is there a way to get the other methods or am I doing this wrong? Should I be using something else to implement code completion?

Shane

Yes. Delphi itself compiles (!!) the current unit you're working on behind the scenes; that is why if your code contains errors, the code completion doesn't work in Delphi; it also explains why at certain times it seems to be working very hard when you're not doing those times it recompiles the current unit (and other units it needs), to be able to extract browser info...

The only time when they actually parse the current unit in the IDE is when they do the class completion (CTRL-SHIFT-C) (obviously, since then they cannot compile it, as it's not complete)

The best way would be to keep browser log files of all units somewhere, and use those for the code completion. From time to time these could be updated.

Michael.