[Caml-list] help

[Caml-list] help

[mohammad siddiqui]

Hello,

I am having problem dealing with data structures in OCAML, I am translating 
code from C to OCAML. As shown below one structure contains array of other 
structures as its field. One of the guys here told me that we can’t “inline” 
structures into other structures in OCAML.
I don’t how to deal with this, I already spent like 2 days on it. Should I 
make use of objects and let some member of it array of other objects? I am 
not sure this will do the purpose. I really appreciate if anyone helps me 
out this.

For example if try to change the value model.supvec. (1).words. (1).wnum, it 
changes the values of all elements of the array model.supvec. I tried using 
references, mutable fields, Array.set function for modifying the contents of 
individual element.

The structures in OCAML are:

type word = { mutable wnum:int;	       (* word number *)
	     mutable weight:float };;	       (* word weight *)

type doc =  {
	      mutable docnum:int;              (* Document ID *)
	      mutable   queryid:
	      mutable costfactor:float;
	      mutable  twonorm_sq:float;
	      mutable   words:word array
	     };;

type model =         {
                        mutable   sv_num:int;
                        mutable  at_upper_bound:int;
                        mutable  b:float;
                        mutable  supvec:doc array;
			 mutable alpha: int array;

                       } ;;

The Structures in C are

typedef struct word {
int wnum;
float weight;
} WORD;

typedef struct doc {
  long    docnum;
  long    queryid;
  double  costfactor;
  double  twonorm_sq;
  WORD    *words;
} DOC;



typedef struct model {
  long    sv_num;
  long    at_upper_bound;
  double   b;
  DOC     **supvec;
  Double  *alpha;
} MODEL;

_________________________________________________________________
FREE pop-up blocking with the new MSN Toolbar – get it now! 
http://toolbar.msn.com/go/onm00200415ave/direct/01/

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] help

[Jon Harrop]

I sent you this as an e-mail on Friday:

The following types are a start for your conversion:

type fnum=int and fval=float;

type word = { wnum: fnum ref; weight: fval }
type doc = { docnum: int; queryid: int; costfactor: float; twonorm_sq: float; 
words: word array }
type model = { sv_num: int; at_upper_bound: int; b: float; supvec: doc array; 
alpha: float };

The following line creates an example "model" with its values filled in:

let test_model = { sv_num=0; at_upper_bound=0; b=0.; supvec=Array.make 1 
{ docnum=0; queryid=0; costfactor=0.; twonorm_sq=0.; words=Array.make 1 
{ wnum=ref 0; weight=0. } }; alpha=0. };

I deliberately made "wnum" a reference so that it can be set:

test_model.supvec.(0).words.(0).wnum := 1;

If you want to do a really direct conversion then you probably want to make 
each of the fields in each of the records a reference (as I did with "wnum"). 
However, you may wish to exercise the functional programming side of ocaml 
and write functions which replace data structures, instead of altering them 
in an imperative style. Although this is unintuitive at first, because an 
imperative language would do lots of copying and deleting of data structures, 
the ocaml compiler is very adept at performing only the changes which are 
made to a data structure...

HTH.

Cheers,
Jon.

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] help

[Henri DF]

> I sent you this as an e-mail on Friday:
> 
> The following types are a start for your conversion:
> 
> type fnum=int and fval=float;
> 
> type word = { wnum: fnum ref; weight: fval }
> type doc = { docnum: int; queryid: int; costfactor: float; twonorm_sq: float; 
> words: word array }
> type model = { sv_num: int; at_upper_bound: int; b: float; supvec: doc array; 
> alpha: float };
> 
> The following line creates an example "model" with its values filled in:
> 
> let test_model = { sv_num=0; at_upper_bound=0; b=0.; supvec=Array.make 1 
> { docnum=0; queryid=0; costfactor=0.; twonorm_sq=0.; words=Array.make 1 
> { wnum=ref 0; weight=0. } }; alpha=0. };
> 
> I deliberately made "wnum" a reference so that it can be set:

any reason not to make it a mutable ? 

henri

> test_model.supvec.(0).words.(0).wnum := 1;
> 
> If you want to do a really direct conversion then you probably want to make 
> each of the fields in each of the records a reference (as I did with "wnum"). 
> However, you may wish to exercise the functional programming side of ocaml 
> and write functions which replace data structures, instead of altering them 
> in an imperative style. Although this is unintuitive at first, because an 
> imperative language would do lots of copying and deleting of data structures, 
> the ocaml compiler is very adept at performing only the changes which are 
> made to a data structure...
> 
> HTH.
> 
> Cheers,
> Jon.
> 
> -------------------
> To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
> Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> 

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] help

[mohammad siddiqui]

John,

I tried exactly what you have told. Its is fine when we just have one 
element in an array of words. If we have more than one, changing the value 
of one changes the values of the rest of the elements.

for example , if the test_model is initialiazed like this:

let test_model =
{ sv_num=0; at_upper_bound=0; b=0.; supvec=Array.make 2
{ docnum=0; queryid=0; costfactor=0.; twonorm_sq=0.; words=Array.make 1
{ wnum=ref 0; weight=0. } }; alpha=0. };

now if i change the values , it chnages all th elements of the array 
"supvec" in model

test_model.supvec.(0).words.(0).wnum := 1;

# test_model;;
- : model =
{sv_num = 0; at_upper_bound = 0; b = 0.;
supvec =
  [|{docnum = 0; queryid = 0; costfactor = 0.; twonorm_sq = 0.;
     words = [|{wnum = {contents = 1}; weight = 0.}|]};
    {docnum = 0; queryid = 0; costfactor = 0.; twonorm_sq = 0.;
     words = [|{wnum = {contents = 1}; weight = 0.}|]}|];
alpha = 0.}

Thanks,

Mohammad S Siddiqui.


it changes the values of wnum in all elements of model.supvec.
{ sv_num=0; at_upper_bound=0; b=0.; supvec=[
{ docnum=0; queryid=0; costfactor=0.; twonorm_sq=0.; words=Array.make 1
{ wnum=ref 0; weight=0. } }; alpha=0. };


>From: Jon Harrop <jdh30@cam.ac.uk>
>To: caml-list@inria.fr
>Subject: Re: [Caml-list] help
>Date: Sun, 25 Apr 2004 13:13:37 +0100
>
>
>I sent you this as an e-mail on Friday:
>
>The following types are a start for your conversion:
>
>type fnum=int and fval=float;
>
>type word = { wnum: fnum ref; weight: fval }
>type doc = { docnum: int; queryid: int; costfactor: float; twonorm_sq: 
>float;
>words: word array }
>type model = { sv_num: int; at_upper_bound: int; b: float; supvec: doc 
>array;
>alpha: float };
>
>The following line creates an example "model" with its values filled in:
>
>let test_model = { sv_num=0; at_upper_bound=0; b=0.; supvec=Array.make 1
>{ docnum=0; queryid=0; costfactor=0.; twonorm_sq=0.; words=Array.make 1
>{ wnum=ref 0; weight=0. } }; alpha=0. };
>
>I deliberately made "wnum" a reference so that it can be set:
>
>test_model.supvec.(0).words.(0).wnum := 1;
>
>If you want to do a really direct conversion then you probably want to make
>each of the fields in each of the records a reference (as I did with 
>"wnum").
>However, you may wish to exercise the functional programming side of ocaml
>and write functions which replace data structures, instead of altering them
>in an imperative style. Although this is unintuitive at first, because an
>imperative language would do lots of copying and deleting of data 
>structures,
>the ocaml compiler is very adept at performing only the changes which are
>made to a data structure...
>
>HTH.
>
>Cheers,
>Jon.
>
>-------------------
>To unsubscribe, mail caml-list-request@inria.fr Archives: 
>http://caml.inria.fr
>Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: 
>http://caml.inria.fr/FAQ/
>Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

_________________________________________________________________
Test your ‘Travel Quotient’ and get the chance to win your dream trip! 
http://travel.msn.com

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

[Caml-list] Re: help

[Brian Hurt]

On Sun, 25 Apr 2004, mohammad siddiqui wrote:

> Hello,
> 
> I am having problem dealing with data structures in OCAML, I am translating 
> code from C to OCAML. As shown below one structure contains array of other 
> structures as its field. One of the guys here told me that we can\x12t \x13inline\x14 
> structures into other structures in OCAML.

That was me.

Let's consider the C code:

typedef struct {
    int a;
    int b;
} foo_t;

typedef struct {
    int c;
    foo_t * arr;
} bar_t;

void f(bar_t * ptr) {
    /* Change a foo_t */
    ptr->arr[ptr->c].a += 1;
    ptr->arr[ptr->c].b -= 1;
}

Fairly simply, and basically what you want to do, right?  Now, Ocaml can't 
"inline" (unbox is what I should have said) the foo_t's into the array.  
So you can't do the above code.  What you *can* implement code like (C 
again):

typedef struct {
    int c;
    foo_t ** arr;
} bar_t;

void f(bar_t ptr) {
    ptr->arr[ptr->c]->a += 1;
    ptr->arr[ptr->c]->b -= 1;
}

Notice the extra level of indirection- arr is no longer a pointer to an 
array of foo_t's, but instead a pointer to an array of pointers to 
foo_t's.

Now, the precise translation of this, second version of the code, is this:

type foo_t = { mutable a: int; mutable b: int };;
type bar_t = { c: int; arr: foo_t array };;

let f ptr = 
    ptr.arr.(ptr.c).a <- ptr.arr.(ptr.c).a + 1;
    ptr.arr.(ptr.c).b <- ptr.arr.(ptr.c).b - 1
;;

But we can do better than this, by converting the structures to tuples, 
giving us:

type foo_t = int * int;;
type bar_t = int * (foo_t array);;

let f ((c, arr) : bar_t) =
    arr.(c) <-
        begin
            match (arr.(c)) with
                | (a, b) -> (a+1, b-1)
       end
;;

Here we're taking advantage of the fact that the foo_t's are boxed (not 
inlined), by simply throwing the old tuple out and replacing it with a new 
tuple with the correct new values.

Does this help?

-- 
"Usenet is like a herd of performing elephants with diarrhea -- massive,
difficult to redirect, awe-inspiring, entertaining, and a source of
mind-boggling amounts of excrement when you least expect it."
                                - Gene Spafford 
Brian

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] help

[Jon Harrop]

On Sunday 25 April 2004 5:30 pm, mohammad siddiqui wrote:
> I tried exactly what you have told. Its is fine when we just have one
> element in an array of words. If we have more than one, changing the value
> of one changes the values of the rest of the elements.

Sorry, yes. Now I come to think of it, it's kind of obvious that a mutable 
field in a record would also get copied. My fault. So we don't want to use 
Array.make to generate this structure...

How about using Array.init? This function takes the size of the array and a 
function which generates an element in the array given the index of the 
element within the array, e.g.:

# Array.init 4 (fun i -> i*i);;
- : int array = [|0; 1; 4; 9|]

So we get:

let test_model = { sv_num=0; at_upper_bound=0; b=0.; supvec=Array.init 4 (fun 
i -> { docnum=0; queryid=0; costfactor=0.; twonorm_sq=0.; words=[| { wnum=ref 
(i*i); weight=0. } |] }); alpha=0. };;

  val test_model : model =
  {sv_num = 0; at_upper_bound = 0; b = 0.;
   supvec =
    [|{docnum = 0; queryid = 0; costfactor = 0.; twonorm_sq = 0.;
       words = [|{wnum = {contents = 0}; weight = 0.}|]};
      {docnum = 0; queryid = 0; costfactor = 0.; twonorm_sq = 0.;
       words = [|{wnum = {contents = 1}; weight = 0.}|]};
      {docnum = 0; queryid = 0; costfactor = 0.; twonorm_sq = 0.;
       words = [|{wnum = {contents = 4}; weight = 0.}|]};
      {docnum = 0; queryid = 0; costfactor = 0.; twonorm_sq = 0.;
       words = [|{wnum = {contents = 9}; weight = 0.}|]}|];
   alpha = 0.}

HTH!

Cheers,
Jon.

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] help

[Matt Gushee]

On Sun, Apr 25, 2004 at 01:53:48AM +0000, mohammad siddiqui wrote:
> Hello,
> 
> I am having problem dealing with data structures in OCAML, I am translating 
> code from C to OCAML. As shown below one structure contains array of other 
> structures as its field. One of the guys here told me that we can?t 
> ?inline? structures into other structures in OCAML.
> I don?t how to deal with this, I already spent like 2 days on it. Should I 
> make use of objects and let some member of it array of other objects? I am 
> not sure this will do the purpose. I really appreciate if anyone helps me 
> out this.
> 
> For example if try to change the value model.supvec. (1).words. (1).wnum, 
> it changes the values of all elements of the array model.supvec. I tried 
> using references, mutable fields, Array.set function for modifying the 
> contents of individual element.
> 
> The structures in OCAML are:
> 
> type word = { mutable wnum:int;	       (* word number *)
> 	     mutable weight:float };;	       (* word weight *)
> 
> type doc =  {
> 	      mutable docnum:int;              (* Document ID *)
> 	      mutable   queryid:
> 	      mutable costfactor:float;
> 	      mutable  twonorm_sq:float;
> 	      mutable   words:word array
> 	     };;
> 
> type model =         {
>                        mutable   sv_num:int;
>                        mutable  at_upper_bound:int;
>                        mutable  b:float;
>                        mutable  supvec:doc array;
> 			 mutable alpha: int array;
> 
>                       } ;;
> 
> The Structures in C are
> 
> typedef struct word {
> int wnum;
> float weight;
> } WORD;
> 
> typedef struct doc {
>  long    docnum;
>  long    queryid;
>  double  costfactor;
>  double  twonorm_sq;
>  WORD    *words;
> } DOC;
> 
> 
> 
> typedef struct model {
>  long    sv_num;
>  long    at_upper_bound;
>  double   b;
>  DOC     **supvec;
>  Double  *alpha;
> } MODEL;
> 
> _________________________________________________________________
> FREE pop-up blocking with the new MSN Toolbar ? get it now! 
> http://toolbar.msn.com/go/onm00200415ave/direct/01/
> 
> -------------------
> To unsubscribe, mail caml-list-request@inria.fr Archives: 
> http://caml.inria.fr
> Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: 
> http://caml.inria.fr/FAQ/
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

On Sun, Apr 25, 2004 at 04:30:56PM +0000, mohammad siddiqui wrote:

> I tried exactly what you have told. Its is fine when we just have one 
> element in an array of words. If we have more than one, changing the value 
> of one changes the values of the rest of the elements.

That's not exactly what's happening, though your example makes it look
that way.

> for example , if the test_model is initialiazed like this:
> 
> let test_model =
> { sv_num=0; at_upper_bound=0; b=0.; supvec=Array.make 2
> { docnum=0; queryid=0; costfactor=0.; twonorm_sq=0.; words=Array.make 1
> { wnum=ref 0; weight=0. } }; alpha=0. };

Okay, we can see the cause of the problem here: you are initializing an
array with Array.make, where the initial value includes a ref. The
resulting array will have unique members, *except* that each 'wnum'
field points to a single, shared ref value. That's the problem.


> now if i change the values , it chnages all th elements of the array 
> "supvec" in model
> 
> test_model.supvec.(0).words.(0).wnum := 1;

As you probably understand by now, what this is really doing is changing
all instances of the shared value stored in the 'wnum' field. Evidently 
that was the only field you tried to change, in which case it would
appear that all the elements of 'supvec' were being changed.

Here are three possible solutions:

  1. Make wnum a mutable field instead of a ref.

  2. Replace the array member instead of setting the ref:

       test_model.supvec.(0).words <- newvalue

     (but then I guess there's no point in 'wnum' being either a ref or
      a mutable field)

  3. Don't use Array.make to initialize the array. Instead, use a 
     recursive function that starts with an empty array, and adds new
     elements with Array.append.

-- 
Matt Gushee                 When a nation follows the Way,
Englewood, Colorado, USA    Horses bear manure through
mgushee@havenrock.com           its fields;
http://www.havenrock.com/   When a nation ignores the Way,
                            Horses bear soldiers through
                                its streets.
                                
                            --Lao Tzu (Peter Merel, trans.)

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] help

[Matt Gushee]

On Sun, Apr 25, 2004 at 06:06:23PM +0100, Jon Harrop wrote:
> On Sunday 25 April 2004 5:30 pm, mohammad siddiqui wrote:
> > I tried exactly what you have told. Its is fine when we just have one
> > element in an array of words. If we have more than one, changing the value
> > of one changes the values of the rest of the elements.
> 
> Sorry, yes. Now I come to think of it, it's kind of obvious that a mutable 
> field in a record would also get copied.

Yes, a mutable field *or* a ref (which invalidates one of the solutions
I suggested). Actually, a ref is just a predefined record type that has a 
mutable field and is supported with a special assignment operator.

  # let wnum = { contents = 0 };;
  val wnum : int ref = {contents = 0}

-- 
Matt Gushee                 When a nation follows the Way,
Englewood, Colorado, USA    Horses bear manure through
mgushee@havenrock.com           its fields;
http://www.havenrock.com/   When a nation ignores the Way,
                            Horses bear soldiers through
                                its streets.
                                
                            --Lao Tzu (Peter Merel, trans.)

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] help

[skaller]

On Mon, 2004-04-26 at 02:30, mohammad siddiqui wrote:
> John,
> 
> I tried exactly what you have told. Its is fine when we just have one 
> element in an array of words. If we have more than one, changing the value 
> of one changes the values of the rest of the elements.

I suggest you first try a *purely functional* approach.
The reason is: its easier! It will work exactly as you expect,
no suprises .. I promise <g>.

Get rid of references and mutable fields. Forget efficiency
concerns at this time.

Here is how to make a doc, containing words, where i have
simplified the data structures to demonstrate:

type wrd = string
type doc = { id:int; words: wrd list }

let add_word d w= { d with words = w::d.words }

(* NOTE: functional update syntactic sugar saves 
explicitly copying each field .. unmentioned fields
get copied automatically *)

let empty = { id=99; words=[] }
let d = ref empty
;;

List.iter
(fun w -> d := add_word !d w)
["Hello";"This";"is";"a";"Document";"of";"words"]
;;

List.iter print_endline (!d.words)
;;


Note here i AM using a reference to hold the doc,
but the technique of adding a word to a doc
is purely functional: you get a brand new doc
for every word you add.

EXERCISE: use a fold to get rid of the 'd' variable,
making this little program *purely* functional
(up to the printing part anyhow :D

QUESTION: how fast is the functional technique
compared with using mutable fields or references
to modify a data structure in place??

I don't know the answer . but I'm going to GUESS.
The functional technique is 10% FASTER!

-- 
John Skaller, mailto:skaller@users.sf.net
voice: 061-2-9660-0850, 
snail: PO BOX 401 Glebe NSW 2037 Australia
Checkout the Felix programming language http://felix.sf.net



-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] help

[Jon Harrop]

On Sunday 25 April 2004 1:45 pm, Henri DF wrote:
> > I deliberately made "wnum" a reference so that it can be set:
>
> any reason not to make it a mutable ?

No good reason. ;)

I've just been looking at the assembler output from ocamlopt on this. I think 
that a mutable record field represents one fewer levels of indirection 
compared to a reference record field. On a contrived example, using mutable 
is twice as fast.

Comparing a mutable record field to a reference (not in a record), the latter 
comes out as slightly quicker in my timings. But I think this is 
insignificant as the assembler output seems to be identical.

Cheers,
Jon.

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

Re: [Caml-list] help

[Andrew Lenharth]

On Mon, Apr 26, 2004 at 08:13:42PM +0100, Jon Harrop wrote:
> On Sunday 25 April 2004 1:45 pm, Henri DF wrote:
> > > I deliberately made "wnum" a reference so that it can be set:
> >
> > any reason not to make it a mutable ?
> 
> No good reason. ;)

Take a look at how it is implemented in Pervasives.
>From the manual:

type 'a ref = {
   	mutable contents : 'a;
}

So, yea, unless you like another layer of indirection, just using
mutable should save you a little bit.

Andrew

-- 
"It will work in practice, yes. But will it work in theory?"
--- a french diplomat's comment, recalled by Madeleine  Albright

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners