Re: Rationalizing EXPLAIN VERBOSE output

Gavin Sherry wrote:
> On Sun, 10 Mar 2002, Bruce Momjian wrote:
> 
> > Tom Lane wrote:
> 
> > > Seems kinda ugly.  But maybe same idea with repeated VERBOSE,
> > > a la some Unix commands ("more -v's get you more detail"):
> > > 
> > >     EXPLAIN [ANALYZE] [VERBOSE] [ VERBOSE ... ] statement;
> > > 
> 
> > 
> > I was never a fan of the -v -v more-verbose options, and I don't see any
> > case where we use such behavior in our code.  We do use detail levels
> > for debug, and that is fairly common.
> 
> I agree. This is fine under Unix, but command arguments are not really a
> grammar. Yacc doesn't enjoy terminal repetition and for good reason: it
> usually suggests a clumsy grammar. 
> 
> Personally, I think that Tom's code should go into standard EXPLAIN.

I am confused.  Which grammar do you like?

> As for how to returning explain data as a SELECT. I think I prefer
> Oracle's idea of output tables with a Postgres twist. EXPLAIN could then
> be something like:
> 
> EXPLAIN [VERBOSE] [SET ID='...' ] [INTO [TEMP] <table>] <query>
> 
> If 'table' exists, EXPLAIN would check if it is a valid explain output
> table (correct attr names, types) and if so insert the results of explain,
> one tuple per line of output. ID would be a text identifier of the output.
> 
> If the table didn't exist, it would be created. TEMP means that the table
> is removed at the end of the session.
> 
> Is this overkill?

That was my initial reaction.  :-)

--  Bruce Momjian                        |  http://candle.pha.pa.us pgman@candle.pha.pa.us               |  (610)
853-3000+  If your life is a hard drive,     |  830 Blythe Avenue +  Christ can be your backup.        |  Drexel Hill,
Pennsylvania19026