[Farrago-developers] findbugs

[John V. S. <js...@lu...>]

[cc'ing farrago-dev]

I logged the first two findings below for me as FRG-270, and the last 
one for Rushan as LDB-162.

People complain about operator overloading in C++, but in the last case 
below, due to JDK 1.5 autounboxing, we have a situation where

(colSel1 < colSel2) means compare by value, whereas

(colSel1 == colSel2) means compare by reference, for the same colSel1 
and colSel2!  Sheesh.

JVS

Boris Chen wrote:
> Findbugs works fairly well, but so far, only as ad hoc. Havent' had the time to work on getting it to be a repeatably used thing with maintained excludes, etc. Removal of warnings in general would be nice
> 
> 
> Hmm. still can't find what zelaine described, but here are 3 other instances of comparisons by reference
> 
> 
> 
>  //open/lu/dev/farrago/src/org/eigenbase/jmi/JmiChangeSet.java
>                 // Avoid unnecessary updates in the case where ordinals
>                 // are already correct.
>                 if (oldOrdinal != newOrdinal) {
>                     target.refSetValue("ordinal", newOrdinal);
>                 }
>             }
>         }
>     }
> 
>  //open/lu/dev/farrago/src/net/sf/farrago/ddl/DdlValidator.java
>     {
>         stmtValidator.setParserPosition(null);
>         if (action == VALIDATE_CREATION) {
>             return invokeHandler(modelElement, "validateDefinition");
>         } else if (action == VALIDATE_MODIFICATION) {
>             return invokeHandler(modelElement, "validateModification");
>         } else if (action == VALIDATE_DELETION) {
>             return invokeHandler(modelElement, "validateDrop");
>         } else if (action == VALIDATE_TRUNCATION) {
>             return invokeHandler(modelElement, "validateTruncation");
>         } else {
>             throw new AssertionError();
>         }
> 
> 
>      
> //open/lu/dev/farrago/src/com/lucidera/lcs/LcsIndexOptimizer.java   
>         public int compare(SargColumnFilter filter1, SargColumnFilter filter2)
>         {
>             // sort based on the selectivity if stats are available; otherwise,
>             // just sort on column number to ensure that results are
>             // deterministic
>             Double colSel1 = filter1.getSelectivity(tabStats);
>             Double colSel2 = filter2.getSelectivity(tabStats);
>             if (colSel1 != null && colSel2 != null) {
>                 return
>                 (colSel1 < colSel2) ? -1 :
>                     ((colSel1 == colSel2) ? 0 : 1); 
>             } else {
>                 return (filter1.columnPos - filter2.columnPos);                  
>             }
>         }
>         
>     }