Re: extended stats objects are the only thing written like "%s"."%s"

On 30.08.21 20:06, Alvaro Herrera wrote:
> On 2021-Aug-28, Tom Lane wrote:
> 
>> I think what we are doing there is following the message style
>> guideline that says to put double quotes around inserted strings.
>> In this case schema.object (as a whole) is the inserted string.
>> People often confuse this with SQL double-quoted identifiers, but it
>> has nothing whatsoever to do with SQL's rules.  (It's easier to make
>> sense of this rule in translations where the quote marks are not
>> ASCII double-quotes ... like your example with «nice».)
>>
>> In short: Justin is right, this should not be done this way.
> 
> I don't agree with the way we're applying the message guidelines here,
> but since this is the only place where we do this, I've changed it to
> the idiomatic way of quoting names.

I agree that the current situation is not satisfactory.  We should think 
about extending the guidelines to cover this.

Note that it's not necessarily enough to say, leave \"%s\".\"%s\" 
untranslated.  For example, this could create inconsistencies with 
analogous messages that don't include a schema qualification.  Also, 
unless we are being careful about escaping double-quoted strings inside 
the substituted strings, it wouldn't be entirely correct either.

A comprehensive approach across the tree would be preferable, perhaps 
with additional APIs to support it.  Also, the question when schema 
qualifications should be printed or not should be answered.