On Sun, 12 Jan 2020 15:48:05 -0700
"Keith Medcalf" <kmedc...@dessus.com> wrote:
> SQL is simply a Structured Query Language that can be overlaid on a
> variety of datastore models, one of which is the Relational model.
I think that's misleading, Keith, for a couple of reasons.
SQL was invented by IBM expressly and specfically for its "relational"
product. Its wordiness is the product of its roots in "4th generation"
languages in vogue at the time, the promise of which was to permit
users to "write their own reports". Hence the strict select-from-where
syntax, meant to be so simple that untrained users could figure it
out.
While its true, as you say, that many pre-relational systems (and
post-relational ones, later, feh) added SQL on their shingles, those
were never coherent implementations. They were a way to say Yes! when
the question was, "Do you support SQL?"
> Many (most in fact) datastores require that all instances of the same
> "column" in an "entity" be the same type
To the extent that SQL implements relational algebra/calculus, its
utility and consistency *requires* that each column be of a defined
type. For example, if the query includes,
where A between 0 and 1
and A is a *not* a numeric value, then the query is reduced from
rigorous first-order predicate logic to gibberish.
SQLite serves a particular niche very well. The choice not to enforce
type constraints for declared datatypes biases the system toward
ease-of-insertion. That has advantages in the SQLite problem domain,
but the trade-off comes at a price: it makes SQLite *harder* to use in a
rigorous way.
--jkl
_______________________________________________
sqlite-users mailing list
sqlite-users@mailinglists.sqlite.org
http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
