Generic sequences in seq.el

Generic sequences in seq.el

[Yuan Fu]

[-- Attachment #1: Type: text/plain, Size: 784 bytes --]

I'm trying to "upgrade" some methods from supporting list to supporting
generic sequence.
However, I couldn't figure out how to change the value of a sequence
destructively.
E.g., if I want to insert a element into a list, I can do something like

    (setf (nthcdr n seq) (cons element (nthcdr n seq)))

to insert element into the nth position in seq

But with sequences, I don't see any generic functions that can do similar
work.
Am I missing something?

BTW, what I'm doing now is define my own generic function, like

    (cl-defgeneric my-insert-at (elt seq)))

The idea is I'll implement this function for the generic sequence
that I'll define in the future. Is that a correct approach?
Other destructive functions that I couldn't find alternatives are delete,
push, pop, etc.

[-- Attachment #2: Type: text/html, Size: 2165 bytes --]

Re: Generic sequences in seq.el

[Stefan Monnier]

>     (setf (nthcdr n seq) (cons element (nthcdr n seq)))
> to insert element into the nth position in seq

I think with `seq` you'd do something like

   (seq-concatenate 'list (seq-subseq seq 0 n)
                          (list element)
                          (seq-subseq seq n))

Of course, it's different because it's non-destructive, but insertion
into an array can't be done non-destructively anyway.

> The idea is I'll implement this function for the generic sequence
> that I'll define in the future. Is that a correct approach?
> Other destructive functions that I couldn't find alternatives are delete,
> push, pop, etc.

We currently don't have any data-structure that can be efficiently
modified in-place via push/pop (note that even lists don't qualify
because push/pop doesn't modify the list in place: they replace a list
with another without modifying any existing list (unless the PLACE is
the car/cdr of a list, of course, but that's a separate issue)).


        Stefan

Re: Generic sequences in seq.el

[Yuan Fu]

[-- Attachment #1: Type: text/plain, Size: 1763 bytes --]

>   (seq-concatenate 'list (seq-subseq seq 0 n)
>                          (list element)
>                          (seq-subseq seq n))

That returns a list, not the sequence. So inserting something into the
sequence
makes it not a sequence anymore?

> We currently don't have any data-structure that can be efficiently
> modified in-place via push/pop (note that even lists don't qualify
> because push/pop doesn't modify the list in place: they replace a list
> with another without modifying any existing list (unless the PLACE is
> the car/cdr of a list, of course, but that's a separate issue)).

Thanks, it's good to know.

Yuan.

On Fri, Dec 21, 2018 at 9:24 AM Stefan Monnier <monnier@iro.umontreal.ca>
wrote:

> >     (setf (nthcdr n seq) (cons element (nthcdr n seq)))
> > to insert element into the nth position in seq
>
> I think with `seq` you'd do something like
>
>    (seq-concatenate 'list (seq-subseq seq 0 n)
>                           (list element)
>                           (seq-subseq seq n))
>
> Of course, it's different because it's non-destructive, but insertion
> into an array can't be done non-destructively anyway.
>
> > The idea is I'll implement this function for the generic sequence
> > that I'll define in the future. Is that a correct approach?
> > Other destructive functions that I couldn't find alternatives are delete,
> > push, pop, etc.
>
> We currently don't have any data-structure that can be efficiently
> modified in-place via push/pop (note that even lists don't qualify
> because push/pop doesn't modify the list in place: they replace a list
> with another without modifying any existing list (unless the PLACE is
> the car/cdr of a list, of course, but that's a separate issue)).
>
>
>         Stefan
>
>
>

[-- Attachment #2: Type: text/html, Size: 4637 bytes --]

Re: Generic sequences in seq.el

[Stefan Monnier]

>>   (seq-concatenate 'list (seq-subseq seq 0 n)
>>                          (list element)
>>                          (seq-subseq seq n))
>
> That returns a list, not the sequence.

I'm not sure which sequence is "the" sequence, but a list is
definitely a sequence.


        Stefan

Re: Generic sequences in seq.el

[Yuan Fu]

[-- Attachment #1: Type: text/plain, Size: 623 bytes --]

You are right. I can implement my own
method to return objects that are not vector, string or list.

Still, it would be convenient to have a function
that can destructively modify a sequence. What is
the reason to not have one?

Yuan

On Sun, Dec 23, 2018 at 9:34 AM Stefan Monnier <monnier@iro.umontreal.ca>
wrote:

> >>   (seq-concatenate 'list (seq-subseq seq 0 n)
> >>                          (list element)
> >>                          (seq-subseq seq n))
> >
> > That returns a list, not the sequence.
>
> I'm not sure which sequence is "the" sequence, but a list is
> definitely a sequence.
>
>
>         Stefan
>

[-- Attachment #2: Type: text/html, Size: 1623 bytes --]

Re: Generic sequences in seq.el

[Stefan Monnier]

> Still, it would be convenient to have a function
> that can destructively modify a sequence. What is
> the reason to not have one?

Two reasons:
1- Side-effects are bad.
2- There hasn't yet been a compelling reason to add one.
3- As mentioned before "We currently don't have any data-structure that
   can be efficiently modified in-place via push/pop".


        Stefan