Hi Regina;
Thank you for the explanations.
Da: Regina Henschel
Hi Pedro,
Pedro Giffuni schrieb:
Hi;
While digging around Calc I see some things that could be ordered better.
One change that I am considering is moving RAND() to the
Analysis scaddin and bringing the error erf() function in base Calc.
This has reasons of course: in the case of RAND() I want it in the same
file as RANDBETWEEN() so I can use them cleanly.. erf() is something
I would consider doing with Boost, and I don't want to have to link boost
into the scaddin for just one function.
There are probably other changes that can be done too. Just wanted to make
sure if there is some criteria to keep some function in or out of the base
Calc
or in a scaddin.
Long time ago Addin was used for those functions which are not available in a
standard Excel installation, but need an addin installed in Excel. Another
reason for Addin has been, to show, how own functions can be defined without
integrating them directly into the interpreter (example ROT13). I have never
done such thing and do not know whether it still works. But there are two
issues about it, https://issues.apache.org/ooo/show_bug.cgi?id=101386 and
https://issues.apache.org/ooo/show_bug.cgi?id=98149
In the meantime Excel has moved the functions to the standard installation and
most of these functions are specified in ODF. Besides the task to give a way
to define own functions, there is no need for scaddin at all.
So considering that all, I think, that at least all functions, which are
defined in ODF, should go to the normal sc.
OK, I don't really want to move things around very much, plus these things have
to be done in careful order.
I think I found a way to do what I need without moving things around so we will
see.
I do prefer to ask around before doing real changes so bear with me ;-).
I do not know, whether a move has any influence on opening old documents,
which do not use the ODF namespace oc.
I do not understand your comment on erf. The function erf and erfc are moved
to rtl::math with https://issues.apache.org/ooo/show_bug.cgi?id=97091
What I see is that we are keeping in rtl::math a lot of things that the system
usually provides. It's
good to have them there if they are going to be used in Addins but I am not
sure the implementations
are actually too good. In the case of erf, I think anything related to the
Normal Distribution is
absolutely critical and essential and I would have expected the implementation
to be along with
the other stats functions.
TBH, I haven't even started to look at the stats stuff. I am very busy with
other projects.
You are considering far-reaching changes. Therefore I suggest to speak with
Eike before doing it.
I am actually going very slow, I am surprised that no one had done these things
already but then
I only looked at this after the pain of updating the ancient boost we were
carrying.
Of course discussion with people that know this better is welcome and that's
why I bring the
subject to the public lists. I suspect Eike is reading and that he will chime
in if he notices my
sharp axe dangerously near his baby ;-).
Pedro.
