Richard Huxton writes:

That's a good one. The bug is actually independent of the function
definition, but you cannot easily reproduce it in psql, because psql cuts
out -- comment before sending the command to the server. Here's how one
could do it:

cmd=$(echo -e "SELECT 'hello world'\n-- SELECT 'goodbye world'\n::text;")
psql -c "$cmd"

The problem is strings of this form:

'foo'
'bar'

This is equivalent to 'foobar'. Comments are also allowed between the
parts:

'foo'
-- abc
'bar'

Still equivalent to 'foobar'. In your case it's scanning the string
similar to

'hello world'
-- SELECT 'goodbye world
'\n::text;

Hence the complain the the string is not terminated.

The bug here is that the scanner doesn't know that a newline (or end of
input) is a required as part of a -- comment. If I change the rule

comment ("--"{non_newline}*)

in scan.l to

comment ("--"{non_newline}*){newline}

then the example works. This does not cover the case of a comment at the
end of the input, but a solution shall be forthcoming.

--
Peter Eisentraut peter_e (AT) gmx (DOT) net


---------------------------(end of broadcast)---------------------------
TIP 3: if posting/reading through Usenet, please send an appropriate
subscribe-nomail command to majordomo (AT) postgresql (DOT) org so that your
message can get through to the mailing list cleanly

On Wednesday 08 October 2003 20:56, Peter Eisentraut wrote:
Well, I try ;-)

You intrigue me sir...

Ah - now if I ever new that, I'd forgotten it.

[snip]
So given the "comment" defn in scan.l, it's seeing the quote as the next token
in the input stream? (Wracks brain thinking back to compiler technologies
class in a decade a long, long way from here).

Ah - in perl you'd be looking for a pattern anchored with a $ - I see what you
mean.

Thanks for the explanation Peter.

--
Richard Huxton
Archonet Ltd

---------------------------(end of broadcast)---------------------------
TIP 1: subscribe and unsubscribe commands go to majordomo (AT) postgresql (DOT) org

Peter Eisentraut <peter_e (AT) gmx (DOT) net> writes:
I think the minimum-damage place to fix this is by requiring \n after
{comment} in the horiz_whitespace rule. As is, it's possible for
xqcat to match to a second quote that is in the body of a -- comment.

regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 6: Have you searched our list archives?

http://archives.postgresql.org

Tom Lane writes:

You mean like this?

horiz_whitespace ({horiz_space}|{comment}{newline})
whitespace_with_newline ({horiz_whitespace}*{newline}{whitespace}*)

That doesn't work. The offending comment in the example is matched to
{whitespace} after {newline} in the second rule. The {horiz_whitespace}*
is matched to empty. We could do

special_whitespace ({space}+|{comment}{newline})
horiz_whitespace ({horiz_space}|{comment})
whitespace_with_newline ({horiz_whitespace}*{newline}{special_whitespace}* )

--
Peter Eisentraut peter_e (AT) gmx (DOT) net


---------------------------(end of broadcast)---------------------------
TIP 9: the planner will ignore your desire to choose an index scan if your
joining column's datatypes do not match