Re: Derivates docs

"Andreas J. Guelzow" wrote:
Neil Booth wrote:
For example, Excel cluttering the namespace with the bond
functions like it does is not ideal, IMO.  If we create
any Gnumeric-specific functions for bonds, we should give
them a common prefix.

And I would rather go the other way: any Excel function we don't like
should be pre (or post) fixed while the `good' ones (whether from XL or
not)need no prefix.

Eye of the beholder, no?

I suggest namespaces can solve some of this.  Allow '::' as part of
names, so number theory functions look like NUMTHEORY::ISPRIME, and
Excel compatibility stuff is XL::PROPER or whatever, then an Excel
compatible spreadsheet mode (say when an XLS is imported) exports the
XL:: namespace into the global one so the functions are accessible as
plain PROPER.  There's no reason that the Excel functions can't be
split into different libraries so that
Andreas's request is met: the Excel functions that seem a good and
useful base can be made available and the dodgy financial stuff
can be omitted.  (That doesn't require a different prefix from XL::
necessarily). Also Andreas mentioned (I think) command completion
for functions which I think is also handled more easily with a
structural distinction of the library prefix.

C got by fine for the most part with simple prefixes for library
functions, but the process of unwinding that in C++ has been
pretty horrible: namespace std which doesn't have writev() in it,
preprocessor defines that run amok, etc.  Conclusion might be it's
better to make allowances for namespaces early.  Further, I can
think of no language where the addition of a simple namespace
facility was a net loss or inconvenience, and this one would be
pretty transparent.

The core question I can see is the syntax-space issue. '::' doesn't
mean anything now in Excel as far as I know.  What if Excel 2004
decided to use it for COM method invocation or .NET services or
some such?  Does that seem an important risk?


Cc:d to jody gnome org as well, since my messages appear to fall
into a pit of twisty moderators. I've tried to join the list but
at my work address so I think this might bounce.

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