[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Orgmode] Re: org-mode feature is surely a bug!

From: Carsten Dominik
Subject: [Orgmode] Re: org-mode feature is surely a bug!
Date: Fri, 29 May 2009 16:44:06 +0200

Hi Michael,

Thanks for your mail.

what version of Org-mode are you using?  We have changed this
quite some time ago to (float 8) (I see now that the docs still say
5, but this is not the case in the code, it is 8, except maybe
in very old releases).

The reason why it is not
larger is more a display issue and a computation issue.  Since
Org-mode is pure plain text, we need to write all significant
digits into the table, we do not have the option to make the
display a shorter version of an underlying more accurate number,
and writing out all 16 digits of a double precision number
would make the table columns wide and the tables unreadable.

We could have #+PRECISION, but you can also do

M-x customize-variable RET org-calc-default-modes RET
and change it there, for all your files, at least for now,
until I have a better option for per-file settings.

Hope this helps.  Please write next time to address@hidden
(this is our mailing list) so that a searchable record of answers
and questions is created.

Best wishes

- Carsten

On May 29, 2009, at 4:22 PM, Dr. Michael Dowling wrote:

Hello Carsten,

I was adapting org-mode to my applications when I noticed that it could not add up! I had a table in which one column was a column of figures, and I wanted the sum, and org-mode came up with the wrong answer. I was flabbergasted when I discovered that this behaviour was documented as a

Specifically, the offending line was

calc-float-format  (float 5)

Apparently, if the calc default of 12 were used, the tables could look
ugly, so it was thought better to produce a pretty table than a correct

At first I thought that this might be merely a display problem, that
internal calculations would be more accurate, and that I could get the
extra precision that I needed using a printf style print, but no.  The
printed value indeed had an extra digit, but that right most digit was a
meaningless '0'.

I've changed this in my .emacs file, but I would urge a re-think on

As a numerical analyst, I feel comfortable with rounding errors, and
precision, and I accept that it is impossible to supply exact answers to all calculations, but working with only decimal 5 digit precision I find
rather frightening.  (How calc came up with a default of 12 I don't
know; it's capable of arbitary precision, but anything less that the
precision of a double precision of 64 bits (about 16 decimal places),
and all the paraphernalia of guard digits etc. that the hardware
floating point units use is odd.)

Perhaps a line like


would be nice so that the user can choose his precision on file by file
basis, and keeping the above value at 12, the calc default.  I sixth
digit after a decimal point for a tax return is clearly too much, but
for financial calculations involving millions, this would be
unacceptable. It follows that this really should not be hard coded, but
set on a file by file basis.

I don't want to sound negative, though; org-mode is really nifty, and,
although I have still to get into it properly, I have great expectations
that I am sure will be fulfilled beyond all my hopes!  Great work!

Mike Dowling

Dr. Michael L. Dowling
Gaußstr. 27
38106 Braunschweig

reply via email to

[Prev in Thread] Current Thread [Next in Thread]