On Sat, 12 Jun 2021 at 10:54, Andrea Aime <andrea.a...@geo-solutions.it> wrote:
> Hi all, > PMD recently released version 6.35, with a new check borrowed from the > Sonar family called > "Cognitive complexity". > > In short, it's an improvement over cyclomatic complexity that accounts for > the difficulty in understanding > code rather than just counting the decision points in a method. > You can read a short paper about it here: > https://www.sonarsource.com/docs/CognitiveComplexity.pdf > (10 pages excluding the appendixes, well worth it IMHO). > > I took this new check for a spin... it measures the cognitive complexity > of methods, with a threshold > of 15. That's a non-starter, we have way too many methods above that > threshold. > I've then tried out 50 and 100, still way too many (oh my gosh!). > > Pumping it up to 200 (!!!) provides a list of candidates that seems a bit > more reasonable to attack. > For example, gt-main only has two methods above that threshold. Here they > are (brace for cognitive impact!) > > - "Conditionals hurricane with love for chained property calls" in > DefaultTemporalPrimitive > > <https://github.com/geotools/geotools/blob/f66689b5305eaf381e21dc2feec8011138d4cb24/modules/library/main/src/main/java/org/geotools/temporal/object/DefaultTemporalPrimitive.java#L79> > - "Gonna fit my entire program in a single method" in FilterDOMParser > > <https://github.com/geotools/geotools/blob/14d27a2db448454c64e8f38e9977565cc400d52b/modules/library/main/src/main/java/org/geotools/filter/FilterDOMParser.java#L113> > > I think I recognise that method :-) but in my defence I've learned a lot in the last 20 years! This is different from considering just method length, for example, some > time ago Gabriel was proposing > to look PMD maximum method length check, which flagged only a few methods, > including "reproject" which > is a 400 lines long method... but only has a cognitive complexity of > little above 100, I believe thanks to > its flatter structure. > > So, what do you think? Do you mind having this kind of check as part of > the ones performed by PMD? > We can start with a score of 200 and go down little by little, until time > and interest are exhausted. > > +1 Ian
_______________________________________________ GeoTools-Devel mailing list GeoTools-Devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/geotools-devel
