Review Request: Share code between makebuilder and external scripts
Morten Volden
mvolden2 at gmail.com
Mon May 14 18:48:44 UTC 2012
> On May 3, 2012, 10:05 a.m., Milian Wolff wrote:
> > outputview/ifilterstrategy.h, line 45
> > <http://git.reviewboard.kde.org/r/104814/diff/1/?file=61753#file61753line45>
> >
> > I severly dislike this api. please change it to something like
> >
> > virtual FilteredItem errorInLine(const QString& line) const = 0;
> >
> > then, adapt the FilteredItem to have a
> >
> > bool isValid() const;
> >
> > furthermore, make it so that a default-constructed FilteredItem() is not seen as valid.
>
> Morten Volden wrote:
> If I understand you correctly. You would like this (in outputmodel.cpp):
>
> FilteredItem item( line );
>
> bool matched = m_filter->isErrorInLine(line, item);
> if( !matched )
> {
> matched = m_filter->isActionInLine(line, item);
> }
>
>
> to be replaced by this:
>
> FilteredItem item = m_filter->errorInLine(line);
> if( !item.isValid ) {
> item = m_filter->actionInLine(line);
> }
>
>
> But wouldn't returning a Filtered item (for both error and action) force a construction of such an object twice (instead of once) pr. input line?
>
> Milian Wolff wrote:
> yes but this is a cheap struct, don't overoptimize at the cost of readability please
>
> also, it should be .isValid() - i.e. call a function here that should decide whether it's valid or not (this should not require an additional data member imo).
>
> Morten Volden wrote:
> IMHO the first is just as readable as the other. However, I realize that I am entering a discussion about aesthetics - which is probably pointless. How about I make some unittest(s) where we can do some measuring. If the first is not significantly faster than the other I'd be happy to change it to what you suggested.
>
> Milian Wolff wrote:
> if you really want to spent the time on measuring the time difference, sure - do that. be sure to compile it in release mode to get compiler optimizations etc.
Constructed a few tests to benchmark the outputmodel with different filtering strategies. The constructed tests do not show any noticeable difference between the two approaches. Changed the code to address the points mentioned in this item. Same applies to the issue below.
- Morten
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://git.reviewboard.kde.org/r/104814/#review13328
-----------------------------------------------------------
On May 6, 2012, 8:59 p.m., Morten Volden wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> http://git.reviewboard.kde.org/r/104814/
> -----------------------------------------------------------
>
> (Updated May 6, 2012, 8:59 p.m.)
>
>
> Review request for KDevelop.
>
>
> Description
> -------
>
> So far I have:
> * Moved some of the filtering code
> - More specifically some of the code for filtering (Compiler output, script output )
>
> * Extended the outputmodel
> - It now inherits from QAbstractListModel and IOutputview model
> - It is now possible to choose a filtering strategy on the outputview outputmodel
> * Plus a few other things to make things work
>
> * Started doing some Unit test of the different filters
>
> The solution currently has four filtering strategies:
>
> * No filter
> * Compiler filter
> * Script Error flter
> * StaticAnalysis Filter (for cpp check, krazy, etc)
>
> I think I have now come to a point where it would be good to get some feedback on the proposed solution
>
> Work identified that remains to be done (Input is welcome):
>
> * Finish tests for filter strategies
> * Move MakeoutputModel tests to outputview and extend them
> * Make Makebuildder use the code in outputview.
> * Make NofilteringStrategy faster (i.e. do not create a filtered item for each line)
> * Make custom buildsystem use the code in outputview - Done (But Needs to be reviewed)
> * Extend filterstrategies with other filters?? (Input on that one is most welcome)
> * Discuss how the filters should be selected via the GUI (and in which use-cases it makes sense to do so)
>
>
> Diffs
> -----
>
> outputview/CMakeLists.txt 5d9e539
> outputview/delegateholder.h PRE-CREATION
> outputview/delegateholder.cpp PRE-CREATION
> outputview/filtereditem.h PRE-CREATION
> outputview/filtereditem.cpp PRE-CREATION
> outputview/ifilterstrategy.h PRE-CREATION
> outputview/ifilterstrategy.cpp PRE-CREATION
> outputview/outputdelegate.h PRE-CREATION
> outputview/outputdelegate.cpp PRE-CREATION
> outputview/outputfilteringstrategies.h PRE-CREATION
> outputview/outputfilteringstrategies.cpp PRE-CREATION
> outputview/outputformats.h PRE-CREATION
> outputview/outputformats.cpp PRE-CREATION
> outputview/outputjob.h c17b592
> outputview/outputjob.cpp 5b9db06
> outputview/outputmodel.h b4c9631
> outputview/outputmodel.cpp 317e3ee
> outputview/tests/CMakeLists.txt PRE-CREATION
> outputview/tests/filteringstrategytest.h PRE-CREATION
> outputview/tests/filteringstrategytest.cpp PRE-CREATION
> outputview/tests/outputviewtest.h PRE-CREATION
> outputview/tests/outputviewtest.cpp PRE-CREATION
> outputview/tests/projects/onefileproject/main.cpp PRE-CREATION
> plugins/execute/nativeappjob.cpp 3eb654c
> plugins/executescript/executescriptoutputmodel.h 180adbd
> plugins/executescript/executescriptoutputmodel.cpp 1c852e9
> plugins/executescript/executescriptplugin.h 4eea6a4
> plugins/executescript/executescriptplugin.cpp e55b514
> plugins/executescript/scriptappconfig.h 2e401ea
> plugins/executescript/scriptappconfig.cpp d7ff984
> plugins/executescript/scriptappjob.h 48400aa
> plugins/executescript/scriptappjob.cpp 3b68ca5
> plugins/externalscript/externalscriptjob.h 1922469
> plugins/externalscript/externalscriptjob.cpp aeb5a34
> plugins/externalscript/externalscriptoutputmodel.cpp 7a3d5d4
> plugins/externalscript/externalscriptplugin.h c87aaa8
> plugins/externalscript/externalscriptplugin.cpp 6d62b63
> vcs/dvcs/dvcsjob.cpp 1e0f187
>
> Diff: http://git.reviewboard.kde.org/r/104814/diff/
>
>
> Testing
> -------
>
> Started working on Unittest for filtering strategies. Plan to move most tests from Makeoutputmodel at some point. Plus used the patch for a week now and nothing seemingly breaks...
>
>
> Screenshots
> -----------
>
> CPPCheckBefore
> http://git.reviewboard.kde.org/r/104814/s/550/
> CppcheckAfter
> http://git.reviewboard.kde.org/r/104814/s/551/
> QuickCompileBefore
> http://git.reviewboard.kde.org/r/104814/s/552/
> QuickCompileAfter
> http://git.reviewboard.kde.org/r/104814/s/553/
>
>
> Thanks,
>
> Morten Volden
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kdevelop-devel/attachments/20120514/87ac56ed/attachment.html>
More information about the KDevelop-devel
mailing list