I am not opposed to this since we are obviously dependent on Hadoop. -- Brock Noland Sent with Sparrow (http://www.sparrowmailapp.com/?sig)
On Wednesday, August 8, 2012 at 6:22 PM, Jarek Jarcec Cecho wrote: > Maybe let me add few words - I know that it might seem that this method is > meant to be used only for HDFS as it's part of the hadoop, but I was using it > on LocalFileSystem as well without any issues. > > Jarcec > > On Thu, Aug 09, 2012 at 12:44:57AM +0200, Jarek Jarcec Cecho wrote: > > Another possibility would be to utilize FileSystem.delete() from Hadoop: > > > > http://hadoop.apache.org/common/docs/current/api/org/apache/hadoop/fs/FileSystem.html > > > > This class seems to be present in current dependency hadoop-core in profile > > hadoop1 or hadoop-common in profile hadoop2. > > > > Jarcec > > > > On Wed, Aug 08, 2012 at 11:24:05PM +0100, Dave Beech wrote: > > > In a few places now in MRUnit we're creating temp files / directories as > > > part of testing (e.g. dist cache testing, MockMapredOutputFormat). > > > > > > These will obviously need to be cleaned up as part of the test execution. > > > But, from experience I've found deleting files/folders from Java to be > > > pretty unreliable (especially if folders are not empty), so usually I'd > > > use > > > commons-io FileUtils.forceDelete() to get the job done. > > > > > > I'd really like to be able to use this method in MRUnit, but adding a new > > > dependency to the POM for one method just seems.... wrong. > > > > > > I don't know what's worse. Adding the dependency, or re-implementing some > > > file deletion code that's already been done "properly" elsewhere. > > > > > > What's your opinion? > > > > > > Thanks, > > > Dave > > > > > > > > > >
