On Thu, 22 Mar 2012, Talha Yalta wrote:
Thanks very much Prof. Cottrell for the rapid fixes.
> That's OK. The data are stored to at least 12 digits in the
> ODS file, as you can see if you adjust the display format. We
> don't pay attention to the fact that they are _formatted_ to
> show only 3 digits.
I am confused. Being aware of the above, I had copied these series
after first using the "=ROUND(x,3) command on the originals. In my
system, in libreoffice, I see the numbers with 3 decimal places even
if I reformat the cell to show 20 digits.
Ah, you're right, the numbers are stored to only 3 digits in
the ods file. I was looking at the file using gnumeric, which
showed extra stuff when I reformatted the cells. But the real
point is that the "extra stuff" is in fact there in memory.
For example, the number 30976.746 cannot be represented in a C
double: the closest you get is 30976.745999999999185.
LibreOffice is just sanitizing what you're allowed to see
based on your "ROUND(x,3)" instruction, it's not magically
producing a floating point number with an infinite number of
trailing zeros.
That said, I've slightly modified gretl's CSV printing
apparatus so it's a bit less inclined to print trailing junk.
Thanks for the correlogram graph page test case. This should
now be OK in CVS and snapshots. The problem was specific to
the case where EPS rather than PDF graphs are produced.
Finally, the tab labels for new scripts show "Untitled"
although I
remember translating an "Untitled" string. Is this something else that
will be marked for translation later?
It's now marked for translation.
Allin