I'm trying to barf binary floating point
values out of a LISP environment and into
a file.
The file streaming stuff seems to be geared
towards bytes and text. This would be fine if
it let you just throw the bits of an IEEE (hah)
standard single-float onto a stream.
I'm obviously missing something simple. Could
someone please put me out of my misery?
I'm using version 4.3.2 of Allegro Common Lisp, BTW.
Ewan Dennis
Innerworkings Ltd.
Ewan Dennis wrote:
>
> I'm trying to barf binary floating point
> values out of a LISP environment and into
> a file.
>
> The file streaming stuff seems to be geared
> towards bytes and text. This would be fine if
> it let you just throw the bits of an IEEE (hah)
> standard single-float onto a stream.
>
> I'm obviously missing something simple. Could
> someone please put me out of my misery?
>
> I'm using version 4.3.2 of Allegro Common Lisp, BTW.
Here are some hints, I hope this helps.
(with-open-file (stream "d:/temp/float-bin" :direction :output :if-exists :supersede :element-type 'unsigned-byte)
(write-binary-float (float-binary 1.23) stream))
(defun float-to-binary (float)
(multiple-value-bind (mantissa exponent sign) (integer-decode-float float)
(let ((value (dpb (ldb (byte 8 0) (+ 150 exponent)) (byte 8 23) (ldb (byte 23 0) mantissa))))
(if (minusp sign)
(dpb 1 (byte 31 1) value)
value))))
;;;
;;; big endian or little endian?
;;;
(defun write-binary-float (float stream)
(write-byte (ldb (byte 8 24) float) stream)
(write-byte (ldb (byte 8 16) float) stream)
(write-byte (ldb (byte 8 8) float) stream)
(write-byte (ldb (byte 8 0) float) stream))
* Ewan Dennis <····@innerworkings.co.uk>
| I'm trying to barf binary floating point values out of a LISP environment
| and into a file.
that is actually not a meaningful request.
| The file streaming stuff seems to be geared towards bytes and text. This
| would be fine if it let you just throw the bits of an IEEE (hah) standard
| single-float onto a stream.
why do you need this? do you need to talk to a different program? do
you think (or know) that reading and writing floating-point numbers is
a real bottle-neck in your application?
| I'm obviously missing something simple. Could someone please put me out
| of my misery?
could be you simply need FASL-WRITE and FASL-READ in Allegro CL 4.3.2.
these functions preserve the notion of type that exists in memory, and
which does _not_ exist in a file to which you have just dumped a bunch of
bytes, just like raw bytes in memory doesn't have crucial type knowledge
in C, but it has in Common Lisp.
as I told someone else: chuck the C mind-set and re-evaluate the problem.
#:Erik
Erik Naggum wrote:
<snipetty>
> why do you need this? do you need to talk to a different program? do
> you think (or know) that reading and writing floating-point numbers is
> a real bottle-neck in your application?
I'm talking to the rest of the world, I'm afraid. My LISP code is not
running
on its own.
> as I told someone else: chuck the C mind-set and re-evaluate the problem.
Nice idea, and I like what little I know about the functional paradigm
but
I'm working on a big project in which the LISP part is only a component.
Cheers for the kick up the proverbial. I understand the argument.
Defense over.
Ewan Dennis
* Ewan Dennis <····@innerworkings.co.uk>
| I'm talking to the rest of the world, I'm afraid. My LISP code is not
| running on its own.
then there are two options: (1) realize that you will never have portable
CL code if it tries to write out bytes from memory, anyway, and then just
do it, or (2) use a foreign function to do it for you and call it with
the floating point values. option (2) is probably a lot easier if you
don't like to experiment with system internals.
#:Erik
>>>>> "Erik" == Erik Naggum <····@naggum.no> writes:
Erik> * Ewan Dennis <····@innerworkings.co.uk>
Erik> | I'm talking to the rest of the world, I'm afraid. My LISP code is not
Erik> | running on its own.
Erik> then there are two options: (1) realize that you will never have portable
Erik> CL code if it tries to write out bytes from memory, anyway, and then just
Erik> do it, or (2) use a foreign function to do it for you and call it with
Erik> the floating point values. option (2) is probably a lot easier if you
Erik> don't like to experiment with system internals.
Why is (1) true? Someone already sent portable code to get at the
bits of a float and pack them in a way that could be sent out to the
outside world.
Ray