My suggestion is aimed at simplicity. It proposes one human-readable
number format by default, not a constant interchange of underlying
types. To me it is very much in the spirit of the Sqlite paradigm.
JS
Andrew Piskorski wrote:
On Wed, Nov 02, 2005 at 06:43:31AM -0600, John Stanton wrote:
Subject: Re: [sqlite] Proposed 3.3.0 changes. Was: 5/2==2
You misunderstood the comment. A schoolchild expects to see 5/2 give
exactly 2.5, not an approximate representation of 2.5 in a floating
I question why we have to use binary integers based on the word size of
particular generations of computers when we are storing data. We have
Talk about feature creep... None of that is on the table in
Dr. Hipp's recent proposed changes. It also sounds likely to conflict
with one of Dr. Hipp's principal goals for SQLite, "simplicity" - and
he means of implementation, not just of use. But, it might still be
worth discussing.
First of all, what is required to meet your goal? (I'm not
sufficiently familiar with the issues to know off the top of my head,
but since you're interested in this...) Is adding bignum support (via
GMP or any number of other such libraries) sufficient, or is more than
that needed to properly represent exact fractions, complex numbers,
and the like?
Tcl just integrated bignum support on the Head, which presumably will
ship with Tcl 8.5. Does that support cover everything you want, or
not? What about the "full numerical tower" provided by many Scheme
implementations?
If intersection of this sort of numerics with databases interests you,
you may want to get in touch with Jean-Claude Wippler. I don't recall
what if any plans he has for bignums and the like, but the vector
oriented nature of his Vlerq and Metakit projects makes it likely that
he does indeed have plans, or at least ideas, along those lines:
http://www.vlerq.org/
http://www.equi4.com/
