I also agree that we have too many different ways of doing things. A set of common rules/ways would definitely be beneficial for the project.
Concerning the command line parsing: I thought that Alexander Alexandrov wanted to unify the command line parsing by replacing both tools with a better one. So this should be fixed sooner or later. The different code styles are just natural if there is no common set of established rules, which are also enforced. I see two solutions: Either enforcing common coding rules or refraining from reformatting the code from other contributors. I don't know whether we can find a common denominator with which everyone can live and which is yet specific enough to make the code base more homogenous. I also agree that the mixed Java/Scala projects make it harder to get started. I've often seen that people confuse the basic types (Scala tuples vs. our own tuples, Java list vs. Scala list, etc.). This is probably something we cannot fix without rewriting parts which are implemented in the "other" language, though. Personally I don't see the different testing styles as critical. Whether one is using JUnit tests, WordSpecLike tests or FlatSpec tests, it should be pretty obvious for everyone where the testing code is written in case that they want to change something. Moreover, most of the time you would not want to change proper defined test cases. In fact, I think that WordSpecLike and FlatSpec let you write better tests, because it encourages people to clearly and more naturally write what they want to test. Compared to some shorter and often not so meaningful junit method names, this is for me a clear advantage. If a test case fails, I want to directly know without having to go through the testing code, what was tested and what went probably wrong. However, this holds only true if we don't use the JUnitRunner to run these tests. Unfortunately, this is currently the case. With JUnitRunner, the actual result output of the tests is often hard to decode. Therefore, we should probably stick to writing testing methods if we use JUnit to run our Scala tests but use the more expressive FlatSpec for pure Scala modules. That said, I would be in favour of some explicitly stated guidelines, because "Lets' keep it in mind" will be forgotten soon. On Mon, Mar 9, 2015 at 8:46 AM, Ufuk Celebi <u...@apache.org> wrote: > Hey Stephan, > > On 08 Mar 2015, at 23:17, Stephan Ewen <se...@apache.org> wrote: > > > Hi everyone! > > > > I would like to start an open discussion about some issue with the > > heterogeneity of the Flink code base. > > Thanks for bringing this up. I agree with your position. The related > discussion about using Guava vs. Validate is a good step into the right > direction. In general, I think it's super hard to get more homogeneity > without enforcing rules (like in the Guava/Validate discussion). I would be > OK with trying to settle on rules and then enforcing them. But I'm not sure > whether that is what you are asking for? Are you more aiming at a "Let's > keep it in mind" kind of thingy?! > > > Here are a few examples: > > > > - Parameter checking is sometimes done with commons-lang3, commons-lang, > > or guava > > - Command line parsing is sometimes done with commons-cli, sometimes with > > scopt. > > I think these are easily enforceable and could also be changed manually > w/o too much hassle. > > > - Code styles are quite different from commit to commit. Spaces, > > indentations, braces. Not a critical thing, but seems to encourage people > > to reformat other people's code, whenever the pass over it, which should > be > > avoided (cluttered diffs, may introduce new bugs actually) > > This is something we could more strictly enforce in pull requests and > generally ask people to refrain from. > > > - Some projects are mixed Java/Scala, which is not perfectly supported by > > the tools so far. It also needs many "fromJava / toJava" conversions and > > makes the entry hurdle into the project higher. > > - Tests are sometimes written as Java Unit tests, sometimes as Scala Unit > > tests (method style), sometimes as Scala Unit Tests (grammar style). > > This is an artifact of the mixed Scala/Java discussion. I agree that this > can be problematic, but I'm not sure how to solve this as long as we mix > Java/Scala in the same modules?! For new code in the runtime, we could > stick to one language. What do you propose here as a solution? > > > I am eager to hear opinions! > > As I've said, I agree with your points, but I think a big issue for new > comers and committers alike is missing documentation in the code. We should > try to keep the discussion we had in that regard in mind as well. > > – Ufuk
