Re: Inconsistency in extended-query-protocol logging

"Guillaume Smet" <guillaume.smet@gmail.com> writes:
> IMHO, it's more logical to remove it as the text after statement: is
> not a statement in the extended query protocol case.

Well, I was never particularly enamored of the idea that we should put
statement: into everything generated by log_statement.  My proposal
would be to generate
statement: querystring            Simple Queryparse <stmt>: querystring        Parsebind <stmt>/<portal>: querystring
Bindexecute <stmt>/<portal>: querystring    Execute
 

or these prefixed with "duration: xxx", as appropriate.  Bruce was
pretty hot about having statement: in there, so the hard part might
be to convince him.

Also, the current code distinguishes a "fetch" from an
already-partially-executed portal ... do you care about that?
        regards, tom lane