Re: ('dog$house' = quote_ident('dog$house')) is surprisingly FALSE

david.g.johnston@gmail.com wrote:

bryn@yugabyte.com wrote:

(3) The PG doc on quote_ident says this in large friendly letters:

Quotes are added only if necessary…

Notice "only". I now know that this is very much not the case. You can compose an effectively unlimited number of different examples along these lines:

select quote_ident('redaktør'); → "redaktør"
create table redaktør(n int); → table successfully created

Yep, and that is precisely what would make for a good bug report. Pointing out that "if necessary" does not indeed match up with the behavior. I suspect it is likely to get changed - everything else being discussed just detracts attention from it.

*BRIEFLY*

What does "make for a good bug report" mean, David? Is it:

(1.1) You, David, or somebody else who has been officially recognized as a PG Contributor (https://www.postgresql.org/community/contributors/) will file the bug, granting it credibility with their imprimatur?

or (1.2) I, Bryn, should file the bug.

About "I suspect it is likely to get changed", do you mean:

(2.1) Change the doc to match quote_ident's current, unpredictable, behavior? (By all means, substitute "hard to describe accurately, precisely, and yet still tersely" for "unpredictable".)

(2.2) Change quote_ident's implementation—and then write new doc to describe the new behavior precisely and accurately? And for this option, the next question is "What's the spec of the changed implementation?"

Notice that the issue is broader than just quote_ident, as this test shows:

I’m not convinced. The discussion has shown that some people are somewhat confused. For example, it was suggested that a name like this:

Compare this with the implementation that I thought, at first, that I could use when I simply believed the doc. (The subject line of this thread hits at the trivial SQL statement that would implement the "language SQL" function.) ANd if that's all there is to it, then when not ship is as a built-in?