ocaml-lib-devel — Developer's mailing list

Re: [Ocaml-lib-devel] regeps

[Richard J. <ri...@an...>]

Not really sure I understand it from a user's point of view.

For me what really matters is that I could write something like:

  if string =3D~ /(a+)(b*)/ then ...

The regular expression (Perl-compatible, please!) ought to be compiled
and checked at compile time, and the compiler should handle caching
compiled expressions containing variables.

Actually, the example syntax above isn't really great.  I'd need some
way to bind captures.

The Pcre module has good and bad features.  Good: It uses Perl-
compatible regexps with fairly minimal fuss.  Bad: All checking/
compilation is done at runtime; the syntax for accessing captures is
very clunky.

The camlp4 regexp module [1] looks much better, but the regexps aren't
fully Perl compatible, and I'm not sure if compilation happens at
compile time or runtime and how it handles caching.

How will your proposal actually work (as in, how will it actually be
used: can you give some example code)?

Rich.

[1] http://web.yl.is.s.u-tokyo.ac.jp/~oiwa/pub/caml/regexp-pp-0.9.3/README.=
match-regexp

--=20
Richard Jones. http://www.annexia.org/ http://www.j-london.com/
Merjis Ltd. http://www.merjis.com/ - improving website return on investment
PTHRLIB is a library for writing small, efficient and fast servers in C.
HTTP, CGI, DBI, lightweight threads: http://www.annexia.org/freeware/pthrli=
b/

Re: [Ocaml-lib-devel] regeps (repost)

[Richard J. <ri...@an...>]

(Repost: original was apparently tagged as sp*m).

Not really sure I understand it from a user's point of view.

For me what really matters is that I could write something like:

  if string =3D~ /(a+)(b*)/ then ...

The regular expression (Perl-compatible, please!) ought to be compiled
and checked at compile time, and the compiler should handle caching
compiled expressions containing variables.

Actually, the example syntax above isn't really great.  I'd need some
way to bind captures.

The Pcre module has good and bad features.  Good: It uses Perl-
compatible regexps with fairly minimal fuss.  Bad: All checking/
compilation is done at runtime; the syntax for accessing captures is
very clunky.

The camlp4 regexp module [1] looks much better, but the regexps aren't
fully Perl compatible, and I'm not sure if compilation happens at
compile time or runtime and how it handles caching.

How will your proposal actually work (as in, how will it actually be
used: can you give some example code)?

Rich.

[1] http://web.yl.is.s.u-tokyo.ac.jp/~oiwa/pub/caml/regexp-pp-0.9.3/README.=
match-regexp

--=20
Richard Jones. http://www.annexia.org/ http://www.j-london.com/
Merjis Ltd. http://www.merjis.com/ - improving website return on investment
MOD_CAML lets you run type-safe Objective CAML programs inside the Apache
webserver. http://www.merjis.com/developers/mod_caml/

Re: [Ocaml-lib-devel] Re: regeps

[skaller <sk...@us...>]

On Sat, 2004-09-11 at 01:02, Jesse Guardiani wrote:
> skaller wrote:
> 
> > I'm playing with my regexp package and I'd like
> > to discuss architecture.
> 
> [...]
> 
> > The goal is a simple, powerful, in Caml regular matching
> > facility, written entirely in Ocaml, and obsoleting Str, PCRE,
> > and Ocamllex.
> > 

> 3.) Do you have some code we can play with? I'd be able to comment
>     on the interface if I could compile something...

Attached.

dfa.ml and dfa.mli contain the basic engine.

test_dfa.ml is a sample test routine,
unfinished.

The engine isn't necessarily the most efficient code,
but it seems to work. No attempt is made to minimise
the number of states.

This engine only accepts one type of character: int.
There are no other intrinsic limitations, however
some of the data structures might get big.

For large character sets, like Unicode, the engine
as it stands will probably use all your memory
and take too long to be practical.

To fix this problem, we need character classes:
typically, many characters are equivalent. 
This is done by a function

	classify : int -> int

which maps a character to a class index which
is then fed to the driver. The regexps also need
to be built in terms of class indices. One way
may be to add a regexp term 

	`REGEXP_in_range (a,b)

however, this does require some fiddling.
EG if you have a range A-Z and a keyword 'IS'
then actually you have 5 classes.

It may be useful to change the character type
from int to  int * 'b where 'b is the actual
character, and int is its class. The automaton
only uses the class. However if you're processing
a stream, and want to extract groups, you want
an array of 'b, not an array of class indicies.

Similarly condider a regexp for a decimal integer:

	{`Start} (9 {`Digit}) *

where `Start sets a variable to 0, and `Digit multiples
by 10 and adds the digit minus '0', thus computing
the numeric value as an attribute for a some token.
Clearly a class of 0-9 won't do even if 0-9 are equivalent,
we need the actual character for the computation.

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

[Ocaml-lib-devel] Re: regeps

[Jesse G. <je...@wi...>]

skaller wrote:

> I'm playing with my regexp package and I'd like
> to discuss architecture.

[...]

> The goal is a simple, powerful, in Caml regular matching
> facility, written entirely in Ocaml, and obsoleting Str, PCRE,
> and Ocamllex.
> 
> Any comments appreciated.

1.) I DEFINATELY think O'Caml needs a 100% pure O'Caml regex module.
    I've wished for this a number of times myself, and if we got it
    then we could remove the dependency on PCRE for things like Dbi's
    postgresql driver.

2.) I know next to nothing about regex implementations. I am simply
    a user.

3.) Do you have some code we can play with? I'd be able to comment
    on the interface if I could compile something...


-- 
Jesse Guardiani, Systems Administrator
WingNET Internet Services,
P.O. Box 2605 // Cleveland, TN 37320-2605
423-559-LINK (v)  423-559-5145 (f)
http://www.wingnet.net

[Ocaml-lib-devel] regeps

[skaller <sk...@us...>]

I'm playing with my regexp package and I'd like
to discuss architecture.

(1) ENGINE

The engine accepts a regular expression of the following type:

type 'a regexp_t =
  [
  | `REGEXP_char of int
  | `REGEXP_seq of 'a regexp_t * 'a regexp_t (** concatenation *)
  | `REGEXP_alt of 'a regexp_t * 'a regexp_t (** alternation *)
  | `REGEXP_aster of 'a regexp_t (** Kleene closure *)
  | `REGEXP_epsilon (** epsilon: null string *)
  | `REGEXP_code of 'a (** associated code *)
  ]

You will note a 'character' is a fixed type, namely int,
and that there is a user determined 'code' constructor
whose purpose will be explained below.

It returns

(a) The alphabet, being a Set of int denoting
each character which is recognized.

(b) A count n of the number of states, which are
numbered 0 to n-1.

(c) A partial mapping from states to codes,
represented by a hashtable, explained below.

(d) A partial mapping from character * state
pairs to states, represented by a hashtable,
called the transition matrix.

Here is the interface:

val compile_regexp : 
  'a regexp_t ->
  CharSet.t * (* alphabet *)
  int *  (* state count *)
  (int, 'a) Hashtbl.t *  (* terminal codes *)
  (CharSet.elt * int, int) Hashtbl.t (* transition matrix *)

If the code mapping fails, there is no code for that state.
If the matrix lookup fails, the transition is to 
the 'error' state.

Now for the fun part :)

Most regexp systems include some weird
notion of 'groups'.

The engine above is not only much simpler,
it is also better.

Basically, a 'code' is a marked epsilon transition.
By tacking codes after a regular expression,
the terminal state for that expression is marked
with an 'accepting' code.

Whilst driving the matrix with a matcher, lexer,
or other driver loop, you can check whether you have
encountered a marked state, and if so, which one.

Consider:

\(1*\)-\(2*\)

We will convert the groups to plain marks:

{`m 1}1*{`m 2}-{`m 3}2{`m 4}

and create an int option array of length 4 initialised
to None. Whenever we get a marked state m1-m4 we
store the string index into the array in the right
slot.

When we're finished .. hey presto, we have our groups.
To implement this we could also use the code type:

type code_t = [`Left of int | `Right of int]

Another use of marks is for classification:

1*{`Ones}|2*{`Twos}

When we get a match, we also know which alternative
matched. Of course you can do this with groups too.

The big difference between marks and groups is that
the marks are an arbitrary (comparable) user defined type, and,
what you do when you encounter one is up to you.

Note groups and marks are equivalent -- regexp engines implement
groups with marks, and marks can always be replaced by groups.
However groups use a positional notation (counting brackets),
whereas marks actually return an arbitrary code.

Using a positional group encoding like Cameleon:

	string option array option

isn't nearly as good as marks: in a production lexer
with 50 tokens you'd have to search for Some string,
and then translate the array index into a token (if you can
remember that position 45 is the keyword 'match' for
example .. :)

In particular this regexp engine *automagically* supports
'in Caml' tokenisers which are are actually more expressive
than Ocamllex. In my test engine I have written this:

let digit = regexp_of_charset "0123456789"
let lower = regexp_of_charset "abcdefghijklmnopqrstuvwxyz"
let upper = regexp_of_charset "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
let letter = alt upper lower
let underscore = regexp_of_char '_'
let idstart = alt letter underscore
let idrest = alt idstart digit
let space = regexp_of_char ' '
let white = many space
let ident = seq idstart (aster idrest)
let num = many digit
let code x = `REGEXP_code x

let tk_ident = seq ident (code `Ident)
let tk_num = seq num (code `Num)
let tk_white = seq white (code `White)

let re = alt tk_ident (alt tk_num tk_white)
let res = many re

I'm using obvious named functions for combinators 

  seq a b = `REGEXP_seq (a,b)

etc, you could also define infix operators for alt and seq.
Of course you can also do: fold_left seq epsilon [a;b;c;d] etc.
Note this isn't Ocamllex code -- its pure Ocaml!

Although the engine creates a Hashtbl, for which lookup
isn't efficient, you can always transform it into
an array and pay the extra space cost.

Also, if you want to avoid the cost of compiling a regexp,
you can write a pure ocaml program which does the conversion,
and either prints out the matrix as Ocaml code, or simply
Marshall it to disk.

(2) DRIVERS

To actually use the engine, we need some drivers such as
'match' and 'lex', and some convenient mark processing
utilities.

(3) PARSERS

We actually use Ocaml as the parser. We can 'sugar' the syntax
with any kind of function. For example we can create a 
type of regexp with groups instead of marks,
and just translate the groups into appropriate marks.

In order to provide Posix and Emacs looking regexps, we actually
write a parser function that produces a regexp.

There is a lot more work to produce usable sugar for
this engine -- eg actually implement groups. I don't yet
have everything because I'm not yet sure exactly what
is needed. 

The goal is a simple, powerful, in Caml regular matching
facility, written entirely in Ocaml, and obsoleting Str, PCRE,
and Ocamllex.

Any comments appreciated.

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

Re: [Ocaml-lib-devel] Re: TAI64

[Richard J. <ri...@an...>]

On Thu, Sep 09, 2004 at 01:35:17PM -0400, Jesse Guardiani wrote:
> Richard Jones wrote:
>=20
> > On the subject of dates and times again, someone with a clue *did*
> > scratch their itch and implement an interesting time module:
>=20
> Well, I think that may depend on whether or not Dan Bernstein has a clue.
>=20
> I use TAI64 with multilog, but I certainly don't know if it's a sane time
> format or not. I generally like Dan's work, but I've never heard much abo=
ut
> TAI64 outside Qmail...

Very true ...

Rich.

--=20
Richard Jones. http://www.annexia.org/ http://www.j-london.com/
Merjis Ltd. http://www.merjis.com/ - improving website return on investment
C2LIB is a library of basic Perl/STL-like types for C. Vectors, hashes,
trees, string funcs, pool allocator: http://www.annexia.org/freeware/c2lib/

[Ocaml-lib-devel] Re: TAI64

[Jesse G. <je...@wi...>]

Richard Jones wrote:

> On the subject of dates and times again, someone with a clue *did*
> scratch their itch and implement an interesting time module:

Well, I think that may depend on whether or not Dan Bernstein has a clue.

I use TAI64 with multilog, but I certainly don't know if it's a sane time
format or not. I generally like Dan's work, but I've never heard much about
TAI64 outside Qmail...

-- 
Jesse Guardiani, Systems Administrator
WingNET Internet Services,
P.O. Box 2605 // Cleveland, TN 37320-2605
423-559-LINK (v)  423-559-5145 (f)
http://www.wingnet.net

[Ocaml-lib-devel] TAI64

[Richard J. <ri...@an...>]

On the subject of dates and times again, someone with a clue *did*
scratch their itch and implement an interesting time module:

http://cvs.sourceforge.net/viewcvs.py/ocnae/src/cf/cf_tai64.mli?rev=3D1.2&v=
iew=3Dmarkup

http://cvs.sourceforge.net/viewcvs.py/ocnae/src/cf/cf_stdtime.mli?rev=3D1.2=
&view=3Dmarkup

I didn't check very closely whether the license was compatible.

Rich.

--=20
Richard Jones. http://www.annexia.org/ http://www.j-london.com/
Merjis Ltd. http://www.merjis.com/ - improving website return on investment
C2LIB is a library of basic Perl/STL-like types for C. Vectors, hashes,
trees, string funcs, pool allocator: http://www.annexia.org/freeware/c2lib/

Re: [Ocaml-lib-devel] Re: Dllist ? IndexTree ?

[Jesse D. G. <je...@wi...>]

Nicolas Cannasse wrote:

>>Nicolas Cannasse wrote:
>>
>>    
>>
>>>What about the two newest data structures dllist and index tree ?
>>>Is there final implementation proposal available somewhere to include in
>>>ExtLib ?
>>>      
>>>
>>Gimme a day or two to finalize Dllist. I had planned to work on it this
>>past weekend (Monday was a Holiday and all), but instead I took a little
>>road trip with my wife and son. I'll try to work on it tonight and maybe
>>have something final by tommorrow.
>>    
>>
>
>It's ok you don't need to hurry that much :)
>  
>

New version available here:
    http://www.wingnet.net/~jesse/ocaml/dllist/index.html

Changes:
- merge -> splice
- jump -> skip
- enums reimplemented
- added rev_enum function

My sourceforge ID is 'trevarthan'. I'll be the maintainer.

Thanks!


--
Jesse Guardiani, Systems Administrator
WingNET Internet Services,
P.O. Box 2605 // Cleveland, TN 37320-2605
423-559-LINK (v)  423-559-5145 (f)
http://www.wingnet.net

[Ocaml-lib-devel] Indexed tree module

[Brian H. <bh...@sp...>]

I'd like to submit this module for  inclusion into extlib.

This is module formerly known as Prince, er, Funarr.  On reflection, I
decided I liked 'index tree' as a name better too.

General notes:

- Changed name

- Everything from Funarr still supported

- I stuck with a weight balanced tree.  Chris Okasaki's Random Access 
Lists are interesting, but they're geared towards list replacement, not 
array replacement.  For example, I can't figure out how to delete elements 
in the middle of a random access list- I can with a weight balanced tree.

- added append, to append two indexed trees in O(log N) time.  The 
returned tree is balanced.

- added splice, return a subset of an indexed tree.  This takes either
O(log N)  where N is the length of the source tree, or O((log M)^2) where
M is the length of the returned tree.  I can't think of a better way of
doing this- suggestions welcomed.

- Added list, array, and enum conversions.

- Added rev function to reverse an indexed tree

- I realized that I could do a binary search on the indexed tree in the 
same amount of time it takes me to get a single element.  So I added two 
binary search functions- search, which returns the found element, and
find_index, which returns the index of the found element.  Note that with 
these functions it now becomes possible to implement a threaded map 
structure, by combining an index tree with a map.

- Haven't tested it yet.

-- 
"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

Re: [Ocaml-lib-devel] Circular Buffer

[Bardur A. <oca...@sc...>]

On Wed, Sep 08, 2004 at 04:47:34PM +0200, Bardur Arantsson wrote:

[sorry about replying to myself. I know, it's bad form...]

I just realized that the interface doc states that operations
are "at most" O(something). This is of course redundant,
at I'll change it to "This operation is O(...)".

-- 
Bardur Arantsson
<ba...@im...>
<ba...@sc...>

If it looks like a duck, quacks like a duck, and walks like a
duck it's probably just a tool of the conspiracy.
                                   Kallahan | http://slashdot.org

[Ocaml-lib-devel] Circular Buffer

[Bardur A. <oca...@sc...>]

Hi all,

I've written an efficient circular (character) buffer
module for a network-related application I'm developing,
and I was wondering if there might be interest in
including it in ExtLib?

The highlights:

- All but two operations are O(1), the only exceptions are
  blit and grow (for obvious reasons).

- No superfluous string copying.

- No string allocation except when creating or growing.

- Interface is very convenient for impementing buffering
  with Unix.read, Unix.write/single_write, or
  IO.input/output and similar functions. (It basically
  uses the same pattern that the ocamlnet/netstring
  Netbuffer module uses for add_inplace which is about as
  convenient as it gets).


I've attached the interface file which contains everything
*I* need, I'm open to suggestions for additions, however.

Any interest at all...?

Cheers,

-- 
Bardur Arantsson
<ba...@im...>
<ba...@sc...>

All warfare is based on deception.
                                         Sun Tzu | The Art of War

Re: [Ocaml-lib-devel] Dllist ? IndexTree ?

[Richard J. <ri...@an...>]

On Tue, Sep 07, 2004 at 10:48:43AM +0200, Nicolas Cannasse wrote:
> What about the two newest data structures dllist and index tree ?
> Is there final implementation proposal available somewhere to include in
> ExtLib ?

I wasn't intending it for inclusion, but I have a very complete CSV
(comma-separated-values) library which could go in.

http://www.merjis.com/developers/csv/

Any licensing changes necessary can be incorporated.

Rich.

--=20
Richard Jones. http://www.annexia.org/ http://www.j-london.com/
Merjis Ltd. http://www.merjis.com/ - improving website return on investment
Learning Objective CAML for C, C++, Perl and Java programmers:
http://www.merjis.com/richj/computers/ocaml/tutorial/

Re: [Ocaml-lib-devel] split & nsplit with support for quote characters?

[Nicolas C. <war...@fr...>]

> I have need for split & nsplit functions that honor (in
> other words, ignore the text between) double quote characters.
> Is anyone interested in something like this for extlib?

I think this is too much particular to be used for ExtLib, and also not so
much efficient. A Regexp/Lexer would fit better here.

Regards,
Nicolas Cannasse

[Ocaml-lib-devel] split & nsplit with support for quote characters?

[Jesse G. <je...@wi...>]

I have need for split & nsplit functions that honor (in
other words, ignore the text between) double quote characters.
Is anyone interested in something like this for extlib?

For example, I have a text file that uses the following format:

key:field1:name0=value0,name1=value1

This file is compiled into a CDB hash using Dustin Sallings
CDB module, then I access it using the key. If a match is found,
the value returned is everything after the first colon
character up to a newline. It is then my job to parse the
value further into field1 and a list of name=value pairs.

This can all be accomplished with split and nsplit, unless
you need to embed a comma character in one of your values,
like this:

mykey:myfield:FOO="BAR0,BAR1,BAR2",SHELL="/bin/sh"

With my current parsing scheme (based on split and nsplit)
this would be broken down as follows:

field1=myfield
FOO="BAR0
BAR1=
BAR2=
SHELL="/bin/sh"

So I need to implement a split and nsplit that honor quotation
characters. Here's what I've got right now (it's ugly, but it
works):

exception Invalid_quote of char;;

(** [qfind ~quote:['"'] s x] returns the starting index of the string [x]
    within the string [s], ignoring characters enclosed in any of the characters
    passed to the [quote] argument.
    
    @raise Invalid_string if [x] is not a substring of [s].
    @raise Invalid_quote if a character in the [quote] list appears in [s], but
    is not terminated properly.
    
    @param quote List of characters inside which we don't look for [sep];
    defaults to '"'.
*)
let qfind ?(quote=['"']) str sub =
        let sublen = ExtString.String.length sub in
        if sublen = 0 then
                0
        else
                let found = ref 0 in
                let len = ExtString.String.length str in
                try
                        let i = ref 0 in
                        while !i <= len - sublen do
                                begin
                                let j = ref 0 in

                                i := if List.mem (String.unsafe_get str (!i)) quote
                                     then begin
                                                incr j;
                                                let q = String.unsafe_get str (!i) in
                                                
                                                (* skip quoted characters *)
                                                while not (List.mem
                                                (String.unsafe_get str (!i + !j)) quote) do
                                                        incr j;
                                                        if (!i + !j) > len -
                                                        sublen then raise
                                                        (Invalid_quote q)
                                                done;
                                                
                                                !i + !j
                                                end
                                     else !i;

                                
                                while String.unsafe_get str (!i + !j) =
                                      String.unsafe_get sub !j do
                                        incr j;
                                        if !j = sublen then begin found := !i; raise Exit; end;
                                done;

                                incr i;
                                end
                        done;
                        raise ExtString.String.Invalid_string
                with
                        Exit -> !found
;;


(** [qsplit ~quote:['"'] s sep] splits the string [s] between the first
    occurrence of [sep], ignoring characters enclosed in any of the characters
    passed to the [quote] argument.
        
    @raise Invalid_string if the separator is not found.
    @raise Invalid_quote if a character in the [quote] list appears in [s], but
    is not terminated properly.
    
    @param quote List of characters inside which we don't look for [sep];
    defaults to '"'.
*)
let qsplit ?(quote=['"']) str sep =
        let p = qfind ~quote:quote str sep in
        let len = ExtString.String.length sep in
        let slen = ExtString.String.length str in
        String.sub str 0 p, String.sub str (p + len) (slen - p - len)
;;

        
(** [nqsplit ~quote:['"'] s sep] splits the string [s] into a list of strings
    which are separated by [sep], ignoring characters enclosed in any of the characters
    passed to the [quote] argument.
        
    @raise Invalid_quote if a character in the [quote] list appears in [s], but
    is not terminated properly.

    @param quote List of characters inside which we don't look for [sep];
    defaults to '"'.
*)
let rec nqsplit ?(quote=['"']) str sep =
        try
                let s1 , s2 = qsplit ~quote:quote str sep in
                s1 :: nqsplit ~quote:quote s2 sep
        with
                ExtString.String.Invalid_string -> [str]
;;


Any interest?

-- 
Jesse Guardiani, Systems Administrator
WingNET Internet Services,
P.O. Box 2605 // Cleveland, TN 37320-2605
423-559-LINK (v)  423-559-5145 (f)
http://www.wingnet.net

Re: [Ocaml-lib-devel] Re: Dllist ? IndexTree ?

[Brian H. <bh...@sp...>]

On Tue, 7 Sep 2004, Jesse Guardiani wrote:

> Nicolas Cannasse wrote:
> 
> > What about the two newest data structures dllist and index tree ?
> > Is there final implementation proposal available somewhere to include in
> > ExtLib ?
> 
> Gimme a day or two to finalize Dllist. I had planned to work on it this
> past weekend (Monday was a Holiday and all), but instead I took a little
> road trip with my wife and son. I'll try to work on it tonight and maybe
> have something final by tommorrow.
> 
> 
> 

I too have been distracted by life.  I'll try to get something akin to a 
release out today.

-- 
"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

Re: [Ocaml-lib-devel] Re: Dllist ? IndexTree ?

[Nicolas C. <war...@fr...>]

> Nicolas Cannasse wrote:
>
> > What about the two newest data structures dllist and index tree ?
> > Is there final implementation proposal available somewhere to include in
> > ExtLib ?
>
> Gimme a day or two to finalize Dllist. I had planned to work on it this
> past weekend (Monday was a Holiday and all), but instead I took a little
> road trip with my wife and son. I'll try to work on it tonight and maybe
> have something final by tommorrow.

It's ok you don't need to hurry that much :)

Nicolas

[Ocaml-lib-devel] Re: Dllist ? IndexTree ?

[Jesse G. <je...@wi...>]

Nicolas Cannasse wrote:

> What about the two newest data structures dllist and index tree ?
> Is there final implementation proposal available somewhere to include in
> ExtLib ?

Gimme a day or two to finalize Dllist. I had planned to work on it this
past weekend (Monday was a Holiday and all), but instead I took a little
road trip with my wife and son. I'll try to work on it tonight and maybe
have something final by tommorrow.


-- 
Jesse Guardiani, Systems Administrator
WingNET Internet Services,
P.O. Box 2605 // Cleveland, TN 37320-2605
423-559-LINK (v)  423-559-5145 (f)
http://www.wingnet.net

[Ocaml-lib-devel] Dllist ? IndexTree ?

[Nicolas C. <war...@fr...>]

What about the two newest data structures dllist and index tree ?
Is there final implementation proposal available somewhere to include in
ExtLib ?

Nicolas Cannasse

Re: [Ocaml-lib-devel] IO.really_nread

[Bardur A. <oca...@sc...>]

On Tue, Aug 31, 2004 at 01:46:45PM +0200, Nicolas Cannasse wrote:

> > Fair enough. :)
> > 
> > But then IO.nread is subject to a similar problem: It will
> > busy-wait on a non-blocking input until more input becomes
> > available instead of returning the string read so far.
> > 
> > I fixed that and added two more really_... functions which
> > do exactly the same as their regular counterparts except
> > that they raise an exception if unable to read/write the
> > requested amount.
> > 
> > I've attached the patch I intend to apply to add the
> > functions
> > 
> > really_nread
> > really_input
> > really_output
> > 
> > and to fix IO.nread when used on non-blocking IO.
> > 
> > Any objections? (Obviously, I'll add declarations and
> > documentation to IO.mli if accepted.)
> 
> Looks ok for me.
> 

Committed.

-- 
Bardur Arantsson
<ba...@im...>
<ba...@sc...>

- Leela, you've got mail!
- (sighs)
- It's not spam!
                                    Computer and Leela | Futurama

[Ocaml-lib-devel] OptParse.Opt.set added

[Bardur A. <oca...@sc...>]

Hi all,

I've added a "set" function to the Opt module so that you
can now set the value of an option programmatically. This
can be quite useful if you want to provide e.g. three
distinct options where value of the third can be computed
from any other two. An example from network benchmarking
would be: 1) number of packets, 2) packet size, and 3)
total number of bytes to send. Instead of having to
declaring three more variables to hold the "filled-in"
values, you can just set the missing option's value directly
and proceed as normal.

Unfortunately it couldn't be implemented without a _tiny_
amount of breakage, but you will _only_ see this if you've
implemented your own option types... and even then you
only have to provide one more trivial closure for the
option definition. Since almost nobody is affected by this
change I decided to commit immediately.

The diff is attached.

Cheers,

-- 
Bardur Arantsson
<ba...@im...>
<ba...@sc...>

Who is General Failure, and why is he reading my disk?

Re: [Ocaml-lib-devel] IO.really_nread

[Nicolas C. <war...@fr...>]

> Fair enough. :)
> 
> But then IO.nread is subject to a similar problem: It will
> busy-wait on a non-blocking input until more input becomes
> available instead of returning the string read so far.
> 
> I fixed that and added two more really_... functions which
> do exactly the same as their regular counterparts except
> that they raise an exception if unable to read/write the
> requested amount.
> 
> I've attached the patch I intend to apply to add the
> functions
> 
> really_nread
> really_input
> really_output
> 
> and to fix IO.nread when used on non-blocking IO.
> 
> Any objections? (Obviously, I'll add declarations and
> documentation to IO.mli if accepted.)

Looks ok for me.

Nicolas

Re: [Ocaml-lib-devel] IO.really_nread

[Bardur A. <oca...@sc...>]

On Tue, Aug 31, 2004 at 09:49:01AM +0200, Nicolas Cannasse wrote:

> > > This will cause an infinite loop on non-blocking IO (r = 0).
> > > You might want to raise Sys_blocked_io in this case.
> > >
> >
> > Hmm... the ***channel adapter*** uses Pervasives.input to read.
> 
> Yes, but not other adapters... Please have a look at specification here :
> http://www.ocaml-programming.de/rec/IO-Classes.html
> 
> Nicolas
> 
> 

Fair enough. :)

But then IO.nread is subject to a similar problem: It will
busy-wait on a non-blocking input until more input becomes
available instead of returning the string read so far.

I fixed that and added two more really_... functions which
do exactly the same as their regular counterparts except
that they raise an exception if unable to read/write the
requested amount.

I've attached the patch I intend to apply to add the
functions

	really_nread
	really_input
	really_output
	
and to fix IO.nread when used on non-blocking IO.

Any objections? (Obviously, I'll add declarations and
documentation to IO.mli if accepted.)

-- 
Bardur Arantsson
<ba...@im...>
<ba...@sc...>

On a posted sign: "Bill stickers will be prosecuted!" Written
next to it: "Bill Stickers is innocent!"

Re: [Ocaml-lib-devel] [BUG] read_i32 produces incorrect results

[Nicolas C. <war...@fr...>]

> I think I found a bug in IO.read_i32 implementation. The current
> implementation is (roughly):

Thanks for the report, this is now fixed on CVS.

Regards,
Nicolas Cannasse

Re: [Ocaml-lib-devel] IO.really_nread

[Nicolas C. <war...@fr...>]

> > This will cause an infinite loop on non-blocking IO (r = 0).
> > You might want to raise Sys_blocked_io in this case.
> >
>
> Hmm... the ***channel adapter*** uses Pervasives.input to read.

Yes, but not other adapters... Please have a look at specification here :
http://www.ocaml-programming.de/rec/IO-Classes.html

Nicolas