Hi,
 I want to develop my own constructs below PIG Latin language. I would like to take PIG Latin code and run my own Java classes to execute the functions instead of running Hadoop. Can somehow get DAG or any other construct created by PIG Latin interpretor ?
Thanks
Varun

The parser is written in antlr.  You can call that parse just as well as Pig
can.  It remains to be seen whether the resulting data structures are usable
for your purposes.

If not, you can grab the antlr source and strip out the Pig specific code.
 That would allow you to have a DAG pretty easily.

On Thu, Mar 10, 2011 at 11:59 PM, varun srivastava
<[EMAIL PROTECTED]>wrote:

> Hi,
>  I want to develop my own constructs below PIG Latin language. I would like
> to take PIG Latin code and run my own Java classes to execute the functions
> instead of running Hadoop. Can somehow get DAG or any other construct
> created by PIG Latin interpretor ?
> Thanks
> Varun
>
>
>

You can work off either the antlr AST or the logical plan created from the antlr AST .  If you use the logical plan, you can also use pig's logical validation and optimizer rules.
You can have a look at the functions in PigServer.Graph such as parseQuery(), compile() .

Right now, there are no public API's that expose this. So they can change in future.

-Thejas
On 3/10/11 11:59 PM, "varun srivastava" <[EMAIL PROTECTED]> wrote:

Hi,
 I want to develop my own constructs below PIG Latin language. I would like to take PIG Latin code and run my own Java classes to execute the functions instead of running Hadoop. Can somehow get DAG or any other construct created by PIG Latin interpretor ?
Thanks
Varun