Shallow copy of a record

Shallow copy of a record

[Chris King]

Maybe I missed it in the docs, but is there a way to create a shallow
copy of a record without modifying any of its contents?  I know
there's the {foo with a=b} construct, but this only works with at
least one assignment; I want to create an unmodified copy of the
record (akin to the {< >} construct for objects).

Thanks!

Re: [Caml-list] Shallow copy of a record

[Jacques Garrigue]

From: Chris King <colanderman@gmail.com>

> Maybe I missed it in the docs, but is there a way to create a shallow
> copy of a record without modifying any of its contents?  I know
> there's the {foo with a=b} construct, but this only works with at
> least one assignment; I want to create an unmodified copy of the
> record (akin to the {< >} construct for objects).

This does not exist, and with good reason: there is no way in the type
system to define a function which works on all records, but only
records. So to do the copy you need to know at least one of the labels
of the record, which in turn gives you its type.

Another approach, which is potentially unsafe, is to use the Obj.dup
function. You can make it a bit safer by ensuring (dynamically) that
you only copy caml blocks or values:

  let copy (v : 'a) : 'a =
    let obj = Obj.repr v in
    if Obj.is_int obj then v
    else if Obj.is_block obj && let tag = Obj.tag obj in
            tag < Obj.lazy_tag || tag = Obj.double_array_tag
    then Obj.obj (Obj.dup obj)
    else invalid_arg "shallow copy"

Jacques Garrigue

Re: [Caml-list] Shallow copy of a record

[Kurt Welgehausen]

It's not clear whether you require a general polymorphic
copy function or you just need to make a copy of a
particular record. In the latter case, something like

  let y = {x with a = x.a}

will do what you want (even though it's not very pretty).

Best regards

Re: [Caml-list] Shallow copy of a record

[Chris King]

On Sat, 22 Jan 2005 08:17:19 -0600, Kurt Welgehausen <kwel@kwel.net> wrote:

> It's not clear whether you require a general polymorphic
> copy function or you just need to make a copy of a
> particular record.

Just the particular record.

> In the latter case, something like
> 
>   let y = {x with a = x.a}
> 
> will do what you want (even though it's not very pretty).

Yeah, that's what I thought of first; I was just hoping there was a
way to do it without the redundant assign (like {x} or something).

Re: [Caml-list] Shallow copy of a record

[Chris King]

On Sat, 22 Jan 2005 15:49:09 +0900 (JST), Jacques Garrigue
<garrigue@math.nagoya-u.ac.jp> wrote:

> This does not exist, and with good reason: there is no way in the type
> system to define a function which works on all records, but only
> records. So to do the copy you need to know at least one of the labels
> of the record, which in turn gives you its type.

Isn't it the foo in {foo with a=b} that determines the type, though?

> Another approach, which is potentially unsafe, is to use the Obj.dup
> function.

That seems like it would do the trick for now - thanks!

Re: [Caml-list] Shallow copy of a record

[Chris King]

On Sat, 22 Jan 2005 10:59:27 +0000, Jon Harrop <jon@jdh30.plus.com> wrote:

> Why do you want to do this?

My specific need is to make a copy of the structure returned by
tcgetattr to be used to reset the terminal state (i.e. one seperate
from the "working" copy used to change the terminal state).  I realize
that I can just call tcgetattr twice (and indeed that's probably
preferable in a functional language) but my C background wants me just
to make a copy :).  (I've actually thought of other reasons not to do
this, but the problem still bugs me!)

Re: [Caml-list] Shallow copy of a record

[Michal Moskal]

On Sat, 22 Jan 2005 11:55:19 -0500, Chris King <colanderman@gmail.com> wrote:
> On Sat, 22 Jan 2005 15:49:09 +0900 (JST), Jacques Garrigue
> <garrigue@math.nagoya-u.ac.jp> wrote:
> 
> > This does not exist, and with good reason: there is no way in the type
> > system to define a function which works on all records, but only
> > records. So to do the copy you need to know at least one of the labels
> > of the record, which in turn gives you its type.
> 
> Isn't it the foo in {foo with a=b} that determines the type, though?

No, consider:

  let copy foo = {foo with a = b}

and now, if there were no assign:

  let copy foo = {foo}

what should be the type of copy?

-- 
: Michal Moskal :: http://nemerle.org/~malekith/ :: GCS !tv h e>+++ b++
: No, I will *not* fix your computer............ :: UL++++$ C++ E--- a?

Re: [Caml-list] Shallow copy of a record

[Chris King]

On Sat, 22 Jan 2005 18:55:40 +0100, Michal Moskal
<michal.moskal@gmail.com> wrote:

> No, consider:
> 
>   let copy foo = {foo with a = b}
> 
> and now, if there were no assign:
> 
>   let copy foo = {foo}
> 
> what should be the type of copy?

Ah okay, I didn't think about that.  But I suppose in cases like that,
the type could be forced, as in "let copy foo = { (foo:bar) }".