On Thu, 11 Sep 2003, Tom Lane wrote:
> People seem to be assuming that this feature needs to be impervious to
> whitespace and being adjacent to other things. I believe we could make
> it a good deal more robust if both the opening and closing markers
> (whatever they are) are required to stand alone on a line. For example
>
> {'
> }'
>
> represents the empty string, and
>
> {'
> x = 'text';
> }'
>
> represents '\tx = \'text\';'. I think the rule would need to be that
> the newline just after the opening marker, and the newline just before
> the closing marker, are not included in the resulting string literal.
>
> The point is we don't have to make case like {'xxx}' be recognized as an
> occurrence of this construct, and on the whole I think it's a lot safer
> if we don't. Keep in mind that the string you are trying to quote is
> often code in a language that is not SQL. Therefore, choosing markers
> on the grounds that they won't appear in SQL is not good enough.
Agreed. But I think it would be very nice to not invent yet another block
quoting mechanism if possible. I don't understand the resistance to here
documents, since they solve the problem Tom brought up -- the token will
by definition never be a problem because it's alone on a line, and the
user chooses it and can avoid problematic ones. Or to combine COPY style
(but not its under-the-hood operation) with here documents, why not:
INSERT INTO sometable (5, <<\.
a
very long
string
\.
);
Is there a reason not to use here documents?
Jon