Ok thanks

On Wed, Feb 16, 2011 at 3:35 PM, Ramesh, Amit <[email protected]> wrote:

>
> No, joins are only possible on fields common to all the aliases in the
> join.
>
>
> On 2/16/11 2:56 PM, "sonia gehlot" <[email protected]> wrote:
>
> > Thank you very much Amit.
> >
> > One more question in the same way if I want to join multiple tables
> >
> >> select blah, blah
> >> From
> >> page_events pe
> >   Left Join referrer ref
> >         on ref.id = pe.id
> >> Left Join page_events pe_pre
> >> on pe.day = pe_pre.day
> >> And pe.session_id = pe_pre.session_id
> >> And pe.page_seq_num = pe_pre.page_seq_num + 1
> >
> > Can we do this in one statement in Pig?
> >
> > Thanks again for your help.
> >
> > Sonia
> >
> > On Wed, Feb 16, 2011 at 2:40 PM, Ramesh, Amit <[email protected]>
> wrote:
> >
> >>
> >> You can just do:
> >>
> >> join_pe_pre = JOIN page_events BY (day, session_id, page_seq_num) LEFT
> >> OUTER, page_events_pre BY (day, session_id, page_seq_num + 1);
> >>
> >> Amit
> >>
> >>
> >> On 2/16/11 2:09 PM, "sonia gehlot" <[email protected]> wrote:
> >>
> >>> Hi All,
> >>>
> >>> I am new to Hadoop and I started exploring Pig since last month. I have
> >> few
> >>> question I have to replicate some SQL query to Pig that has left join
> for
> >>> example:
> >>>
> >>> select blah, blah
> >>> From
> >>> page_events pe
> >>> Left Join page_events pe_pre
> >>> on pe.day = pe_pre.day
> >>> And pe.session_id = pe_pre.session_id
> >>> And pe.page_seq_num = pe_pre.page_seq_num + 1
> >>>
> >>> So I wanted to confirm is this is the right and only way to do multi
> >> column
> >>> join in Pig? Or we can do this in some other way?
> >>>
> >>> join1_pe_pre = JOIN page_events BY day LEFT OUTER,  page_events_pre BY
> >> day ;
> >>>
> >>> join2_pe_pre = JOIN join1_pe_pre BY session_id LEFT OUTER,
> >> page_events_pre
> >>> BY session_id ;
> >>>
> >>> join3_pe_pre = JOIN join2_pe_pre BY page_seq_num LEFT OUTER,
> >> page_events_pre
> >>> BY page_seq_num +1 ;
> >>>
> >>> Thanks for your help.
> >>>
> >>> Sonia
> >>
> >>
>
>

Reply via email to