PROPOSAL : Specify nullable argument (original) (raw)

Reinier Zwitserloot reinier at zwitserloot.com
Tue Mar 17 19:58:06 PDT 2009

That sounds like great news. I'll do some more research.

--Reinier Zwitserloot

On Mar 18, 2009, at 03:49, Jeremy Manson wrote:

I don't know what the current state of the 305 proposal is, but Bill has been talking about at least three annotations in informal conversations.

Jeremy On Tue, Mar 17, 2009 at 5:27 PM, Reinier Zwitserloot <reinier at zwitserloot.com> wrote: Adding non-nullity to the type system would be a great boon, IMO, but its far more complicated than just having a @NonNull annotation.

I've written up a proposal a while ago here: http://www.zwitserloot.com/2008/10/23/non-null-in-static-languages/ Suffice to say it gets very complicated because you need 4 nullity states (Never-Allows-Null, Definitely-Allows-Null, Might-Allow-Null, Legacy (like raw generics, for interop with old code), AND you need a separate process to promote or demote generics (So that you can say: Eventhough this is a Map that maps NonNull String to NonNull String, if you call get() on this, you get a @Nullable String back, because null will flow out if I can't find the key. In other words, the 'V' generics parameter needs to be demoted to @Nullable V somehow). Read the blog entry for a more elaborate explanation with examples. I'd love something like that in java, but the proposal as it stands needs a source 1.7; keyword (because having no modifier is taken as 'never null', which is obviously not backwards compatible), and is very complicated, so: Definitely not project coin material. Incidentally, I think jsr305 is fundamentally broken. It doesn't have the states needed to convey nullity info; it needs at least 2 annotations, probably 3 to be able to differentiate between code that simply didn't specify anything, and code that did, instead of the one annotation that jsr305 gives. --Reinier Zwitserloot

On Mar 17, 2009, at 21:41, Neal Gafter wrote: Did you intend this to be a complete proposal? Do you like this better than @NonNull, which will probably be included in Java 7 under jsr 305? On Tue, Mar 17, 2009 at 1:25 PM, Olivier Chorier <lapsus63 at gmail.com> wrote: Example : public void getListOfProducts(Company owner, !String facultativeName) { .... } The '!' should prevent the developper that the argument is able to be null. Invoking getListOfProducts(null, "name") could throw a compilation error/warning. Or maybe better : public void getListOfProducts(!Company owner, String facultativeName) { .... } Here, the '!' indicates a mandatory argument (not null).

However, I don't know if the '!' marker could be the most appropriate.

More information about the coin-dev mailing list