On Thu, 03 Jul 2014 19:00:21 +0200
RSmith <[email protected]> wrote:
> > A vague citation to a million anonymous programs of unknown quality
> > is not a convincing reason to think otherwise. And you and I both
> > know that a random sample of 1 million programs will contain
> > roughly 999000 crappy ones. :-)
> >
>
> The citation is not vague, he did not mean *any* selection of a
> million programs, but indeed all of the millions of programs
> commercially available running sqlite on many billions of devices
> (refer: www.sqlite.org home page) - I would contend that it will be
> hard to find even 1% really bad programs among a million of these
> that do sqlite an injustice for merely existing - but that is just a
> contention since I have no resources to prove it and your estimate of
> "crappy" is unquantifiable.
If the answer to every question, concern, and suggestion is that it
already works for millions of programs, there is no point in discussion
or further development. Just use what's there because it's already
perfect.
Mr. Rubin-Smith is asking a semantic question: what does each error
code *mean*? Does the error indicate malformed input,
resource contention, constraint violation, program logic error, or
system failure? Is it within the user's capacity to correct? Will the
same operation, retried later with the same input, possibly succeed?
If his question is unusual, then it's fair to say so too are such
programmers unusual. I would guess the number of programmers who
understand what is meant by "the semantics of the function" tracks
closely with the number who've studied section 2 of the Unix manual.
Depending on what "studied" means, that might well be 1% of the
programming population; I'd venture, two guesses deep, that 20% is
generous. Otherwise Haskell would be popular and multithreading never
would have been invented.
Look back again at his requirements. He's not keeping games scores for
an address book on an iPhone. He won't have (I suppose) millions of
users and billions of devices. What he will have, SQLite permitting,
is a single, vital, rock-solid application. If I were in his position,
I'd be asking the same questions. I could only hope to use the same
degree of tact.
By the way, they're good questions. If answered in the documentation,
the lives of those millions of programmers will be improved, whether or
not they notice.
A productive answer to his query would be, gee, a table or three.
functions
---------
function
error
errors
------
error
type
error_types
----------
type
description
That might work. AFAIK an error means the same thing no matter which
function it is returned by or what global configuration is in force. If
that proves not to be the case, the functions table would need to be
more elaborate.
By way of example, cf.
http://technet.microsoft.com/en-us/library/aa937495(v=sql.80).aspx.
DB-Library associates a severity with each error. The notation
"DBPROCESS is dead" indicates the connection handle cannot be used
anymore. The programmer knows, per the severity, how to proceed with
each error.
--jkl
_______________________________________________
sqlite-users mailing list
[email protected]
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
