This works:

a = load 'thing';
b = foreach (group a all) {
  sorted = order a by $0;
  top = limit sorted 100;
  generate flatten(top);
}

This does not (neither in pig8 nor in pig9, to my knowledge):

a = load 'thing';
b = foreach (group a all) {
  top = limit (order a by $0) 100;
  generate flatten(top);
}

Just curious if this is a design choice, design limitation, annoying to
implement, etc. Thought I'd bring it up.

Jon, your man for "nested foreach" posts to the dev listserv

I think the answer is:
annoying to implement and with not so compelling use cases.

The nested expressions take a different path in the parser in order to
distinguish them from the non-nested ones. So you would probably need to
replicate some code here and there to make this work. It can be done if you
think it is worth doing.

Cheers,
--
Gianmarco

On Fri, Nov 11, 2011 at 21:22, Jonathan Coveney <[EMAIL PROTECTED]> wrote:

> This works:
>
> a = load 'thing';
> b = foreach (group a all) {
>  sorted = order a by $0;
>  top = limit sorted 100;
>  generate flatten(top);
> }
>
> This does not (neither in pig8 nor in pig9, to my knowledge):
>
> a = load 'thing';
> b = foreach (group a all) {
>  top = limit (order a by $0) 100;
>  generate flatten(top);
> }
>
> Just curious if this is a design choice, design limitation, annoying to
> implement, etc. Thought I'd bring it up.
>
> Jon, your man for "nested foreach" posts to the dev listserv
>

It's not the most important issue by far (I think the scoped names issue is
more frustrating, as far as usability), it's just a bit counterintuitive
that the syntax would not be consistent. What reason is there (besides
implementation) for a user to assume that syntax inside of the foreach and
outside of the foreach would be different (except for the nested foreach
which is a documented technical limitation that someone is working on)?

2011/11/12 Gianmarco De Francisci Morales <[EMAIL PROTECTED]>

> I think the answer is:
> annoying to implement and with not so compelling use cases.
>
> The nested expressions take a different path in the parser in order to
> distinguish them from the non-nested ones. So you would probably need to
> replicate some code here and there to make this work. It can be done if you
> think it is worth doing.
>
> Cheers,
> --
> Gianmarco
>
>
>
> On Fri, Nov 11, 2011 at 21:22, Jonathan Coveney <[EMAIL PROTECTED]>
> wrote:
>
> > This works:
> >
> > a = load 'thing';
> > b = foreach (group a all) {
> >  sorted = order a by $0;
> >  top = limit sorted 100;
> >  generate flatten(top);
> > }
> >
> > This does not (neither in pig8 nor in pig9, to my knowledge):
> >
> > a = load 'thing';
> > b = foreach (group a all) {
> >  top = limit (order a by $0) 100;
> >  generate flatten(top);
> > }
> >
> > Just curious if this is a design choice, design limitation, annoying to
> > implement, etc. Thought I'd bring it up.
> >
> > Jon, your man for "nested foreach" posts to the dev listserv
> >
>

I think there is no good reason to leave it like this.
I was already thinking about making a major update to Pig grammar in order
to simplify and rationalize it.
See also PIG-2138 for example.
If I manage to bring myself to work on it, this improvement should
definitely go in as well.

Could you open a Jira to keep track of the idea?

Cheers,
--
Gianmarco

On Sun, Nov 13, 2011 at 02:25, Jonathan Coveney <[EMAIL PROTECTED]> wrote:

> It's not the most important issue by far (I think the scoped names issue is
> more frustrating, as far as usability), it's just a bit counterintuitive
> that the syntax would not be consistent. What reason is there (besides
> implementation) for a user to assume that syntax inside of the foreach and
> outside of the foreach would be different (except for the nested foreach
> which is a documented technical limitation that someone is working on)?
>
> 2011/11/12 Gianmarco De Francisci Morales <[EMAIL PROTECTED]
> >
>
> > I think the answer is:
> > annoying to implement and with not so compelling use cases.
> >
> > The nested expressions take a different path in the parser in order to
> > distinguish them from the non-nested ones. So you would probably need to
> > replicate some code here and there to make this work. It can be done if
> you
> > think it is worth doing.
> >
> > Cheers,
> > --
> > Gianmarco
> >
> >
> >
> > On Fri, Nov 11, 2011 at 21:22, Jonathan Coveney <[EMAIL PROTECTED]>
> > wrote:
> >
> > > This works:
> > >
> > > a = load 'thing';
> > > b = foreach (group a all) {
> > >  sorted = order a by $0;
> > >  top = limit sorted 100;
> > >  generate flatten(top);
> > > }
> > >
> > > This does not (neither in pig8 nor in pig9, to my knowledge):
> > >
> > > a = load 'thing';
> > > b = foreach (group a all) {
> > >  top = limit (order a by $0) 100;
> > >  generate flatten(top);
> > > }
> > >
> > > Just curious if this is a design choice, design limitation, annoying to
> > > implement, etc. Thought I'd bring it up.
> > >
> > > Jon, your man for "nested foreach" posts to the dev listserv
> > >
> >
>

Here we go:

https://issues.apache.org/jira/browse/PIG-2360

I need to work on getting up to speed on how ANTLR works...

2011/11/13 Gianmarco De Francisci Morales <[EMAIL PROTECTED]>

> I think there is no good reason to leave it like this.
> I was already thinking about making a major update to Pig grammar in order
> to simplify and rationalize it.
> See also PIG-2138 for example.
> If I manage to bring myself to work on it, this improvement should
> definitely go in as well.
>
> Could you open a Jira to keep track of the idea?
>
> Cheers,
> --
> Gianmarco
>
>
>
> On Sun, Nov 13, 2011 at 02:25, Jonathan Coveney <[EMAIL PROTECTED]>
> wrote:
>
> > It's not the most important issue by far (I think the scoped names issue
> is
> > more frustrating, as far as usability), it's just a bit counterintuitive
> > that the syntax would not be consistent. What reason is there (besides
> > implementation) for a user to assume that syntax inside of the foreach
> and
> > outside of the foreach would be different (except for the nested foreach
> > which is a documented technical limitation that someone is working on)?
> >
> > 2011/11/12 Gianmarco De Francisci Morales <
> [EMAIL PROTECTED]
> > >
> >
> > > I think the answer is:
> > > annoying to implement and with not so compelling use cases.
> > >
> > > The nested expressions take a different path in the parser in order to
> > > distinguish them from the non-nested ones. So you would probably need
> to
> > > replicate some code here and there to make this work. It can be done if
> > you
> > > think it is worth doing.
> > >
> > > Cheers,
> > > --
> > > Gianmarco
> > >
> > >
> > >
> > > On Fri, Nov 11, 2011 at 21:22, Jonathan Coveney <[EMAIL PROTECTED]>
> > > wrote:
> > >
> > > > This works:
> > > >
> > > > a = load 'thing';
> > > > b = foreach (group a all) {
> > > >  sorted = order a by $0;
> > > >  top = limit sorted 100;
> > > >  generate flatten(top);
> > > > }
> > > >
> > > > This does not (neither in pig8 nor in pig9, to my knowledge):
> > > >
> > > > a = load 'thing';
> > > > b = foreach (group a all) {
> > > >  top = limit (order a by $0) 100;
> > > >  generate flatten(top);
> > > > }
> > > >
> > > > Just curious if this is a design choice, design limitation, annoying
> to
> > > > implement, etc. Thought I'd bring it up.
> > > >
> > > > Jon, your man for "nested foreach" posts to the dev listserv
> > > >
> > >
> >
>