Reading 16 bit floats from stream?

Reading 16 bit floats from stream?

[Maurizio Colucci]

[-- Attachment #1: Type: text/plain, Size: 212 bytes --]

Hi,

I need to read a sequence of single precision floats (16 bits) from a binary
file. How can I do that?

In case you are curious, I am parsing an MD2 file for a game. :-)

Thanks a lot for any help,

Maurizio

[-- Attachment #2: Type: text/html, Size: 310 bytes --]

Re: Reading 16 bit floats from stream?

[Maurizio Colucci]

[-- Attachment #1: Type: text/plain, Size: 256 bytes --]

On 10/16/06, Maurizio Colucci <maurizio.colucci@gmail.com> wrote:
>
> Hi,
>
> I need to read a sequence of single precision floats (16 bits) from a
> binary file. How can I do that?
>

Sorry, I mean 32 bits. Single precision floats are 32 bits.


Maurizio

[-- Attachment #2: Type: text/html, Size: 648 bytes --]

Re: [Caml-list] Re: Reading 16 bit floats from stream?

[Gerd Stolpmann]

Am Montag, den 16.10.2006, 18:03 +0200 schrieb Maurizio Colucci:
> 
> 
> On 10/16/06, Maurizio Colucci <maurizio.colucci@gmail.com> wrote:
>         Hi,
>         
>         I need to read a sequence of single precision floats (16 bits)
>         from a binary file. How can I do that? 
> 
> Sorry, I mean 32 bits. Single precision floats are 32 bits.

Well, have heard of even 8 bit floats...

32 bit floats can be read. First read the 4 bytes and make an int32 of
them. Then simply call Int32.float_of_bits.

Gerd

-- 
------------------------------------------------------------
Gerd Stolpmann * Viktoriastr. 45 * 64293 Darmstadt * Germany 
gerd@gerd-stolpmann.de          http://www.gerd-stolpmann.de
Phone: +49-6151-153855                  Fax: +49-6151-997714
------------------------------------------------------------

Re: [Caml-list] Re: Reading 16 bit floats from stream?

[Maurizio Colucci]

[-- Attachment #1: Type: text/plain, Size: 758 bytes --]

On 10/16/06, Gerd Stolpmann <info@gerd-stolpmann.de> wrote:
>
> Am Montag, den 16.10.2006, 18:03 +0200 schrieb Maurizio Colucci:
> >
> >
> > On 10/16/06, Maurizio Colucci <maurizio.colucci@gmail.com> wrote:
> >         Hi,
> >
> >         I need to read a sequence of single precision floats (16 bits)
> >         from a binary file. How can I do that?
> >
> > Sorry, I mean 32 bits. Single precision floats are 32 bits.
>
> Well, have heard of even 8 bit floats...
>
> 32 bit floats can be read. First read the 4 bytes


I assume you mean four read_bytes calls...

and make an int32 of
> them.


How do I create an int32 with four bytes?

Then simply call Int32.float_of_bits.


I get this last one, but I can't imagine the previous step.

Thanks

Maurizio

[-- Attachment #2: Type: text/html, Size: 1606 bytes --]

Re: [Caml-list] Re: Reading 16 bit floats from stream?

[Tom]

[-- Attachment #1: Type: text/plain, Size: 51 bytes --]

What about one read_int call? I guess it exists...

[-- Attachment #2: Type: text/html, Size: 55 bytes --]

Re: [Caml-list] Re: Reading 16 bit floats from stream?

[Martin Jambon]

On Mon, 16 Oct 2006, Maurizio Colucci wrote:

> I assume you mean four read_bytes calls...
>
> and make an int32 of
>> them.
>
>
> How do I create an int32 with four bytes?

I can't recall any predefined function. You can read the bytes one by 
one, and use Int32.shift_left and Int32.logor to create the int32 value.

> Then simply call Int32.float_of_bits.
>
>
> I get this last one, but I can't imagine the previous step.
>
> Thanks
>
> Maurizio
>

--
Martin Jambon, PhD
http://martin.jambon.free.fr

Re: [Caml-list] Re: Reading 16 bit floats from stream?

[Tom]

[-- Attachment #1: Type: text/plain, Size: 151 bytes --]

val Pervasives.input_binary_int : in_channel -> int
val Pervasives.read_int : unit -> int

val Marshal.from_channel : in_channel -> 'a

For example...

[-- Attachment #2: Type: text/html, Size: 542 bytes --]

Re: [Caml-list] Re: Reading 16 bit floats from stream?

[Jacques Garrigue]

From: Tom <tom.primozic@gmail.com>

> val Pervasives.input_binary_int : in_channel -> int

This will indeed read a 32-bit integer, but on a 32-bit architecture
ocaml int's are 31-bit, so you loose the highest bit.
Note that on a 64-bit architecture this is sufficient for 32-bit
integers!

> val Pervasives.read_int : unit -> int

This one reads a string...

> val Marshal.from_channel : in_channel -> 'a

And this one reads a marshalled value, which must include a special
header indicating its ocaml representation.

So if you really need to read a full 32-bit integer, none of these
will work, and you must do the job by hand.

Jacques Garrigue

Re: [Caml-list] Re: Reading 16 bit floats from stream?

[Maurizio Colucci]

[-- Attachment #1: Type: text/plain, Size: 1059 bytes --]

On 10/17/06, Jacques Garrigue <garrigue@math.nagoya-u.ac.jp> wrote:
>
> From: Tom <tom.primozic@gmail.com>
>
> > val Pervasives.input_binary_int : in_channel -> int
>
> This will indeed read a 32-bit integer, but on a 32-bit architecture
> ocaml int's are 31-bit, so you loose the highest bit.
> Note that on a 64-bit architecture this is sufficient for 32-bit
> integers!
>
> > val Pervasives.read_int : unit -> int
>
> This one reads a string...
>
> > val Marshal.from_channel : in_channel -> 'a
>
> And this one reads a marshalled value, which must include a special
> header indicating its ocaml representation.
>
> So if you really need to read a full 32-bit integer, none of these
> will work, and you must do the job by hand.
>
> Jacques Garrigue


Thanks to everybody. I found a library called extlib which provides a module
called IO. I solved the problem with

let l =  Int32.float_of_bits (IO.read_real_i32 input)

BTW, I am a bit surprised the standard library does not offer a way to do
that. A simple task can become frustrating.

Bye

Maurizio

[-- Attachment #2: Type: text/html, Size: 1479 bytes --]

Re: [Caml-list] Re: Reading 16 bit floats from stream?

[Gerd Stolpmann]

Am Dienstag, den 17.10.2006, 16:02 +0200 schrieb Maurizio Colucci:

> BTW, I am a bit surprised the standard library does not offer a way to
> do that. A simple task can become frustrating.

I must say that I do not see the point. Although O'Caml is for
high-level tasks it is unavoidable that you must sometimes look at the
details and understand what's going on under the hood. This is more a
matter of the programmer's general knowledge, and not about how good or
bad a certain library is (which includes what _other_ people consider as
important).

Just to scare you. Before release 3.08 there was no Int32.float_of_bits
in O'Caml, and one had to program such a function with bit shifting
primitives. See how ugly this is (look for the now commented-out
function fp8_of_fp4):

https://gps.dynxs.de/svn/lib-ocamlnet2/trunk/code/src/rpc/rtypes.ml

Gerd

-- 
------------------------------------------------------------
Gerd Stolpmann * Viktoriastr. 45 * 64293 Darmstadt * Germany 
gerd@gerd-stolpmann.de          http://www.gerd-stolpmann.de
Phone: +49-6151-153855                  Fax: +49-6151-997714
------------------------------------------------------------