I agree that pig does not have loop probably for a good reason.

but currently I need to write a code to find the transitive closures of
many edges in a graph.
so I need to iterate a code snippet several times, so finally I can find a
connected component of size 2^N

right now I just copy-paste the snippet several times.

I guess I could take out the snippet and make it into a separate pig
script, and load and store intermediate data
at the beginning and end. but loading data is kind of a waste.

any suggestions?

Thanks
Yang

If I understand you correctly, and you want to find out what the components of a graph are, the trans closure probably is not the way to go as this is quadratic on the number of vertices, hence not scalable. There are other ways that require basically the same number of iterations (order of log(the number of vertices)), and has space requirements that are linear in vertex cardinality.

On Jun 20, 2012, at 7:11 PM, Yang wrote:

> but currently I need to write a code to find the transitive closures of
> many edges in a graph.
> so I need to iterate a code snippet several times, so finally I can find a
> connected component of size 2^N

Would embedding Pig in java or other languages work?

http://pig.apache.org/docs/r0.10.0/cont.html#embed-java
On Jun 20, 2012, at 7:12 PM, Yang <[EMAIL PROTECTED]> wrote:

> I agree that pig does not have loop probably for a good reason.
>
> but currently I need to write a code to find the transitive closures of
> many edges in a graph.
> so I need to iterate a code snippet several times, so finally I can find a
> connected component of size 2^N
>
> right now I just copy-paste the snippet several times.
>
> I guess I could take out the snippet and make it into a separate pig
> script, and load and store intermediate data
> at the beginning and end. but loading data is kind of a waste.
>
> any suggestions?
>
> Thanks
> Yang

Thanks Prashant for the pointer, this is definitely a good approach

On Wed, Jun 20, 2012 at 7:15 PM, Prashant Kommireddi <[EMAIL PROTECTED]>wrote:

> Would embedding Pig in java or other languages work?
>
> http://pig.apache.org/docs/r0.10.0/cont.html#embed-java
>
>
> On Jun 20, 2012, at 7:12 PM, Yang <[EMAIL PROTECTED]> wrote:
>
> > I agree that pig does not have loop probably for a good reason.
> >
> > but currently I need to write a code to find the transitive closures of
> > many edges in a graph.
> > so I need to iterate a code snippet several times, so finally I can find
> a
> > connected component of size 2^N
> >
> > right now I just copy-paste the snippet several times.
> >
> > I guess I could take out the snippet and make it into a separate pig
> > script, and load and store intermediate data
> > at the beginning and end. but loading data is kind of a waste.
> >
> > any suggestions?
> >
> > Thanks
> > Yang
>

Yang -- have you seen Hortonworks' blogpost on this?

http://hortonworks.com/blog/transitive-closure-in-apache-pig/

Norbert

On Wed, Jun 20, 2012 at 10:15 PM, Prashant Kommireddi
<[EMAIL PROTECTED]>wrote:

> Would embedding Pig in java or other languages work?
>
> http://pig.apache.org/docs/r0.10.0/cont.html#embed-java
>
>
> On Jun 20, 2012, at 7:12 PM, Yang <[EMAIL PROTECTED]> wrote:
>
> > I agree that pig does not have loop probably for a good reason.
> >
> > but currently I need to write a code to find the transitive closures of
> > many edges in a graph.
> > so I need to iterate a code snippet several times, so finally I can find
> a
> > connected component of size 2^N
> >
> > right now I just copy-paste the snippet several times.
> >
> > I guess I could take out the snippet and make it into a separate pig
> > script, and load and store intermediate data
> > at the beginning and end. but loading data is kind of a waste.
> >
> > any suggestions?
> >
> > Thanks
> > Yang
>

As mentioned mbedding is the way to go now. We have done it before with bash scripts and our custom workflow system.

On Jun 20, 2012, at 10:20 PM, "Norbert Burger" <[EMAIL PROTECTED]> wrote:

> Yang -- have you seen Hortonworks' blogpost on this?
>
> http://hortonworks.com/blog/transitive-closure-in-apache-pig/
>
> Norbert
>
> On Wed, Jun 20, 2012 at 10:15 PM, Prashant Kommireddi
> <[EMAIL PROTECTED]>wrote:
>
>> Would embedding Pig in java or other languages work?
>>
>> http://pig.apache.org/docs/r0.10.0/cont.html#embed-java
>>
>>
>> On Jun 20, 2012, at 7:12 PM, Yang <[EMAIL PROTECTED]> wrote:
>>
>>> I agree that pig does not have loop probably for a good reason.
>>>
>>> but currently I need to write a code to find the transitive closures of
>>> many edges in a graph.
>>> so I need to iterate a code snippet several times, so finally I can find
>> a
>>> connected component of size 2^N
>>>
>>> right now I just copy-paste the snippet several times.
>>>
>>> I guess I could take out the snippet and make it into a separate pig
>>> script, and load and store intermediate data
>>> at the beginning and end. but loading data is kind of a waste.
>>>
>>> any suggestions?
>>>
>>> Thanks
>>> Yang
>>

Will Duckworth  Senior Vice President, Software Engineering  | comScore, Inc.(NASDAQ:SCOR)
o +1 (703) 438-2108 | m +1 (301) 606-2977 | mailto:[EMAIL PROTECTED]
.....................................................................................................

Introducing Mobile Metrix 2.0 - The next generation of mobile behavioral measurement
www.comscore.com/MobileMetrix

hehehe, thanks, I'll post my version slightly later :)

On Wed, Jun 20, 2012 at 7:19 PM, Norbert Burger <[EMAIL PROTECTED]>wrote:

> Yang -- have you seen Hortonworks' blogpost on this?
>
> http://hortonworks.com/blog/transitive-closure-in-apache-pig/
>
> Norbert
>
> On Wed, Jun 20, 2012 at 10:15 PM, Prashant Kommireddi
> <[EMAIL PROTECTED]>wrote:
>
> > Would embedding Pig in java or other languages work?
> >
> > http://pig.apache.org/docs/r0.10.0/cont.html#embed-java
> >
> >
> > On Jun 20, 2012, at 7:12 PM, Yang <[EMAIL PROTECTED]> wrote:
> >
> > > I agree that pig does not have loop probably for a good reason.
> > >
> > > but currently I need to write a code to find the transitive closures of
> > > many edges in a graph.
> > > so I need to iterate a code snippet several times, so finally I can
> find
> > a
> > > connected component of size 2^N
> > >
> > > right now I just copy-paste the snippet several times.
> > >
> > > I guess I could take out the snippet and make it into a separate pig
> > > script, and load and store intermediate data
> > > at the beginning and end. but loading data is kind of a waste.
> > >
> > > any suggestions?
> > >
> > > Thanks
> > > Yang
> >
>

well, I tried to give a pure pig latin version without using Udf, but now
it seems too cumbersome without udf:
2 of the problems I can't solve is:
for a bag [ (id : int,  extra_info: chararray)], how to generate the tuple
with the smallest id.

another is : is there  a version of MIN() for chararray?

Thanks
yang

On Wed, Jun 20, 2012 at 7:22 PM, Yang <[EMAIL PROTECTED]> wrote:

> hehehe, thanks, I'll post my version slightly later :)
>
>
> On Wed, Jun 20, 2012 at 7:19 PM, Norbert Burger <[EMAIL PROTECTED]>wrote:
>
>> Yang -- have you seen Hortonworks' blogpost on this?
>>
>> http://hortonworks.com/blog/transitive-closure-in-apache-pig/
>>
>> Norbert
>>
>> On Wed, Jun 20, 2012 at 10:15 PM, Prashant Kommireddi
>> <[EMAIL PROTECTED]>wrote:
>>
>> > Would embedding Pig in java or other languages work?
>> >
>> > http://pig.apache.org/docs/r0.10.0/cont.html#embed-java
>> >
>> >
>> > On Jun 20, 2012, at 7:12 PM, Yang <[EMAIL PROTECTED]> wrote:
>> >
>> > > I agree that pig does not have loop probably for a good reason.
>> > >
>> > > but currently I need to write a code to find the transitive closures
>> of
>> > > many edges in a graph.
>> > > so I need to iterate a code snippet several times, so finally I can
>> find
>> > a
>> > > connected component of size 2^N
>> > >
>> > > right now I just copy-paste the snippet several times.
>> > >
>> > > I guess I could take out the snippet and make it into a separate pig
>> > > script, and load and store intermediate data
>> > > at the beginning and end. but loading data is kind of a waste.
>> > >
>> > > any suggestions?
>> > >
>> > > Thanks
>> > > Yang
>> >
>>
>
>