Precision setting for spreadsheet calculation

Precision setting for spreadsheet calculation

[Michael Brand]

Hi all,


1) I tried to set the precision for the spreadsheet calculation to p20 like 
mentioned in the Org info manual "3.5.2 Formula syntax for Calc":
I got

| 0.1666666700000000 |
#+TBLFM: $1 = 1 / 2 * 3; %.16f p20

but I expect

| 0.1666666666666667 |
#+TBLFM: $1 = 1 / 2 * 3; %.16f p20

What am I missing? org-version is the current 6.34c, GNU Emacs version is 21.1.1


Thanks,
Michael

Re: Precision setting for spreadsheet calculation

[Carsten Dominik]

On Feb 26, 2010, at 9:16 PM, Michael Brand wrote:

> Hi all,
>
>
> 1) I tried to set the precision for the spreadsheet calculation to  
> p20 like mentioned in the Org info manual "3.5.2 Formula syntax for  
> Calc":
> I got
>
> | 0.1666666700000000 |
> #+TBLFM: $1 = 1 / 2 * 3; %.16f p20
>
> but I expect
>
> | 0.1666666666666667 |
> #+TBLFM: $1 = 1 / 2 * 3; %.16f p20

You need p20f20

p20 is for internal computation, f20 for youput, and the table
gets the value through output.  Confusing, yes....


HTH

- Carsten

>
> What am I missing? org-version is the current 6.34c, GNU Emacs  
> version is 21.1.1
>
>
> Thanks,
> Michael
>
>
> _______________________________________________
> Emacs-orgmode mailing list
> Please use `Reply All' to send replies to the list.
> Emacs-orgmode@gnu.org
> http://lists.gnu.org/mailman/listinfo/emacs-orgmode

- Carsten

Re: Precision setting for spreadsheet calculation

[Michael Brand]

Carsten Dominik wrote:
> You need p20f20
> 
> p20 is for internal computation, f20 for youput, and the table
> gets the value through output.  Confusing, yes....

| 0.1666666666666667 |
#+TBLFM: $1 = 1 / 2 * 3; %.16f p20 f20

like expected. But instead of

| 0.166666666666666660 |
#+TBLFM: $1 = 1 / 2 * 3; %.18f p20 f20

which I get I would expect

| 0.166666666666666667 |
#+TBLFM: $1 = 1 / 2 * 3; %.18f p20 f20

Re: Precision setting for spreadsheet calculation

[Carsten Dominik]

Hi Michael,

these are the internals of calc, about which I cannot do much.

- Carsten

On Feb 26, 2010, at 10:12 PM, Michael Brand wrote:

> Carsten Dominik wrote:
>> You need p20f20
>> p20 is for internal computation, f20 for youput, and the table
>> gets the value through output.  Confusing, yes....
>
> | 0.1666666666666667 |
> #+TBLFM: $1 = 1 / 2 * 3; %.16f p20 f20
>
> like expected. But instead of
>
> | 0.166666666666666660 |
> #+TBLFM: $1 = 1 / 2 * 3; %.18f p20 f20
>
> which I get I would expect
>
> | 0.166666666666666667 |
> #+TBLFM: $1 = 1 / 2 * 3; %.18f p20 f20

- Carsten

Re: Precision setting for spreadsheet calculation

[Michael Brand]

Carsten Dominik wrote:
> these are the internals of calc, about which I cannot do much.

But

| 0.16666666666666666667 |
#+TBLFM: $1 = 1 / 2 * 3; p20 f20

shows that the internals of calc are ok, isn't it?

> Michael Brand wrote:
>> But instead of
>>
>> | 0.166666666666666660 |
>> #+TBLFM: $1 = 1 / 2 * 3; %.18f p20 f20
>>
>> which I get I would expect
>>
>> | 0.166666666666666667 |
>> #+TBLFM: $1 = 1 / 2 * 3; %.18f p20 f20

Re: Precision setting for spreadsheet calculation

[Carsten Dominik]

On Feb 26, 2010, at 10:58 PM, Michael Brand wrote:

> Carsten Dominik wrote:
>> these are the internals of calc, about which I cannot do much.
>
> But
>
> | 0.16666666666666666667 |
> #+TBLFM: $1 = 1 / 2 * 3; p20 f20
>
> shows that the internals of calc are ok, isn't it?

Hi Michael,

you are completely right, I was wrong.

The other issue that is playing into this is that when
using a printf format specifier, the number is first
converted into a Lisp number, which is a double, with
16 significant digits...  At this step, the additional
digits are lost.

Maybe you find this interesting enough so that I can
persuade you to write a FAQ entry about these issues? :-)

- Carsten


>
>> Michael Brand wrote:
>>> But instead of
>>>
>>> | 0.166666666666666660 |
>>> #+TBLFM: $1 = 1 / 2 * 3; %.18f p20 f20
>>>
>>> which I get I would expect
>>>
>>> | 0.166666666666666667 |
>>> #+TBLFM: $1 = 1 / 2 * 3; %.18f p20 f20

- Carsten

Re: Precision setting for spreadsheet calculation

[Michael Brand]

Carsten Dominik wrote:
 > when using a printf format specifier, the number
 > is first converted into a Lisp number, which is
 > a double, with 16 significant digits
Thanks for the clarification.

> Maybe you find this interesting enough so that I can
> persuade you to write a FAQ entry about these issues? :-)
Yes, within a few days I will come back with this and/or a texinfo patch for 
the Org info manual, at least chapter "3.5.2 Formula syntax for Calc".

Re: Precision setting for spreadsheet calculation

[Michael Brand]

Hi Carsten

Michael Brand wrote:
> Yes, within a few days I will come back with this and/or a texinfo patch 
> for the Org info manual, at least chapter "3.5.2 Formula syntax for Calc".

The patch below is my proposal for the manual. Please feel free to improve it.

I also played around quite a bit with the precision settings and will put the 
resulting examples into the Worg FAQ. But I would like to postpone my Worg 
FAQ update because its link to the manual would be confusing before the 
manual update itself will have been made.

Is there a small issue with the behavior of the section link
http://orgmode.org/manual/index.html#Top ?
With Firefox 3.0.5 it sometimes jumps to where the html source fragment
<a name="Top"></a>
lies like expected but sometimes it jumps to where the html source fragment
<link title="Top" rel="start" href="#Top">
lies which could be annoying.

Michael


patch for org.texi, `---' file was taken from org-version 6.34c:
=============================================================================
--- a/doc/org.texi	2010-01-18 00:10:46.000000000 +0100
+++ b/doc/org.texi	2010-03-02 20:00:00.000000000 +0100
@@ -2128,8 +2128,11 @@
  @code{org-calc-default-modes}.

  @example
-p20           @r{switch the internal precision to 20 digits}
-n3 s3 e2 f4   @r{normal, scientific, engineering, or fixed display format}
+p20           @r{set the internal Calc calculation precision to 20 digits}
+n3 s3 e2 f4   @r{Normal, scientific, engineering, or fixed}
+              @r{format of the result of Calc passed back to Org.}
+              @r{Calc formatting is unlimited in precision as}
+              @r{long as the Calc calculation precision is greater.}
  D R           @r{angle modes: degrees, radians}
  F S           @r{fraction and symbolic modes}
  N             @r{interpret all fields as numbers, use 0 for non-numbers}
@@ -2139,8 +2142,16 @@
  @end example

  @noindent
-In addition, you may provide a @code{printf} format specifier to
-reformat the final result.  A few examples:
+Unless you use large integer numbers or high-precision-calculation
+and -display for floating point numbers you may alternatively provide a
+@code{printf} format specifier to reformat the Calc result after it has been
+passed back to Org instead of letting Calc already do the
+formatting@footnote{The @code{printf} reformatting is limited in precision
+because the value passed to it is converted into an @code{integer} or
+@code{double}.  The @code{integer} is limited in size by truncating the
+signed value to 32 bits.  The @code{double} is limited in precision to 64
+bits overall which leaves approximately 16 significant decimal digits.}.
+A few examples:

  @example
  $1+$2                @r{Sum of first and second field}
=============================================================================

Re: Precision setting for spreadsheet calculation

[Carsten Dominik]

Hi Michael,

very nice, thanks, I have applied the patch.

- Carsten

On Mar 2, 2010, at 9:15 PM, Michael Brand wrote:

> Hi Carsten
>
> Michael Brand wrote:
>> Yes, within a few days I will come back with this and/or a texinfo  
>> patch for the Org info manual, at least chapter "3.5.2 Formula  
>> syntax for Calc".
>
> The patch below is my proposal for the manual. Please feel free to  
> improve it.
>
> I also played around quite a bit with the precision settings and  
> will put the resulting examples into the Worg FAQ. But I would like  
> to postpone my Worg FAQ update because its link to the manual would  
> be confusing before the manual update itself will have been made.
>
> Is there a small issue with the behavior of the section link
> http://orgmode.org/manual/index.html#Top ?
> With Firefox 3.0.5 it sometimes jumps to where the html source  
> fragment
> <a name="Top"></a>
> lies like expected but sometimes it jumps to where the html source  
> fragment
> <link title="Top" rel="start" href="#Top">
> lies which could be annoying.
>
> Michael
>
>
> patch for org.texi, `---' file was taken from org-version 6.34c:
> = 
> = 
> = 
> = 
> = 
> = 
> = 
> ======================================================================
> --- a/doc/org.texi	2010-01-18 00:10:46.000000000 +0100
> +++ b/doc/org.texi	2010-03-02 20:00:00.000000000 +0100
> @@ -2128,8 +2128,11 @@
> @code{org-calc-default-modes}.
>
> @example
> -p20           @r{switch the internal precision to 20 digits}
> -n3 s3 e2 f4   @r{normal, scientific, engineering, or fixed display  
> format}
> +p20           @r{set the internal Calc calculation precision to 20  
> digits}
> +n3 s3 e2 f4   @r{Normal, scientific, engineering, or fixed}
> +              @r{format of the result of Calc passed back to Org.}
> +              @r{Calc formatting is unlimited in precision as}
> +              @r{long as the Calc calculation precision is greater.}
> D R           @r{angle modes: degrees, radians}
> F S           @r{fraction and symbolic modes}
> N             @r{interpret all fields as numbers, use 0 for non- 
> numbers}
> @@ -2139,8 +2142,16 @@
> @end example
>
> @noindent
> -In addition, you may provide a @code{printf} format specifier to
> -reformat the final result.  A few examples:
> +Unless you use large integer numbers or high-precision-calculation
> +and -display for floating point numbers you may alternatively  
> provide a
> +@code{printf} format specifier to reformat the Calc result after it  
> has been
> +passed back to Org instead of letting Calc already do the
> +formatting@footnote{The @code{printf} reformatting is limited in  
> precision
> +because the value passed to it is converted into an @code{integer} or
> +@code{double}.  The @code{integer} is limited in size by truncating  
> the
> +signed value to 32 bits.  The @code{double} is limited in precision  
> to 64
> +bits overall which leaves approximately 16 significant decimal  
> digits.}.
> +A few examples:
>
> @example
> $1+$2                @r{Sum of first and second field}
> = 
> = 
> = 
> = 
> = 
> = 
> = 
> ======================================================================
>

- Carsten

Re: Precision setting for spreadsheet calculation

[Carsten Dominik]

On Mar 11, 2010, at 10:29 AM, Eric S Fraga wrote:

> On Thu, 11 Mar 2010 07:28:38 +0100, Carsten Dominik <carsten.dominik@gmail.com 
> > wrote:
>>
>> Hi,
>>
>> is anyone able to reproduce this?  I am not, so far....
>>
>> - Carsten
>>
>> Begin forwarded message:
>>
>>> From: Michael Brand <michael.brand@alumni.ethz.ch>
>>> Date: March 5, 2010 9:38:56 AM GMT+01:00
>>> To: Carsten Dominik <carsten.dominik@gmail.com>
>>> Subject: Re: [Orgmode] Precision setting for spreadsheet calculation
>>>
>>> Carsten Dominik wrote:
>>>> Which is the link where you click on to make this happen?
>>> New version complete:
>>>
>>> Is there a small issue with the behavior of the section link
>>> http://orgmode.org/manual/index.html#Top ?
>>> It sometimes jumps to the rendered text line
>>> `Next: Introduction, Previous: (dir), Up: (dir)'  (the anchor,
>>> expected)
>>> and sometimes to
>>> `The Org Manual'  (top of page, wrong)
>>> with Firefox 3.0.5, 3.6 and Windows Explorer 8.0
>>>
>>> When checking http://orgmode.org/manual/index.html at the W3C
>>> validator there are some errors and warnings but I can not help
>>> because I am not a HTML expert:
>>> http://validator.w3.org/check?uri=http%3A%2F%2Forgmode.org%2Fmanual%2Findex.html&charset=(detect+automatically)&doctype=Inline&group=0
>>
>> - Carsten
>
> Carsten,
>
> I can (sort of consistently) reproduce this:
>
> - go to http://orgmode.org/manual/
>
> - click on "12.5 HTML export" (I'm having some problems with HTML
>  export; more on this anon) which takes you to
>
>  http://orgmode.org/manual/HTML-export.html#HTML-export
>
> - click on Exporting at the top which takes you to
>
>  http://orgmode.org/manual/Exporting.html#Exporting
>
> - click on Top which now takes you to
>
>  http://orgmode.org/manual/index.html#Top
>
>  and which is not what was expected, I guess (although it *is* what I
>  would have expected...).

This is where it should go and what I would expect.  I guess I am  
still not getting what the problem is.  Michael?


- Carsten

Re: Precision setting for spreadsheet calculation

[Eric S Fraga]

On Thu, 11 Mar 2010 12:35:42 +0100, Carsten Dominik <carsten.dominik@gmail.com> wrote:
> 
> 
> On Mar 11, 2010, at 10:29 AM, Eric S Fraga wrote:
> 
> > On Thu, 11 Mar 2010 07:28:38 +0100, Carsten Dominik
> > <carsten.dominik@gmail.com > wrote:
> >>
> >> Hi,
> >>
> >> is anyone able to reproduce this?  I am not, so far....
> >>
> >> - Carsten
> >>
> >> Begin forwarded message:
> >>
> >>> From: Michael Brand <michael.brand@alumni.ethz.ch>
> >>> Date: March 5, 2010 9:38:56 AM GMT+01:00
> >>> To: Carsten Dominik <carsten.dominik@gmail.com>
> >>> Subject: Re: [Orgmode] Precision setting for spreadsheet calculation
> >>>
> >>> Carsten Dominik wrote:
> >>>> Which is the link where you click on to make this happen?
> >>> New version complete:
> >>>
> >>> Is there a small issue with the behavior of the section link
> >>> http://orgmode.org/manual/index.html#Top ?
> >>> It sometimes jumps to the rendered text line
> >>> `Next: Introduction, Previous: (dir), Up: (dir)'  (the anchor,
> >>> expected)
> >>> and sometimes to
> >>> `The Org Manual'  (top of page, wrong)
> >>> with Firefox 3.0.5, 3.6 and Windows Explorer 8.0
> >>>
> >>> When checking http://orgmode.org/manual/index.html at the W3C
> >>> validator there are some errors and warnings but I can not help
> >>> because I am not a HTML expert:
> >>> http://validator.w3.org/check?uri=http%3A%2F%2Forgmode.org%2Fmanual%2Findex.html&charset=(detect+automatically)&doctype=Inline&group=0
> >>
> >> - Carsten
> >
> > Carsten,
> >
> > I can (sort of consistently) reproduce this:
> >
> > - go to http://orgmode.org/manual/
> >
> > - click on "12.5 HTML export" (I'm having some problems with HTML
> >  export; more on this anon) which takes you to
> >
> >  http://orgmode.org/manual/HTML-export.html#HTML-export
> >
> > - click on Exporting at the top which takes you to
> >
> >  http://orgmode.org/manual/Exporting.html#Exporting
> >
> > - click on Top which now takes you to
> >
> >  http://orgmode.org/manual/index.html#Top
> >
> >  and which is not what was expected, I guess (although it *is* what I
> >  would have expected...).
> 
> This is where it should go and what I would expect.  I guess I am
> still not getting what the problem is.  Michael?

I guess one point is that Top takes you to different places in the
index.html file. If I go from http://orgmode.org/manual/Exporting.html
by clicking on Top, I get placed at the manually created index of
topics the top line displayed (which is not the first line of the
page, if you understand what I mean) is the one with Next etc.  If I
follow the procedure above, I get placed on the same page but at the
actual top of that page which is the (automatically generated?) table
of contents.

Either destination within index.html is obviously fine (at least to
me!).  I guess Michael's query must have been due to the fact that
depending on where you click from, you get placed in a different place
in the index.html file?

HTH,
eric

Re: Precision setting for spreadsheet calculation

[Michael Brand]

Eric S Fraga wrote:
> Either destination within index.html is obviously fine (at least to
> me!).  I guess Michael's query must have been due to the fact that
> depending on where you click from, you get placed in a different place
> in the index.html file?

Almost yes. Even the _same_ section_link can take you to different places at 
different times. Only now I found the probably easiest way to reproduce: 
start with
http://orgmode.org/manual/index.html
and repeat these two steps as many times as you like

1) click `1 Introduction' (or any other first level node) in the frame on the 
left side
2) click any `Top' (`1 Introduction' has two of them) in the frame on the 
right side or `Org Mode Manual' in the frame on the left side

Every second time doing these steps I am taken to the info_node_navigation
`Next: Introduction, Previous: (dir), Up: (dir)'
This is what I expect since following a `Top' from a info_node_navigation 
should take you simply to a info_node_navigation again in my opinion.
But every other second time I am taken to
`The Org Manual'
This is only the very top of the whole page but not the HTML_anchor
<a name="Top"></a>

Re: Precision setting for spreadsheet calculation

[Eric S Fraga]

On Thu, 11 Mar 2010 14:49:28 -0500, Nick Dokos <nicholas.dokos@hp.com> wrote:
> 
> Michael Brand <michael.brand@alumni.ethz.ch> wrote:

[...]

> 
> o I visit http://orgmode.org/manual/index.html
> o I click `1 Introduction' in the frame on the left side.
> o I click 'Previous: Top' in the Navigation line - I don't see a "frame
>   on the right side" at all.
> 
> Repeating the last two steps ad nauseam always takes me to the same place:
> the Navigation line above "Org Mode Manual"; it never takes me to the very
> top "The Org Manual". I also tried other first level nodes with the same
> results.
> 
> Firefox 3.0.18 on Ubuntu 8.10 (Jaunty).
> 
> HTH,
> Nick

In case this is a useful data point, I am using Firefox 3.5.5 on Debian.

Re: Precision setting for spreadsheet calculation

[Michael Brand]

On 10-03-02 9:15 PM, Michael Brand wrote:
> I also played around quite a bit with the precision settings and will
> put the resulting examples into the Worg FAQ. But I would like to
> postpone my Worg FAQ update because its link to the manual would be
> confusing before the manual update itself will have been made.

I added
http://orgmode.org/worg/org-faq.php#table-high-precision
and
http://orgmode.org/worg/org-faq.php#table-float-fraction

Re: Precision setting for spreadsheet calculation

[Carsten Dominik]

On Apr 18, 2010, at 3:41 PM, Michael Brand wrote:

> On 10-03-02 9:15 PM, Michael Brand wrote:
>> I also played around quite a bit with the precision settings and will
>> put the resulting examples into the Worg FAQ. But I would like to
>> postpone my Worg FAQ update because its link to the manual would be
>> confusing before the manual update itself will have been made.
>
> I added
> http://orgmode.org/worg/org-faq.php#table-high-precision
> and
> http://orgmode.org/worg/org-faq.php#table-float-fraction

Hi Michael,

this is really very useful, I learned a lot from reading these two new  
FAQs.

Thank you for your efforts.

- Carsten