| Re: [cdt-dev] Static Include Analysis (ReDHead) |
On Fri, 2010-05-21 at 17:08 +0200, Kos wrote: > > Let me ask you a question: > My primary focus in this idea is to relieve the programmer of the need > of manually adding includes (and forward declarations) while coding. > Is this also in the scope of your project? Nice that I'm not the only person working on those issues. Actually I was amazed to read you wiki-page and see how much of it matches many features I already have or intend to have. So the scope of my project matches pretty well with with your focus. Here some of the matches: * Find indirect includes used ( called "directly include referenced files" in ReDHead) * Find all files that needs to be included ( part of "organize includes" in ReDHead) * Remove all unused include generation: * (1) path to included file: One can retrieve IIncludeReferences from the class ICProject. When generating includes I check on all of those an chose the shortest path (segment count of IPath class) * (2) What to incude ReDHead can decide if a forward declaration is enough and either include a file containing a forward declaration or the one containing the declaration. * Library headers: Right now I am working on a solution to see that <set> gets included instead of <bits/stl_set.h>. I also intend to provide a method to the user so he can define certain files as "public" headers which will cause ReDHead to include these file (set) instead of the one containing the concrete declaration (bits/stl_set.h). * (3) Position to include ReDhead can find proper include positions which consider (1) existing includes (2) conditional macros (containing includes) (3) include guards (4) position of first declaration in file. unused Includes: * Finding unused includes works well and more or less fast (also on bigger projects) with the help of IIndex and some caching. * When a declaration can be reached through several ways form the source file (through includes) ReDHead can chose in a smart way which one is the better to remove. * Marking on type (codan) Kirstin Weber mentioned that marking unused includes while typing would be nice. This is already implemented and works pretty well. * Markus's comments The problems about the #undef and #pragma statements cannot be avoided which means there should probably never be auto-removal in CDT. instead, the user will have to decide it he wants to apply any proposals. Lukas
Attachment:
smime.p7s
Description: S/MIME cryptographic signature