Not really a bug but...

Not really a bug but...

[Jon Harrop]

Any chance of changing the semantics of string literals so they aren't static?

# let f() = "a";;
val f : unit -> string = <fun>
# let s = f();;
val s : string = "a"
# s.[0] <- 'b';;
- : unit = ()
# f();;
- : string = "b"

-- 
Dr Jon D Harrop, Flying Frog Consultancy Ltd.
OCaml for Scientists
http://www.ffconsultancy.com/products/ocaml_for_scientists/?e

Re: [Caml-list] Not really a bug but...

[Alain Frisch]

Jon Harrop wrote:
> Any chance of changing the semantics of string literals so they aren't static?

A simple Camlp4 syntax extension is enough to turn any literal "foobar"
into String.copy "foobar". (Except it doesn't work because of format
strings.)  If it were done automatically, there would be a penalty for
the common case of immutable strings; to avoid it, you'd need to lift
constant literals out of abstractions, which is not very nice.

-- Alain

Re: [Caml-list] Not really a bug but...

[Christophe TROESTLER]

On Sat, 16 Jun 2007, Alain Frisch <Alain.Frisch@inria.fr> wrote:
> 
> Jon Harrop wrote:
> > Any chance of changing the semantics of string literals so they aren't static?
> 
> A simple Camlp4 syntax extension is enough to turn any literal "foobar"
> into String.copy "foobar". (Except it doesn't work because of format
> strings.)  If it were done automatically, there would be a penalty for
> the common case of immutable strings; to avoid it, you'd need to lift
> constant literals out of abstractions, which is not very nice.

As a side question, any chance of seeing a compiler flag to treat
strings in an immutable way (so as to restrict their mutability to
some modules only)?

ChriS

Re: [Caml-list] Not really a bug but...

[Loup Vaillant]

2007/6/16, Alain Frisch <Alain.Frisch@inria.fr>:
> Jon Harrop wrote:
> > Any chance of changing the semantics of string literals so they aren't static?
> [...]
> If it were done automatically, there would be a penalty for
> the common case of immutable strings; to avoid it, you'd need to lift
> constant literals out of abstractions, which is not very nice.

By the way, why Ocaml didn't take the Java path, i.e. making truly
immutable strings, And provide mutable string buffers as well? Any
chance of seing someone exploring that path? (Some usefull features
would then be fast consing and catenation, and some easier string
sharing).

Loup

Re: [Caml-list] Not really a bug but...

[Richard Jones]

On Sat, Jun 16, 2007 at 12:10:35PM +0200, Loup Vaillant wrote:
> 2007/6/16, Alain Frisch <Alain.Frisch@inria.fr>:
> >Jon Harrop wrote:
> >> Any chance of changing the semantics of string literals so they aren't 
> >static?
> >[...]
> >If it were done automatically, there would be a penalty for
> >the common case of immutable strings; to avoid it, you'd need to lift
> >constant literals out of abstractions, which is not very nice.
> 
> By the way, why Ocaml didn't take the Java path, i.e. making truly
> immutable strings, And provide mutable string buffers as well? Any
> chance of seing someone exploring that path? (Some usefull features
> would then be fast consing and catenation, and some easier string
> sharing).

Mutable strings are useful!  - I can use them as bitfields, general
data areas (for ioctl, mlock), etc  See:
http://et.redhat.com/~rjones/hvcalls/

Rich.

-- 
Richard Jones
Red Hat

Re: [Caml-list] Not really a bug but...

[skaller]

On Sat, 2007-06-16 at 12:10 +0200, Loup Vaillant wrote:
> 2007/6/16, Alain Frisch <Alain.Frisch@inria.fr>:
> > Jon Harrop wrote:
> > > Any chance of changing the semantics of string literals so they aren't static?
> > [...]
> > If it were done automatically, there would be a penalty for
> > the common case of immutable strings; to avoid it, you'd need to lift
> > constant literals out of abstractions, which is not very nice.
> 
> By the way, why Ocaml didn't take the Java path, i.e. making truly
> immutable strings, And provide mutable string buffers as well?

It didn't because it didn't .. :)

Strings are mutable because they can be, without impacting
the run time/garbage collector: Ocam's razor says use
one type when it will do the work of two.

It's unfortunate that in the abstract, Ocaml string are
entirely the worst possible multi-function data type as a result,
since they neither offer the advantages of immutability nor the
advantages of being variable length: mutable fixed length strings
are almost useless .. but not entirely: they're still useful as I/O
buffers.

-- 
John Skaller <skaller at users dot sf dot net>
Felix, successor to C++: http://felix.sf.net

Re: [Caml-list] Not really a bug but...

[Jon Harrop]

On Saturday 16 June 2007 18:01:21 skaller wrote:
> mutable fixed length strings
> are almost useless .. but not entirely: they're still useful as I/O
> buffers.

Like Richard, I use fixed-length mutable strings quite a lot. This complicated 
the translation of some programs to F# which only provides immutable (but 
unicode) strings.

-- 
Dr Jon D Harrop, Flying Frog Consultancy Ltd.
OCaml for Scientists
http://www.ffconsultancy.com/products/ocaml_for_scientists/?e