Re: Does Codd's view of a relational database differ from that ofDate&Darwin?[M.Gittens]

From: Jan Hidders <jan.hidders_at_REMOVETHIS.pandora.be>
Date: Fri, 24 Jun 2005 00:03:11 GMT
Message-ID: <35Iue.128265$BB4.7099455@phobos.telenet-ops.be>

vc wrote:
> Jan Hidders wrote:

>>vc wrote:
>>
>>Where in the description of the semantics of DAPLEX did it say that the
>>result must be actually printed on a screen?

>
> You are sidetracking here. I did not claim that the result *must* be
> printed on the screen. My claim, so far, has been that since 'print'
> is a pure and unmitigated imperative word, the whole construct ('for
> each ... print') should be treated as having side effects, that is,
> non-declarative.

Well, well, well. It seems I am going to have to recant. Not that I completely agree with the above, but I have been studying the syntax and (formal) semantics of Daplex a bit closer, and it seems that the print function cannot only be used at the end of a query, but pretty much anywhere else in any subexpression. (Yay for orthogonality.) And that means you can have side effects during the iteration (not just afterwards, as I assumed). That ruins a substantial amount of optimization possibilities and, hence, cannot in any way be called declarative.

> The same holds true for the 'return' in XQuery (sorry I missed your
> earlier message asking why I do not think XQuery is pure declarative).

I still think that *that* is nonsense. There are operations in XQuery that have side effects, but the return clause is not one of them. The strange thing about the side effects in XQuery, by the way, is that they actually don't seem to make the language much less declarative or less functional.

Jan Hidders

Received on Thu Jun 23 2005 - 19:03:11 CDT