Re: explain root element for auto-explain

On Thu, Aug 20, 2009 at 9:30 AM, Andrew Dunstan<andrew(at)dunslane(dot)net> wrote:
>
>
> Andrew Dunstan wrote:
>>
>> Bruce Momjian wrote:
>>>
>>> Are we going to publish an XML DTD for EXPLAIN, or have we already?
>>
>> Not a DTD, but I am working on an XML Schema (DTDs are a bit yesterday).
>>
>>
>
> Here is a RelaxNG spec which people might find a bit easier to read. It has
> been autocreated by a little tool called trang, that I used on a very large
> body of explain output that I produced by mangling the regression tests

This is definitely easy to read, especially for XML.

> (and, incidentally, crashing the server in the result - I still have to
> chase that up).

Hmm.

> I have a couple of questions, however. First, in that long list of
> alternatives for a Plan node, can any of them occur more than once?

I don't think so. But I also don't think Index-Cond should be treated
specially, as you have done here.

> Second,
> we are using Item as a child of both Output and Sort-Key nodes. Are they
> really describing the same thing? And in any case, Item is a wonderfully
> non-informative name, as is Output, for that matter.

Well, I can't help Output. That's what 8.4-EXPLAIN calls it. I do
think maybe it should be ripped out of EXPLAIN (VERBOSE) and made a
separate option.

Are they really the same thing? Obviously not. I just needed a way
to make a list of scalars in XML and I picked that for want of
creativity.

> BTW - I know this requires tweaking - those xsd:NCName values will probably
> just become text, for example.

As far as I'm concerned, you're already way ahead producing something
that fits on the screen.

...Robert