Thanks Daniel. Of course, you are right. Turns out I had a bug elsewhere in my
UDF that was making me think this was not working correctly. After fixing that
bug the "define ..." works fine.
Using the following works great:
define INITIALIZED_UDF com.my.udfs.UDF(constructor_params)
Thanks,
Zach
On Tuesday, November 30, 2010 at 4:40 PM, Daniel Dai wrote:
> Pig always instantiate UDF using the construct parameter defined in
> "define" statement. ". CONTAINS_STRINGS(haystack) only pass haystack to
> CONTAINS_STRINGS.exec(). It will not re-initializing the UDF.
>
> Daniel
>
> Zach Bailey wrote:
>
> > I am trying to do what seems like should be a simple task using pig and a
> > UDF I have written but can't seem to figure out the syntax to get it
> > working.
> >
> >
> > At a high level I have a UDF that takes a number of strings that I then
> > want to see if exist in some other strings. So I write a UDF called
> > CONTAINS_ANY that I would like to initialize with the "needles" and then
> > use pig to distribute this search out via hadoop.
> >
> >
> > The problem is I can't figure out what the correct syntax is to initialize
> > the UDF with the "needles" and then use the UDF later once it has been
> > initialized. I have tried the following syntax:
> >
> >
> > define CONTAINS_STRINGS com.my.piggybank.CONTAINS_ANY('string1|string2');
> >
> >
> > and then invoking this by doing
> >
> >
> > filtered = FILTER data BY CONTAINS_STRINGS(haystack);
> >
> >
> > but this ends up re-initializing the UDF with the strings from the
> > haystack which is not what I wanted.
> >
> >
> > Essentially I want to be able to write a UDF that is like the built-in
> > MATCHES function so I can say something like:
> >
> >
> > filtered = FILTER data by haystack CONTAINS_STRINGS('string1|string2');
> >
> >
> > but so far have been unable to find any useful/relevant documentation on
> > how to accomplish this.
> >
> >
> > Thanks a lot for any pointers or help anyone can give.
> >
> >
> > Best,
> > Zach
> >
> >
> >
> >
>
>
>
>
