JEP proposed to target JDK 11: 323: Local-Variable Syntax for Lambda Parameters (original) (raw)

Jeremy Manson jeremymanson at google.com
Mon Jan 29 05:43:49 UTC 2018

We do not require method parameters to be final. It isn't referenced at all in our style guide, but we have very copious internal Java best practices documentation; this is a very slightly modified version of the relevant passage:

On parameters and local variableslink <http://go/java-practices/final#paramslocals>modeedit <https://cs.corp.google.com/#piper///depot/eng/doc/devguide/java/practices/final.md&edit=1&l=76> comment <https://engdoc.corp.google.com/eng/doc/devguide/java/practices/final.md?cl=head#> It would be nice if Java made parameters and local variables final by default. Since the need to allow reassignment is both more rare and more risky, it makes sense to be opt-in. Unfortunately, Java got this backward. Some developers prefer to mark all their parameters and local variables final unless they are reassigned. There are good reasons for this, but it is not clear that the advantages outweigh the costs (lower signal-to-noise, and especially burden of enforcement)—and so this practice is not favored at Google. Remember that the dangers of sloppy reassignment of locals will be limited if you always keep your methods short and focused. And note that as of Java 8 it's no longer necessary to use final on local variables to make them accessible from an inner class. Of course, you can still use final on a local variable when it serves some specific purpose. For example, you might use a blank final <https://en.wikipedia.org/wiki/Final(Java)#Blankfinal> to signal that while you do intend to "modify" (set) the value later, you'll do it only once. Available solution: projects can opt in <http://go/errorprone/bugpatterns#optional> to @Var Error Prone enforcement <http://errorprone.info/bugpattern/Var>. This inverts the default: the Error Prone check requires all modified parameters and local variables to be explictly annotated with @Var.

Jeremy

On Sun, Jan 28, 2018 at 4:02 PM, Jonathan Bluett-Duncan < jbluettduncan at gmail.com> wrote:

Hi Rémi,

You may find my (short) response inline. On 28 January 2018 at 23:47, Remi Forax <forax at univ-mlv.fr> wrote: > Hi Andrew, > > ----- Mail original ----- > > De: "Andrew Haley" <aph at redhat.com> > > À: "Zheka Kozlov" <orionllmain at gmail.com>, "Martijn Verburg" <_ _> martijnverburg at gmail.com> > > Cc: "jdk-dev" <jdk-dev at openjdk.java.net> > > Envoyé: Vendredi 26 Janvier 2018 16:09:34 > > Objet: Re: JEP proposed to target JDK 11: 323: Local-Variable Syntax for > Lambda Parameters > > > On 26/01/18 09:59, Zheka Kozlov wrote: > >> The motivation is to allow final and annotations on lambda parameters: > >> > >> (final var x) → x + 1; > >> (@Nonnull var x) → x + 1; > > > > I'm trying to understand how it makes sense to have something > > that's both variable and final. > > yes, 'final var' is an odd combination, 'var' means more this is a local > variable and less this is something which is variable. > > In Java, you can declare a parameter of a method final which means it's a > local variable that can be only assigned once, given that parameters are > assigned with the calling arguments, it means that this parameter can not > be re-assigned. This is just a compile time 'annotation', because marking a > parameter or a local variable final is not something which is stored in the > classfile (apart if you compile with the -parameters, but most people > don't). > > Some coding conventions, mostly the one from Google mandates that all > method parameters are declared final, so since Java 8, some developers have > asked to be able to declare lambda parameters final. > Oh, do Google's coding conventions really mandate this practice? I've had a look through the Google Java Style guide, but I can't seem to find anything to confirm it. Do you have a source that I can check?

> > In a perfect World, Google should update its coding convention because > mandating to declare local variables final has the nasty side effect of > blurring the distinction between final for a field which as you know has > effects defined by the memory model and final on local variable which has > no effect. So mandating to declare local variables final is harmful. > > > > > -- > > Andrew Haley > > Java Platform Lead Engineer > > Red Hat UK Ltd. <https://www.redhat.com> > > EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671 > > cheers, > Rémi > Cheers, Jonathan

More information about the jdk-dev mailing list