Re: [HACKERS] dollar quoting

Tom Lane wrote:

>Andrew Dunstan <andrew(at)dunslane(dot)net> writes:
>
>
>>Tom Lane wrote:
>>
>>
>>>... But how about
>>>42$foo$
>>>This is a syntax error in 7.4, and we propose to redefine it as an
>>>integer literal '42' followed by a dollar-quote start symbol.
>>>
>>>
>
>
>
>>The test should not succeed anywhere in the string '42$foo$'.
>>
>>
>
>No, it won't. The problem is that it should, because the backend will
>see that as '42' followed by a $foo$ quote start.
>

Ok, I see what you are saying. This mismatch would only happen on
invalid input, though. I believe that what I did will work on all legal
input.

I think that this might be cured by having psql recognise a legal
identifier or keyword and eating it as a word, rather than treating it
as just another set of bytes in the stream. That would enable us to
avoid the lookback in the dollar-quote recognition test altogether. The
attached patch does it that way - the keyword/id test needs to come
right at the end of the loop to avoid clashing with backslash commands,
btw.

I *think* that this way psql will recognise the start of a dollar quote
iff the backend lexer would.

>
>
>
>>Interacting with lexer states would probably be ... unpleasant. Matching
>>a stream oriented lexer with a line oriented CLI would be messy I suspect.
>>
>>
>
>I think it would not be that bad. We'd have to run the lexer on the
>command input buffer and see what state it terminates in.
>
>
>

Yeah. I am not enough of a flex wizard to undertake the task, though. It
would take me lots of time. If we make a decision that we really need
this in order to do dollar quoting in psql I would need some substantial
help, at least.

cheers

andrew