Agreed, same with including a pig-withouthadoop jar. We do publish
such a jar already in maven. This is a non-critical (by which I mean
it doesn't cause failure or wrong results, not that it isn't
important) issue and there is not agreement yet on whether the without
jar should be included in the tar ball. So we shouldn't hold a
release for it.
Alan.
On Jun 23, 2011, at 3:21 PM, Dmitriy Ryaboy wrote:
Moving discussion to a discuss thread so votes can stay on the vote
thread.
Richa, I don't feel those two issues you point out are blockers for a
release, but a patch for 0.10 would be welcome to fix the hardcoding
issue
(gosh I can't imagine how you found that.. :)).
D
On Thu, Jun 23, 2011 at 3:04 PM, Richa Khandelwal
<[email protected]>wrote:
Hi Guys,
I would vote +1 to remove the hadoop jar from the pig.jar. That would
definitely add to the flexibility of using Pig with any version or
variant
of hadoop.
Also, there is "hdfs://" hardcoded at a few places in the pig
source. I can
provide a patch which removes the hard-coding and makes it possible
to use
pig on any distributed filesystem that can be used with hadoop,
based on
pig.properties configuration file. Please let me know.
Thanks,
Richa
On Thu, Jun 23, 2011 at 2:56 PM, Olga Natkovich <[email protected]>
wrote:
My experience with sample command is that people use it for some
quick
testing and debugging not for production problems. This issue has
been
there
since the initial introduction of sample. I still don't see a
strong need
to
hold the beta release for it.
Olga
-----Original Message-----
From: Dmitriy Ryaboy [mailto:[email protected]]
Sent: Thursday, June 23, 2011 10:14 AM
To: [email protected]
Cc: [email protected]
Subject: Re: [VOTE] Release Pig 0.9.0 (candidate 0)
There is a workaround - turning off optimizations, or at least
pushUpFilter. I haven't checked 8 with the new logical plan
disabled,
that
may work too.
If this failed execution, I'd let it pass since there is a
workaround,
but
as is it silently returns incorrect data, with no way for an
analyst to
know
her statistics are now fundamentally wrong. Silent and subtle data
corruption is just about the wrist big we can have.
I can't really block the release, you guys can outvote me. But I'd
rather
you didn't; we can patch this problem today.
On Jun 23, 2011, at 7:47 AM, Alan Gates <[email protected]> wrote:
Are you referring to PIG-2137? I have a few of questions on that
before
I vote for this release candidate or to reroll.
Is this a new issue introduced in 0.9?
Is there a workaround for this?
We have already discussed that 0.9.0 will be beta quality, and a
follow
up release will be needed as users find bugs. As sample is not a
heavily
used feature I am inclined to view this bug as ok. You feel this is
serious
enough to block a beta release?
Alan.
On Jun 22, 2011, at 9:11 PM, Dmitriy Ryaboy wrote:
-1
I discovered a critical bug in how SAMPLE is treated; I don't
think we
should release until it's fixed (Thejas is testing the fix).
D
On Wed, Jun 22, 2011 at 5:56 PM, Olga Natkovich <[email protected]
>
wrote:
I have created a candidate build for Pig 0.9.0 release. This
release
introduces control structures, changes query parser, and performs
semantic
cleanup.
The rat report showed no issues in Java files outside of build
directory.
Keys used to sign the release are available at
http://svn.apache.org/viewvc/pig/trunk/KEYS?view=markup.
Please try it out:
http://people.apache.org/~olga/pig-0.9.0-candidate-0/
Build is also available in maven:
https://repository.apache.org/content/repositories/orgapachepig-042/org/apache/pig/pig/0.9.0/
Should we release this? Vote closes on Monday, June 27.
Olga
