On Sun, 23 Mar 2008 09:23:24 -0700, Ian Osgood wrote:
> On Mar 22, 9:08 pm, Larry Webber <theco...@[EMAIL PROTECTED]
> wrote:
>> On Sat, 22 Mar 2008 19:28:34 -0700, Bruce McFarling wrote:
>> > On Mar 22, 7:57 pm, Larry Webber <theco...@[EMAIL PROTECTED]
> wrote:
>> >> I am using both gforth and an alpha release of vfxlin forth. What is
>> >> the recommended way of writing floating point values to a file? For
>> >> writing integer values there are the "pictured number conversion"
>> >> words. What is the recommended ****table way to do this with gforth
>> >> and vfxlin?
>>
>> > Not because I know the answer, but because I'm interested in hearing
>> > it:
>>
>> > (1) Are you asking about writing a text string representing the
>> > floating point value for later display?
>>
>> > (2) Or are you asking about writing a floating point value in binary
>> > form for later retrieval and computation?
>>
>> Bruce,
>>
>> The latter. The need is to write floating point values to a file. The
>> data will later be read by a mathematical program which will both plot
>> the values and perform certain numerical analysis on the data.
>>
>> A friend was able to "get under the hood" of gforth and figure out how
>> to vector the output to a file but we would like a more straight
>> forward method which is implementation neutral. I don't see a
>> straightforward way of doing this presently.
>>
>> Thanks,
>>
>> Larry
>
> So now you need to tell us the format in which your math program expects
> floating point numbers.
>
> From your comment about revectoring, I'm assuming you require a text
> representation instead of binary IEEE, and you were trying to revector
> F. or one of its variations. Look to the implementation of these words
> (SEE F. SEE FE.) for non-****table factors for getting string
> representations (GNU Forth: F$ SCRATCH). The only ****table word for
> getting decimal text from a FLOAT is REPRESENT ( addr len -- exponent
> sign? valid? ) ( F: f -- ).
>
> http://www.forthfreak.net/dpans/dpans12.htm#12.6.1.2143
>
> From REPRESENT it is relatively straightforward to get an exponential
> form (-0.DDDeNNN) suitable for re-parsing. Much of the complexity of the
> implementation of F. is dealing with pretty-printing issues like
> non-exponential forms, leading and trailing zeros, and limiting
> precision.
>
> This brings up a question: is there a standard for textual
> representation of floating point numbers? I only know of the defacto
> standard set by the C standard library's execrable printf/scanf("%f").
>
> Ian
Ian,
My reply from earlier this evening seemed to have dropped in the bit
bucket so forgive me if two posts show up.
Your assumption about the plotting/math program is correct and I failed
to state this clearly in my earlier post. The plotting/math program
expects the data to be ASCII.
Following the information you (and I believe Bruce McFarling in a
different post) provided me, I wrote some words including REPRESENT and
this does 99% of what is needed. These worked on both vfxlin and gforth,
which are the two implementations we are working with currently.
The only remaining task is to insert a decimal point into the character
field at the appropriate place. The information for this location is
returned on the stack after invoking REPRESENT.
To be honest my friend and I were not aware of REPRESENT and we are both
somewhat rusty with Forth (and rooted back in the F-83 days). I think we
need to brush up on the ANSI words a bit. (:=)>
Thanks Ian - and to all who responded to my request for help.
Larry Webber
|
