Re: Backwards compatibility of FileSystem interface.

2013-05-31 Thread Steve Loughran
(forwarding to common-dev which is where this discussion about APIs should go) 1. FileSystem API may add new methods from version to version. There's some list operation that's gone in too. 2. I don't think there are any guarantees that we don't add new methods to a base class -it's w

Re: Backwards compatibility of FileSystem interface.

2013-05-31 Thread Jay Vyas
For context - It appears that over time the FileSystem has changed the signature to the "createNonRecursive" method, (in 2.x there are 3 signatures, in 1.x, there are 2 signatures). [root@jvyas hadoop-common]# git checkout branch-1.1 Switched to branch 'branch-1.1' [root@jvyas hadoop-common]# grep

Re: Backwards compatibility of FileSystem interface.

2013-05-31 Thread Steve Loughran
Jay, this is is more a common-dev question, as it goes beyond hdfs On 31 May 2013 02:07, Jay Vyas wrote: > Hi : > > Are FileSystem interfaces gauranteed to retain semantics of the old > FileSystem contract? > retain the explicit semantics, yes. Implementations may change some implementation det

Backwards compatibility of FileSystem interface.

2013-05-30 Thread Jay Vyas
Hi : Are FileSystem interfaces gauranteed to retain semantics of the old FileSystem contract? Im noticing that the new FileSystem uses the "CreateFlag" class, which older ones did not, and I'm wondering wether the deprecation of some FileSystem methods (i.e. createNonRecursive) will eventually be