Hashtbl.remove legal within Hashtbl.iter for the same hash table?

Hashtbl.remove legal within Hashtbl.iter for the same hash table?

[David MENTRE]

Hello,

Probably a newbie question but anyway: is it allowed to do a
Hashtbl.remove while doing a Hashtbl.iter on the same hash table?

More precisely, at one point while doing a "Hashtbl.iter f h" my
function "f" is called with something like "f k v". Can I do a
"Hashtbl.remove h k" within the body of "f"?

Sincerely yours,
david
-- 
GPG/PGP key: A3AD7A2A David MENTRE <dmentre@linux-france.org>
 5996 CC46 4612 9CA4 3562  D7AC 6C67 9E96 A3AD 7A2A

Re: [Caml-list] Hashtbl.remove legal within Hashtbl.iter for the same hash table?

[Florent Monnier]

> Hello,
Hello David,

> Probably a newbie question but anyway: is it allowed to do a
> Hashtbl.remove while doing a Hashtbl.iter on the same hash table?
I don't know if it is legal, but at least it works:

# let h = Hashtbl.create 8 ;;

# for i = 0 to pred 8 do
    Hashtbl.add h i (char_of_int((int_of_char 'A') + i))
  done ;;

# Hashtbl.iter (fun i v ->
      if i = 3 then Hashtbl.remove h 5;
      Printf.printf " %d  %c\n" i v) h ;;
 0  A
 1  B
 2  C
 3  D
 4  E
 6  G
 7  H

But perhaps is it implementation dependant, in which case another 
implementation could react in a different way...

The ExtLib reacts in the same way than the standard one.

> More precisely, at one point while doing a "Hashtbl.iter f h" my
> function "f" is called with something like "f k v". Can I do a
> "Hashtbl.remove h k" within the body of "f"?

it seems to work too when we remove the current itered key
(with both implementations)

I don't know if this module was written with in mind to allow this behavior, 
but for what I understand from the manual : "in-place modification" should 
mean that this structure is purely imperative and that this behavior is 
"legal/allowed".
______

Perhaps a more sure method could be to get an enum from the hash table, and 
then iter on this enum (with the ExtLib), see below.
With this method you are sure that *all* the keys will be itered, including 
the hidden contents, which is different than the previous example.
Then you don't need to worry about a perhaps implementation dependant 
behavior.

ocaml -I +/site-lib/extlib extLib.cma
# open ExtLib ;;
# open ExtHashtbl ;;
# let h = Hashtbl.create 8 ;;
# for i = 0 to pred 8 do
    Hashtbl.add h i (char_of_int((int_of_char 'A') + i))
  done ;;

(* this one would not appear with the sdt Hashtbl.iter ! *)
# Hashtbl.add h 6 'Z' ;;

# let iter_all_hashtbl f h =
    let keys = Hashtbl.keys h
    and vals = Hashtbl.values h in
    Enum.iter2 f keys vals
  ;;

# let f i v =
    if i = 3 then Hashtbl.remove h 3;
    Printf.printf " %d  %c\n" i v
  ;;

# iter_all_hashtbl f h ;;
 0  A
 1  B
 2  C
 3  D
 4  E
 5  F
 6  Z
 6  G
 7  H

# iter_all_hashtbl f h ;;
 0  A
 1  B
 2  C
 4  E
 5  F
 6  Z
 6  G
 7  H

-- 
Florent

Re: [Caml-list] Hashtbl.remove legal within Hashtbl.iter for the same hash table?

[Till Varoquaux]

Hatables are arrays of associative lists. When you are iterating over
them removing any element you have already visited should be ok.
Removing elements you haven't visited yet could cause you to encounter
them anyhow.

Adding elements might trigger resizing and then things could get
sketchy (has tables are never down sized).

Till
On Sun, May 11, 2008 at 3:42 PM, Florent Monnier
<fmonnier@linux-nantes.fr.eu.org> wrote:
>> Hello,
> Hello David,
>
>> Probably a newbie question but anyway: is it allowed to do a
>> Hashtbl.remove while doing a Hashtbl.iter on the same hash table?
> I don't know if it is legal, but at least it works:
>
> # let h = Hashtbl.create 8 ;;
>
> # for i = 0 to pred 8 do
>    Hashtbl.add h i (char_of_int((int_of_char 'A') + i))
>  done ;;
>
> # Hashtbl.iter (fun i v ->
>      if i = 3 then Hashtbl.remove h 5;
>      Printf.printf " %d  %c\n" i v) h ;;
>  0  A
>  1  B
>  2  C
>  3  D
>  4  E
>  6  G
>  7  H
>
> But perhaps is it implementation dependant, in which case another
> implementation could react in a different way...
>
> The ExtLib reacts in the same way than the standard one.
>
>> More precisely, at one point while doing a "Hashtbl.iter f h" my
>> function "f" is called with something like "f k v". Can I do a
>> "Hashtbl.remove h k" within the body of "f"?
>
> it seems to work too when we remove the current itered key
> (with both implementations)
>
> I don't know if this module was written with in mind to allow this behavior,
> but for what I understand from the manual : "in-place modification" should
> mean that this structure is purely imperative and that this behavior is
> "legal/allowed".
> ______
>
> Perhaps a more sure method could be to get an enum from the hash table, and
> then iter on this enum (with the ExtLib), see below.
> With this method you are sure that *all* the keys will be itered, including
> the hidden contents, which is different than the previous example.
> Then you don't need to worry about a perhaps implementation dependant
> behavior.
>
> ocaml -I +/site-lib/extlib extLib.cma
> # open ExtLib ;;
> # open ExtHashtbl ;;
> # let h = Hashtbl.create 8 ;;
> # for i = 0 to pred 8 do
>    Hashtbl.add h i (char_of_int((int_of_char 'A') + i))
>  done ;;
>
> (* this one would not appear with the sdt Hashtbl.iter ! *)
> # Hashtbl.add h 6 'Z' ;;
>
> # let iter_all_hashtbl f h =
>    let keys = Hashtbl.keys h
>    and vals = Hashtbl.values h in
>    Enum.iter2 f keys vals
>  ;;
>
> # let f i v =
>    if i = 3 then Hashtbl.remove h 3;
>    Printf.printf " %d  %c\n" i v
>  ;;
>
> # iter_all_hashtbl f h ;;
>  0  A
>  1  B
>  2  C
>  3  D
>  4  E
>  5  F
>  6  Z
>  6  G
>  7  H
>
> # iter_all_hashtbl f h ;;
>  0  A
>  1  B
>  2  C
>  4  E
>  5  F
>  6  Z
>  6  G
>  7  H
>
> --
> Florent
>
> _______________________________________________
> Caml-list mailing list. Subscription management:
> http://yquem.inria.fr/cgi-bin/mailman/listinfo/caml-list
> Archives: http://caml.inria.fr
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs
>



-- 
http://till-varoquaux.blogspot.com/

Re: [Caml-list] Hashtbl.remove legal within Hashtbl.iter for the same hash table?

[Florent Monnier]

> Hatables are arrays of associative lists. When you are iterating over
> them removing any element you have already visited should be ok.
> Removing elements you haven't visited yet could cause you to encounter
> them anyhow.

which means that it is dependent to the order in which the content is itered,
while the manual says :
"The order in which the bindings are passed to f is unspecified."

Re: [Caml-list] Hashtbl.remove legal within Hashtbl.iter for the same hash table?

[Till Varoquaux]

Indeed. The answer you got was, however, based on the actual
implementation not on the documentation. This means that, at some
point, this might evolve and not be valid anymore. If you want to be
on the safe side you should do a fold instead of an iter and collect
all of the items to remove and then remove them in a second pass. The
performance hit shouldn't be as bad as you could expect (ie: I
wouldn't bother unless performance really is critical).
I see it as very unlikely that ocaml's implementation of hashtbl would
evolve in such a way that it would break any code removing previously
visited items during a traversal. Your call.

Till

On 5/11/08, Florent Monnier <fmonnier@linux-nantes.fr.eu.org> wrote:
>> Hatables are arrays of associative lists. When you are iterating over
>> them removing any element you have already visited should be ok.
>> Removing elements you haven't visited yet could cause you to encounter
>> them anyhow.
>
> which means that it is dependent to the order in which the content is
> itered,
> while the manual says :
> "The order in which the bindings are passed to f is unspecified."
>
> _______________________________________________
> Caml-list mailing list. Subscription management:
> http://yquem.inria.fr/cgi-bin/mailman/listinfo/caml-list
> Archives: http://caml.inria.fr
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs
>


-- 
http://till-varoquaux.blogspot.com/

Re: [Caml-list] Hashtbl.remove legal within Hashtbl.iter for the same hash table?

[Martin Jambon]

On Sun, 11 May 2008, Till Varoquaux wrote:

> Indeed. The answer you got was, however, based on the actual
> implementation not on the documentation. This means that, at some
> point, this might evolve and not be valid anymore. If you want to be
> on the safe side you should do a fold instead of an iter and collect
> all of the items to remove and then remove them in a second pass. The
> performance hit shouldn't be as bad as you could expect (ie: I
> wouldn't bother unless performance really is critical).
> I see it as very unlikely that ocaml's implementation of hashtbl would
> evolve in such a way that it would break any code removing previously
> visited items during a traversal. Your call.

If you (Florent) want to rely on the standard Hashtbl module, I would 
strongly advise against making assumptions such as "the implementation is 
unlikely to change". Think of the situation 5 years later, when the 
implementation actually changes, you've left the company 3 years before, 
and the maintainer has to figure why the program returns messed up 
results... 
I bet you wouldn't like to be this maintainer.

It's like drunk driving or forgetting the condom: just don't do it...



Martin


> On 5/11/08, Florent Monnier <fmonnier@linux-nantes.fr.eu.org> wrote:
>>> Hatables are arrays of associative lists. When you are iterating over
>>> them removing any element you have already visited should be ok.
>>> Removing elements you haven't visited yet could cause you to encounter
>>> them anyhow.
>>
>> which means that it is dependent to the order in which the content is
>> itered,
>> while the manual says :
>> "The order in which the bindings are passed to f is unspecified."
>>
>> _______________________________________________
>> Caml-list mailing list. Subscription management:
>> http://yquem.inria.fr/cgi-bin/mailman/listinfo/caml-list
>> Archives: http://caml.inria.fr
>> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
>> Bug reports: http://caml.inria.fr/bin/caml-bugs
>>
>
>
> -- 
> http://till-varoquaux.blogspot.com/
>
> _______________________________________________
> Caml-list mailing list. Subscription management:
> http://yquem.inria.fr/cgi-bin/mailman/listinfo/caml-list
> Archives: http://caml.inria.fr
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs
>

--
http://wink.com/profile/mjambon
http://mjambon.com

Re: [Caml-list] Hashtbl.remove legal within Hashtbl.iter for the same hash table?

[David MENTRE]

Hello Till,

"Till Varoquaux" <till.varoquaux@gmail.com> writes:

> Indeed. The answer you got was, however, based on the actual
> implementation not on the documentation. This means that, at some
> point, this might evolve and not be valid anymore. If you want to be
> on the safe side you should do a fold instead of an iter and collect
> all of the items to remove and then remove them in a second pass.

I'll do that.

Thanks,
d.
-- 
GPG/PGP key: A3AD7A2A David MENTRE <dmentre@linux-france.org>
 5996 CC46 4612 9CA4 3562  D7AC 6C67 9E96 A3AD 7A2A

Re: [Caml-list] Hashtbl.remove legal within Hashtbl.iter for the same hash table?

[Florent Monnier]

> (* this one would not appear with the sdt Hashtbl.iter ! *)
wrong, it appears too

I've made a mistake in the toplevel :)

Re: [Caml-list] Hashtbl.remove legal within Hashtbl.iter for the same hash table?

[Julien Peeters]

[-- Attachment #1: Type: text/plain, Size: 1332 bytes --]

Hi David,

Because the arguments passed to functions could be considered as references
(same memory space), IMHO, if you respect the scope priority it should work.

For instance:

let () =
    let h = Hashtbl.create 7 in
    let f k v = Hashtbl.remove h k in
        Hashtbl.iter f h

Julien.

2008/5/11 David MENTRE <dmentre@linux-france.org>:

> Hello,
>
> Probably a newbie question but anyway: is it allowed to do a
> Hashtbl.remove while doing a Hashtbl.iter on the same hash table?
>
> More precisely, at one point while doing a "Hashtbl.iter f h" my
> function "f" is called with something like "f k v". Can I do a
> "Hashtbl.remove h k" within the body of "f"?
>
> Sincerely yours,
> david
> --
> GPG/PGP key: A3AD7A2A David MENTRE <dmentre@linux-france.org>
>  5996 CC46 4612 9CA4 3562  D7AC 6C67 9E96 A3AD 7A2A
>
> _______________________________________________
> Caml-list mailing list. Subscription management:
> http://yquem.inria.fr/cgi-bin/mailman/listinfo/caml-list
> Archives: http://caml.inria.fr
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs
>



-- 
Julien Peeters
Student in engineering
Institut de Formation des Ingénieurs de Paris-Sud
Orsay, France
Website: http://www.dedilabs.com/~jpeeters

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