Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > comp.lang.clipper.visual-objects > #3659 > unrolled thread

Floats & Real8s

Started byMartin <spam@spam.spam>
First post2012-06-22 12:39 +0100
Last post2012-06-25 20:03 -0400
Articles 15 — 6 participants

Back to article view | Back to comp.lang.clipper.visual-objects


Contents

  Floats & Real8s Martin <spam@spam.spam> - 2012-06-22 12:39 +0100
    Re: Floats & Real8s Johan Nel <johan.nel@xsinet.co.za> - 2012-06-22 05:03 -0700
    Re: Floats & Real8s "Gerhard Bunzel" <g.bunzel@domonet.de> - 2012-06-22 14:19 +0200
      Re: Floats & Real8s Johan Nel <johan.nel@xsinet.co.za> - 2012-06-22 05:32 -0700
        Re: Floats & Real8s "Gerhard Bunzel" <g.bunzel@domonet.de> - 2012-06-22 15:13 +0200
          Re: Floats & Real8s Martin <spam@spam.spam> - 2012-06-22 14:27 +0100
          Re: Floats & Real8s Johan Nel <johan.nel@xsinet.co.za> - 2012-06-22 06:36 -0700
            Re: Floats & Real8s Johan Nel <johan.nel@xsinet.co.za> - 2012-06-22 06:39 -0700
      Re: Floats & Real8s D.J.W. van Kooten <public@ic2remove.this.com> - 2012-06-25 09:53 +0200
        Re: Floats & Real8s Johan Nel <johan.nel@xsinet.co.za> - 2012-06-25 01:18 -0700
          Re: Floats & Real8s D.J.W. van Kooten <public@ic2remove.this.com> - 2012-06-25 11:28 +0200
            Re: Floats & Real8s Johan Nel <johan.nel@xsinet.co.za> - 2012-06-25 03:23 -0700
        Re: Floats & Real8s "Ginny Caughey" <gcaughey.online@hotmail.com> - 2012-06-25 07:10 -0400
          Re: Floats & Real8s Johan Nel <johan.nel@xsinet.co.za> - 2012-06-25 07:20 -0700
    Re: Floats & Real8s "Jamal" <vodotnet_nospam@yahoo.com> - 2012-06-25 20:03 -0400

#3659 — Floats & Real8s

FromMartin <spam@spam.spam>
Date2012-06-22 12:39 +0100
SubjectFloats & Real8s
Message-ID<_PYEr.202570$ra1.89885@fx05.am4>
What are the advantages between the two above please?

Thanks

Martin

[toc] | [next] | [standalone]


#3661

FromJohan Nel <johan.nel@xsinet.co.za>
Date2012-06-22 05:03 -0700
Message-ID<36a58803-2c1c-4e44-ad6a-6c8ec636fd55@l5g2000vbo.googlegroups.com>
In reply to#3659
Hi Martin,

Precision and value range are exactly the same for both.  The FLOAT
however have format properties.

From the help files regarding FLOAT:

Approximate Range Precision .NET Framework Type FLOAT and REAL8:
±5.0 × 10^-324 to ±1.7 × 10^308

Unless you are specifically relying on the formatting capabilities of
FLOAT, it is recommended that you use the REAL8 data type instead.

There is a penalty on processing speed of FLOAT compared to REAL8.

Regarding REAL8:
The REAL8 keyword is synonymous with System.Double in .NET

HTH,

Johan Nel
Vulcan VIP
Durban, South Africa.


On Jun 22, 1:39 pm, Martin <s...@spam.spam> wrote:
> What are the advantages between the two above please?
>
> Thanks
>
> Martin

[toc] | [prev] | [next] | [standalone]


#3662

From"Gerhard Bunzel" <g.bunzel@domonet.de>
Date2012-06-22 14:19 +0200
Message-ID<js1nn6$6pb$1@news.m-online.net>
In reply to#3659
Martin,

..it's simple:

- use ALWAYS FLOATS in VO, if you have floating point numbers
- use REAL8 or REAL4 ONLY as paramters in C-DLL's if required

It I remember right, this info was from RvdH


HTH

Gerhard



"Martin" <spam@spam.spam> schrieb im Newsbeitrag 
news:_PYEr.202570$ra1.89885@fx05.am4...
> What are the advantages between the two above please?
>
> Thanks
>
> Martin 

[toc] | [prev] | [next] | [standalone]


#3663

FromJohan Nel <johan.nel@xsinet.co.za>
Date2012-06-22 05:32 -0700
Message-ID<094e2820-d2cb-414c-9cb7-598c6aac9f6f@k5g2000vbf.googlegroups.com>
In reply to#3662
Gerhard,

I did a quick search and it appears from what I have found that the
opposite is actually advocated.

Use REAL8 instead of FLOAT.

If you use late bound code yes things will be converted to FLOAT
internally, but if you strict type your code REAL8 should have a speed
advantage even in VO.

Regards,

Johan.

On Jun 22, 2:19 pm, "Gerhard Bunzel" <g.bun...@domonet.de> wrote:
> Martin,
>
> ..it's simple:
>
> - use ALWAYS FLOATS in VO, if you have floating point numbers
> - use REAL8 or REAL4 ONLY as paramters in C-DLL's if required
>
> It I remember right, this info was from RvdH
>
> HTH
>
> Gerhard
>
> "Martin" <s...@spam.spam> schrieb im Newsbeitragnews:_PYEr.202570$ra1.89885@fx05.am4...
>
>
>
>
>
>
>
> > What are the advantages between the two above please?
>
> > Thanks
>
> > Martin

[toc] | [prev] | [next] | [standalone]


#3664

From"Gerhard Bunzel" <g.bunzel@domonet.de>
Date2012-06-22 15:13 +0200
Message-ID<js1qsu$eh6$1@news.m-online.net>
In reply to#3663
Johan,

if it's better to use REAL8, why does VO use FLOAT's for all floating point 
numbers inside the runtime?
REAL8 is only used in a handful of functions or to do some conversion for 
OLE.


Here you will find something about that problem:
https://groups.google.com/group/comp.lang.clipper.visual-objects/browse_thread/thread/b695ed7d54574383/d387fadde9f1d80f


Thats's a short extract from Ginny's post:
...
I attended the same session in Germany that Fru gave that Geoff presented. 
And
Geoff is right - Fru advised using FLOATs unless you have a specific
interrop-related need to do otherwise. If he (or other members of the CA dev
team) has a newer opinion, that's ok too and I'm sure a quote from the 
material
you have would help clear things up.

Ginny


So I think, FLOAT's are to be used within VO.



HTH

Gerhard




"Johan Nel" <johan.nel@xsinet.co.za> schrieb im Newsbeitrag 
news:094e2820-d2cb-414c-9cb7-598c6aac9f6f@k5g2000vbf.googlegroups.com...
Gerhard,

I did a quick search and it appears from what I have found that the
opposite is actually advocated.

Use REAL8 instead of FLOAT.

If you use late bound code yes things will be converted to FLOAT
internally, but if you strict type your code REAL8 should have a speed
advantage even in VO.

Regards,

Johan.

On Jun 22, 2:19 pm, "Gerhard Bunzel" <g.bun...@domonet.de> wrote:
> Martin,
>
> ..it's simple:
>
> - use ALWAYS FLOATS in VO, if you have floating point numbers
> - use REAL8 or REAL4 ONLY as paramters in C-DLL's if required
>
> It I remember right, this info was from RvdH
>
> HTH
>
> Gerhard
>
> "Martin" <s...@spam.spam> schrieb im 
> Newsbeitragnews:_PYEr.202570$ra1.89885@fx05.am4...
>
>
>
>
>
>
>
> > What are the advantages between the two above please?
>
> > Thanks
>
> > Martin

[toc] | [prev] | [next] | [standalone]


#3665

FromMartin <spam@spam.spam>
Date2012-06-22 14:27 +0100
Message-ID<5p_Er.231193$2H1.3450@fx15.am4>
In reply to#3664
thanks all

On 22/06/2012 14:13, Gerhard Bunzel wrote:
> Johan,
>
> if it's better to use REAL8, why does VO use FLOAT's for all floating point
> numbers inside the runtime?
> REAL8 is only used in a handful of functions or to do some conversion for
> OLE.
>
>
> Here you will find something about that problem:
> https://groups.google.com/group/comp.lang.clipper.visual-objects/browse_thread/thread/b695ed7d54574383/d387fadde9f1d80f
>
>
> Thats's a short extract from Ginny's post:
> ...
> I attended the same session in Germany that Fru gave that Geoff presented.
> And
> Geoff is right - Fru advised using FLOATs unless you have a specific
> interrop-related need to do otherwise. If he (or other members of the CA dev
> team) has a newer opinion, that's ok too and I'm sure a quote from the
> material
> you have would help clear things up.
>
> Ginny
>
>
> So I think, FLOAT's are to be used within VO.
>
>
>
> HTH
>
> Gerhard
>
>
>
>
> "Johan Nel" <johan.nel@xsinet.co.za> schrieb im Newsbeitrag
> news:094e2820-d2cb-414c-9cb7-598c6aac9f6f@k5g2000vbf.googlegroups.com...
> Gerhard,
>
> I did a quick search and it appears from what I have found that the
> opposite is actually advocated.
>
> Use REAL8 instead of FLOAT.
>
> If you use late bound code yes things will be converted to FLOAT
> internally, but if you strict type your code REAL8 should have a speed
> advantage even in VO.
>
> Regards,
>
> Johan.
>
> On Jun 22, 2:19 pm, "Gerhard Bunzel" <g.bun...@domonet.de> wrote:
>> Martin,
>>
>> ..it's simple:
>>
>> - use ALWAYS FLOATS in VO, if you have floating point numbers
>> - use REAL8 or REAL4 ONLY as paramters in C-DLL's if required
>>
>> It I remember right, this info was from RvdH
>>
>> HTH
>>
>> Gerhard
>>
>> "Martin" <s...@spam.spam> schrieb im
>> Newsbeitragnews:_PYEr.202570$ra1.89885@fx05.am4...
>>
>>
>>
>>
>>
>>
>>
>>> What are the advantages between the two above please?
>>
>>> Thanks
>>
>>> Martin
>
>

[toc] | [prev] | [next] | [standalone]


#3666

FromJohan Nel <johan.nel@xsinet.co.za>
Date2012-06-22 06:36 -0700
Message-ID<eefa5dfe-76a6-4d49-a170-bf84daa29833@cu1g2000vbb.googlegroups.com>
In reply to#3664
Hi Gerhard,

I think here is the "latest" regading the matter by Don Caton in a
Message in this group called Performance part 2:

My recommendation is to avoid using the FLOAT data type, _especially_
in
VO.

The FLOAT data type is an archaic data type that goes back to the
dbase
days.  In addition to the numeric data, it also carries around
formatting information (length and decimals).  This allows you to pull
a
float out of a database record and have it print with the same number
of
digits and decimals as the database field has, for example.

Back in the old days of line printers and text mode screens and fixed
pitch fonts this may have made sense, and if you still need this
capability, then use FLOAT.

However, in VO a FLOAT is stored in garbage collected memory whereas a
REAL8 is stored on the stack (or directly inline in a class).  This
makes a REAL8 much more efficient to access because it eliminates the
indirection of reading the value through a GC reference.   And by
avoiding FLOATs, you eliminate unnecessary GC activity.  Every local
or
ivar that contains a FLOAT must be tracked and updated by the GC after
a
collection, and of course, floats that fall out of scope must be
collected and discarded.  A REAL8 has none of this overhead.

Then a bit further in the message:

Bottom line, avoid FLOAT in both languages unless you need the
formatting information.  A better approach, IMO, would be to use
REAL8s
and then let your display and report code format the data as
necessary.

You didn't mention REAL4, but I'll bring that up anyhow.  There's
little
need for REAL4 in either language, unless you're interfacing with
external code that requires a 4-byte fp value.  AFAIK, the CPU is
still
going to perform fp operations at its highest level of precision, and
in
.NET, 4-byte fp values are promoted internally to the highest
precision
possible for the given platform.  There may be a loss of precision
when
the value is stored back into a local that is typed as REAL4 though.

--
Don

Regards,

Johan
On Jun 22, 3:13 pm, "Gerhard Bunzel" <g.bun...@domonet.de> wrote:
> Johan,
>
> if it's better to use REAL8, why does VO use FLOAT's for all floating point
> numbers inside the runtime?
> REAL8 is only used in a handful of functions or to do some conversion for
> OLE.
>
> Here you will find something about that problem:https://groups.google.com/group/comp.lang.clipper.visual-objects/brow...
>
> Thats's a short extract from Ginny's post:
> ...
> I attended the same session in Germany that Fru gave that Geoff presented.
> And
> Geoff is right - Fru advised using FLOATs unless you have a specific
> interrop-related need to do otherwise. If he (or other members of the CA dev
> team) has a newer opinion, that's ok too and I'm sure a quote from the
> material
> you have would help clear things up.
>
> Ginny
>
> So I think, FLOAT's are to be used within VO.
>
> HTH
>
> Gerhard
>
> "Johan Nel" <johan....@xsinet.co.za> schrieb im Newsbeitragnews:094e2820-d2cb-414c-9cb7-598c6aac9f6f@k5g2000vbf.googlegroups.com...
> Gerhard,
>
> I did a quick search and it appears from what I have found that the
> opposite is actually advocated.
>
> Use REAL8 instead of FLOAT.
>
> If you use late bound code yes things will be converted to FLOAT
> internally, but if you strict type your code REAL8 should have a speed
> advantage even in VO.
>
> Regards,
>
> Johan.
>
> On Jun 22, 2:19 pm, "Gerhard Bunzel" <g.bun...@domonet.de> wrote:
>
>
>
>
>
>
>
> > Martin,
>
> > ..it's simple:
>
> > - use ALWAYS FLOATS in VO, if you have floating point numbers
> > - use REAL8 or REAL4 ONLY as paramters in C-DLL's if required
>
> > It I remember right, this info was from RvdH
>
> > HTH
>
> > Gerhard
>
> > "Martin" <s...@spam.spam> schrieb im
> > Newsbeitragnews:_PYEr.202570$ra1.89885@fx05.am4...
>
> > > What are the advantages between the two above please?
>
> > > Thanks
>
> > > Martin

[toc] | [prev] | [next] | [standalone]


#3667

FromJohan Nel <johan.nel@xsinet.co.za>
Date2012-06-22 06:39 -0700
Message-ID<a4b56f9d-5082-4b5b-927f-9bbca742e722@6g2000vbv.googlegroups.com>
In reply to#3666
Sorry forgot to include the link:

https://groups.google.com/group/comp.lang.clipper.visual-objects/browse_frm/thread/32edf37c7593c6c9/98895c13851a7e20?lnk=gst&q=REAL8+vs+FLOAT#98895c13851a7e20

Regards,

Johan

[toc] | [prev] | [next] | [standalone]


#3670

FromD.J.W. van Kooten <public@ic2remove.this.com>
Date2012-06-25 09:53 +0200
Message-ID<4m5gu7d5q574prbj3lgko0iti57e8d1f0v@4ax.com>
In reply to#3662
On Fri, 22 Jun 2012 14:19:14 +0200, "Gerhard Bunzel"
<g.bunzel@domonet.de> wrote:

Hello Gerhard,

>- use ALWAYS FLOATS in VO, if you have floating point numbers
>- use REAL8 or REAL4 ONLY as paramters in C-DLL's if required
>
>It I remember right, this info was from RvdH
>

I remember that too. In fact I've replaced all real8's in our program
years ago  because of recommendations on conferences.

In this NG: Ginny wrote on 11-12-2000:

"The reason to use FLOATs instead of REAL anything is precisely
because the GC takes care of the memory for them. Have you had any
problems with FLOATs? I never have, and CA's official recommendation
is to use FLOATs rather than REAL whatever."

and Paolo:

"1) Think about functions with USUAL parameters.
    To use those values you have no choice but wrap them into FLOATs
because a USUAL can  contain a FLOAT.
2) Some VO functions return FLOATs."

and Geoff on 9-10-2011:

"REAL4 and REAL8 are static memory objects and FLOAT is of course
dynamic, and hence managed by the GC.  All can be used to represent
floating point numbers by CA highly recommends only the usage of
FLOAT. In fact, at Devcon Germany last year, Angswillie expressed
extreme dismay at the thought of anyone avoiding floats for REAL8. He
should know - he wrote the code <g>."


So that's why I chose FLOAT...

Dick

[toc] | [prev] | [next] | [standalone]


#3671

FromJohan Nel <johan.nel@xsinet.co.za>
Date2012-06-25 01:18 -0700
Message-ID<cb56671a-a671-4acd-a62c-5cf5c0a6c12c@cu1g2000vbb.googlegroups.com>
In reply to#3670
Hi Dick,

Don't you think 2000 a bit outdated?

Have you had a look at the message from Don Caton in 2006?

Regards,

Johan.

On Jun 25, 9:53 am, D.J.W. van Kooten <pub...@ic2remove.this.com>
wrote:
> On Fri, 22 Jun 2012 14:19:14 +0200, "Gerhard Bunzel"
>
> <g.bun...@domonet.de> wrote:
>
> Hello Gerhard,
>
> >- use ALWAYS FLOATS in VO, if you have floating point numbers
> >- use REAL8 or REAL4 ONLY as paramters in C-DLL's if required
>
> >It I remember right, this info was from RvdH
>
> I remember that too. In fact I've replaced all real8's in our program
> years ago  because of recommendations on conferences.
>
> In this NG: Ginny wrote on 11-12-2000:
>
> "The reason to use FLOATs instead of REAL anything is precisely
> because the GC takes care of the memory for them. Have you had any
> problems with FLOATs? I never have, and CA's official recommendation
> is to use FLOATs rather than REAL whatever."
>
> and Paolo:
>
> "1) Think about functions with USUAL parameters.
>     To use those values you have no choice but wrap them into FLOATs
> because a USUAL can  contain a FLOAT.
> 2) Some VO functions return FLOATs."
>
> and Geoff on 9-10-2011:
>
> "REAL4 and REAL8 are static memory objects and FLOAT is of course
> dynamic, and hence managed by the GC.  All can be used to represent
> floating point numbers by CA highly recommends only the usage of
> FLOAT. In fact, at Devcon Germany last year, Angswillie expressed
> extreme dismay at the thought of anyone avoiding floats for REAL8. He
> should know - he wrote the code <g>."
>
> So that's why I chose FLOAT...
>
> Dick

[toc] | [prev] | [next] | [standalone]


#3672

FromD.J.W. van Kooten <public@ic2remove.this.com>
Date2012-06-25 11:28 +0200
Message-ID<bmagu7pjmpujtbiqfku93upp308svu9i3q@4ax.com>
In reply to#3671
On Mon, 25 Jun 2012 01:18:32 -0700 (PDT), Johan Nel
<johan.nel@xsinet.co.za> wrote:

Hello Johan,
>
>Don't you think 2000 a bit outdated?

No. VO hasn't changed in the way it treats REAL/FLOAT variables as far
as I know. 

>Have you had a look at the message from Don Caton in 2006?
>
Sure I did. I participated in that thread as well. Don's approach is
partially based on his main project, Vulcan, and partially in
contradiction of what the original CA designers, Robert and Geoff
thought at that time. As I wrote already then: I don't know anymore
who has the right argument: the original designer of CA or the Vulcan
architect who has spent a small bit of his time on minimal VO
maintenance....

I've replaced all my REAL's to FLOAT based on that advise and I think
Martin should choose whatever he think's is true <g>

Dick

[toc] | [prev] | [next] | [standalone]


#3673

FromJohan Nel <johan.nel@xsinet.co.za>
Date2012-06-25 03:23 -0700
Message-ID<a1f59e90-431f-47e4-8a9b-ee3c8de389ea@w24g2000vby.googlegroups.com>
In reply to#3672
Hi Dick,

I agree everybody should use what they feel comfy with.  I do remember
seeing a message even from Geoff that said REAL8 all the time.

From my perspective, I did find using REAL8 to be faster in
mathematical programming than FLOAT.

Would love if Robert can maybe comment seeing he has done most of the
later work on VO.

Regards,

Johan.

On Jun 25, 11:28 am, D.J.W. van Kooten <pub...@ic2remove.this.com>
wrote:
> On Mon, 25 Jun 2012 01:18:32 -0700 (PDT), Johan Nel
>
> <johan....@xsinet.co.za> wrote:
>
> Hello Johan,
>
>
>
> >Don't you think 2000 a bit outdated?
>
> No. VO hasn't changed in the way it treats REAL/FLOAT variables as far
> as I know.
>
> >Have you had a look at the message from Don Caton in 2006?
>
> Sure I did. I participated in that thread as well. Don's approach is
> partially based on his main project, Vulcan, and partially in
> contradiction of what the original CA designers, Robert and Geoff
> thought at that time. As I wrote already then: I don't know anymore
> who has the right argument: the original designer of CA or the Vulcan
> architect who has spent a small bit of his time on minimal VO
> maintenance....
>
> I've replaced all my REAL's to FLOAT based on that advise and I think
> Martin should choose whatever he think's is true <g>
>
> Dick

[toc] | [prev] | [next] | [standalone]


#3674

From"Ginny Caughey" <gcaughey.online@hotmail.com>
Date2012-06-25 07:10 -0400
Message-ID<4fe84725$0$1599$c3e8da3$14a0410e@news.astraweb.com>
In reply to#3670
Thanks for remembering me, Dick. And I still always use floats in VO.

Ginny

"D.J.W. van Kooten"  wrote in message 
news:4m5gu7d5q574prbj3lgko0iti57e8d1f0v@4ax.com...

On Fri, 22 Jun 2012 14:19:14 +0200, "Gerhard Bunzel"
<g.bunzel@domonet.de> wrote:

Hello Gerhard,

>- use ALWAYS FLOATS in VO, if you have floating point numbers
>- use REAL8 or REAL4 ONLY as paramters in C-DLL's if required
>
>It I remember right, this info was from RvdH
>

I remember that too. In fact I've replaced all real8's in our program
years ago  because of recommendations on conferences.

In this NG: Ginny wrote on 11-12-2000:

"The reason to use FLOATs instead of REAL anything is precisely
because the GC takes care of the memory for them. Have you had any
problems with FLOATs? I never have, and CA's official recommendation
is to use FLOATs rather than REAL whatever."

and Paolo:

"1) Think about functions with USUAL parameters.
    To use those values you have no choice but wrap them into FLOATs
because a USUAL can  contain a FLOAT.
2) Some VO functions return FLOATs."

and Geoff on 9-10-2011:

"REAL4 and REAL8 are static memory objects and FLOAT is of course
dynamic, and hence managed by the GC.  All can be used to represent
floating point numbers by CA highly recommends only the usage of
FLOAT. In fact, at Devcon Germany last year, Angswillie expressed
extreme dismay at the thought of anyone avoiding floats for REAL8. He
should know - he wrote the code <g>."


So that's why I chose FLOAT...

Dick 

[toc] | [prev] | [next] | [standalone]


#3676

FromJohan Nel <johan.nel@xsinet.co.za>
Date2012-06-25 07:20 -0700
Message-ID<2c630b7f-1ea7-4215-9739-3ff2f70cbe19@s9g2000vbg.googlegroups.com>
In reply to#3674
Ginny and others,

Here is a message from Robert that I just found.

We need to compare apples with apples.

--
What I tried to tell was that mixing REAL4/8 values with usuals or
runtime
functions that return floats isn't going to help: USUALS can't contain
REAL
values and all the floating point runtime functions return FLOATS and
not
REAL values. If you mix these VO will add conversion between the
datatypes
to your code.

Robert van der Hulst
info@heliks.nl
www.heliks.nl
--
I don't know where the confusion comes in, but to me that says REAL8
in pure calculations would be the better type to use.  Yes I would
have a performance penalty when I use some internal functions, but I
tend to stay away in my own development from FLOAT types.  For the few
times I do interface to internals that uses USUAL/FLOAT I can live
with the speed penalty, but in general pure REAL8 computing is a lot
faster.

Johan.

On Jun 25, 1:10 pm, "Ginny Caughey" <gcaughey.onl...@hotmail.com>
wrote:
> Thanks for remembering me, Dick. And I still always use floats in VO.
>
> Ginny
>
> "D.J.W. van Kooten"  wrote in messagenews:4m5gu7d5q574prbj3lgko0iti57e8d1f0v@4ax.com...
>
> On Fri, 22 Jun 2012 14:19:14 +0200, "Gerhard Bunzel"
>
> <g.bun...@domonet.de> wrote:
>
> Hello Gerhard,
>
> >- use ALWAYS FLOATS in VO, if you have floating point numbers
> >- use REAL8 or REAL4 ONLY as paramters in C-DLL's if required
>
> >It I remember right, this info was from RvdH
>
> I remember that too. In fact I've replaced all real8's in our program
> years ago  because of recommendations on conferences.
>
> In this NG: Ginny wrote on 11-12-2000:
>
> "The reason to use FLOATs instead of REAL anything is precisely
> because the GC takes care of the memory for them. Have you had any
> problems with FLOATs? I never have, and CA's official recommendation
> is to use FLOATs rather than REAL whatever."
>
> and Paolo:
>
> "1) Think about functions with USUAL parameters.
>     To use those values you have no choice but wrap them into FLOATs
> because a USUAL can  contain a FLOAT.
> 2) Some VO functions return FLOATs."
>
> and Geoff on 9-10-2011:
>
> "REAL4 and REAL8 are static memory objects and FLOAT is of course
> dynamic, and hence managed by the GC.  All can be used to represent
> floating point numbers by CA highly recommends only the usage of
> FLOAT. In fact, at Devcon Germany last year, Angswillie expressed
> extreme dismay at the thought of anyone avoiding floats for REAL8. He
> should know - he wrote the code <g>."
>
> So that's why I chose FLOAT...
>
> Dick

[toc] | [prev] | [next] | [standalone]


#3677

From"Jamal" <vodotnet_nospam@yahoo.com>
Date2012-06-25 20:03 -0400
Message-ID<jsau7k$ano$1@dont-email.me>
In reply to#3659
Here is my take on this subject.

If am calling a VO function that returns a FLOAT, then I use a FLOAT, 
otherwise I use REAL8.
As others have explained, REAL8 has advantages over FLOAT in terms of GC and 
internal formatting.

So, the choice is obvious.

Jamal

"Martin"  wrote in message news:_PYEr.202570$ra1.89885@fx05.am4...

What are the advantages between the two above please?

Thanks

Martin 

[toc] | [prev] | [standalone]


Back to top | Article view | comp.lang.clipper.visual-objects


csiph-web