Thanks Denny. If you don't have an idea, I will document that in the pom.xml of 
the scribe source, and in the ScribeSource.java file, so the next time thrift 
is upgraded - the developer can manually change the string to a ByteBuffer. 
Does that make sense to you? 


Hari 

-- 
Hari Shreedharan


On Tuesday, February 5, 2013 at 12:08 AM, Denny Ye wrote:

> OK, I'm not sure which is better method to keep the effective code during
> upgrading.
> I will post it if I have a good idea.
> 
> 
> 2013/2/5 Hari Shreedharan <[email protected] 
> (mailto:[email protected])>
> 
> > OK, no problem. I modified the Scribe Source to simply pick up the String
> > and convert it to a byte array. Unfortunately, since we need to maintain
> > the generated code even in between Thrift upgrades, we should probably not
> > edit the generated code by hand. I am actually OK with modifying that class
> > to use ByteBuffer, but just document that when we do a Thrift upgrade, we
> > must change the LogEntry class manually - this is not a good idea, but if
> > you think that makes it effective and gives better performance I won't
> > complain. Do you think you could do this using a script (find/replace), so
> > we can add ant-run plugin to the pom.xml to handle this case?
> > 
> > 
> > Thanks
> > Hari
> > 
> > --
> > Hari Shreedharan
> > 
> > 
> > On Monday, February 4, 2013 at 11:42 PM, Denny Ye wrote:
> > 
> > > hi Hari, the original version of generated code from that Thrift template
> > > has only support message with 'String' type, I think it's ineffective
> > > 
> > 
> > that
> > > deserialized from byte stream to String again. That's CPU sensitive work
> > > and useless in progress of data flow. Thus, I modified code of LogEntry
> > > 
> > 
> > to
> > > support byte array incoming from socket except deserializing.
> > > 
> > > What should I do next?
> > > 
> > > 
> > > 2013/2/5 Hari Shreedharan <[email protected] 
> > > (mailto:[email protected]) (mailto:
> > [email protected] (mailto:[email protected]))>
> > > 
> > > > Denny,
> > > > 
> > > > I modified the file's namespace and the final file looks like this:
> > > > 
> > > > namespace java org.apache.flume.source.scribe
> > > > 
> > > > enum ResultCode
> > > > {
> > > > OK,
> > > > TRY_LATER
> > > > }
> > > > 
> > > > struct LogEntry
> > > > {
> > > > 1: string category,
> > > > 2: string message
> > > > }
> > > > 
> > > > service Scribe
> > > > {
> > > > ResultCode Log(1: list<LogEntry> messages);
> > > > }
> > > > 
> > > > 
> > > > 
> > > > But it looks like the ScribeSource.java (line 154) expects
> > > > LogEntry.message to be "binary" and not "string." Since there are no
> > > > 
> > > 
> > > 
> > 
> > unit
> > > > tests, I cannot verify that making message binary will not break
> > > 
> > 
> > something.
> > > > Could you give me some fix for this?
> > > > 
> > > > 
> > > > Thanks
> > > > Hari
> > > > 
> > > > --
> > > > Hari Shreedharan
> > > > 
> > > > 
> > > > On Monday, February 4, 2013 at 8:05 PM, Hari Shreedharan wrote:
> > > > 
> > > > > Hi Denny, Juhani,
> > > > > 
> > > > > Thanks! I will add this file to the compileThrift profile, so we can
> > > > update the thrift generated code when we upgrade thrift.
> > > > > 
> > > > > Thanks!
> > > > > Hari
> > > > > 
> > > > > --
> > > > > Hari Shreedharan
> > > > > 
> > > > > 
> > > > > On Monday, February 4, 2013 at 6:46 PM, Denny Ye wrote:
> > > > > 
> > > > > > Sorry for late response
> > > > > > 
> > > > > > Here is the scribe.thrift file :
> > > > > > ************************* file separator *********************
> > > > > > namespace java scribe.thrift
> > > > > > 
> > > > > > enum ResultCode
> > > > > > {
> > > > > > OK,
> > > > > > TRY_LATER
> > > > > > }
> > > > > > 
> > > > > > struct LogEntry
> > > > > > {
> > > > > > 1: string category,
> > > > > > 2: string message
> > > > > > }
> > > > > > 
> > > > > > service scribe
> > > > > > {
> > > > > > ResultCode Log(1: list<LogEntry> messages);
> > > > > > }
> > > > > > ************************* file separator *********************
> > > > > > 
> > > > > > To Juhani, I haven't use the facebook template, there have so many
> > > > > > additional interfaces that I don't like it.
> > > > > > 
> > > > > > 
> > > > > > 
> > > > > > 2013/2/5 Juhani Connolly <[email protected] 
> > > > > > (mailto:[email protected])(mailto:
> > [email protected] (mailto:[email protected])) 
> > (mailto:
> > > > [email protected] 
> > > > (mailto:[email protected]) (mailto:
> > > 
> > 
> > [email protected] 
> > (mailto:[email protected])))>
> > > > > > 
> > > > > > > This is the file we want:
> > https://github.com/facebook/**scribe/blob/master/if/scribe.**thrift<
> > > > https://github.com/facebook/scribe/blob/master/if/scribe.thrift>
> > > > > > > 
> > > > > > > It's apache 2 licensed but also has a facebooke copyright
> > header. I
> > > > think
> > > > > > > this is why we didn't originally include it.
> > > > > > > 
> > > > > > > 
> > > > > > > On 02/03/2013 09:14 AM, Hari Shreedharan wrote:
> > > > > > > 
> > > > > > > > Hi all,
> > > > > > > > 
> > > > > > > > The scribe source has some thrift generated code committed,
> > but no
> > > > > > > > ".thrift" file. This makes upgrading to a new version of
> > > > > > > 
> > > > > > 
> > > > > 
> > > > 
> > > 
> > 
> > libthrift
> > > > > > > > impossible.
> > > > > > > > 
> > > > > > > > I believe Denny Ye contributed this code, so Denny, it'd be
> > nice
> > > > if you
> > > > > > > > could submit the ".thrift" file as soon as possible.
> > > > > > > > 
> > > > > > > > 
> > > > > > > > Thanks,
> > > > > > > > Hari
> > > > > > > > 
> > > > > > > 
> > > > > > 
> > > > > 
> > > > 
> > > > 
> > > 
> > 
> > 
> 
> 
> 


Reply via email to