Subject: Re: PostgreSQL vs SQL/XML Standards
To: Pavel Stehule <pavel.stehule@gmail.com>
References: 
 <CAFj8pRDFQJMoKf-X+3p1ASwU02nnXfAXrg83Z3fHUBd-Gwuuxw@mail.gmail.com>
 <FB592C2F-E2F3-42FD-A342-EE41BECD3771@winand.at>
 <CAFj8pRC_bi4uzE3Z-XXuGnLFPP-7cZ-zr2JhGauWm3E8K2MqEg@mail.gmail.com>
 <45C74339-3531-46BD-A595-37183C6730A1@winand.at>
 <CAFj8pRB4NHismecE-SnezKz9OektP2H8t_684TbqQ4Ax3swZwQ@mail.gmail.com>
 <5C283168.8060001@anastigmatix.net>
 <CAFj8pRDRrxyr8=2PEFxP_eZCRqJMU4Jpj0whJM9pX0V4Q4EQvw@mail.gmail.com>
 <5C291741.9060506@anastigmatix.net>
 <CAFj8pRAmSUO1g1qnVgXmE=7bPULYjwDfyO5KDohsHMFesDenmA@mail.gmail.com>
 <5C297BB7.9070509@anastigmatix.net>
 <CAFj8pRCN=0k507nTfvNQuruHvomA4SiC6uGUXY5Gz0crxf-JAQ@mail.gmail.com>
 <5C3A7A90.3070207@anastigmatix.net>
 <CAFj8pRCSa0ekwc8AOUnEsLeJhWbq7_rOyFzruopfgBTaMpUWMA@mail.gmail.com>
 <5C43FAEB.2070000@anastigmatix.net>
 <CAFj8pRDLywawDY-dRsYdjF036iR_Y4em77X1DF7VodUqau_0Vg@mail.gmail.com>
 <5C4401BE.4080606@anastigmatix.net>
 <CAFj8pRCKwDGqP55RTykDq87DXuXStE+3Qy2Hf9Wxh_A-cUnwTw@mail.gmail.com>
Cc: PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
From: Chapman Flack <chap@anastigmatix.net>
Message-ID: <5C449C71.7060903@anastigmatix.net>
Date: Sun, 20 Jan 2019 11:06:09 -0500
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101
 Thunderbird/38.7.0
MIME-Version: 1.0
In-Reply-To: 
 <CAFj8pRCKwDGqP55RTykDq87DXuXStE+3Qy2Hf9Wxh_A-cUnwTw@mail.gmail.com>
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hi,

On 01/20/19 00:26, Pavel Stehule wrote:
>>> column expressions are evaluated once per row, but XPath  expression is
>>> compiled per row too, if I remember well.
>>
>> I looked for evidence of that in the code, but did not find it; the
>> compilation appears to happen in XmlTableSetColumnFilter, which is
>> called from tfuncInitialize.
> 
> it is called per input row.

I could be misunderstanding what you mean by 'input row' here.

If I write a simple xmltable query where the row_expression produces
six rows:

SELECT *
FROM
  xmltable('pg_am/row'
    PASSING table_to_xml('pg_am', true, false, '')
    COLUMNS amname text PATH 'amname');

six rows are produced, though a breakpoint set on XmlTableSetColumnFilter
fires only once, from tfuncInitialize at the start of xmltable's execution.


By contrast, in the regression test example with PATH ''||lower(_path)||'c',
four rows are produced and the breakpoint fires four times.

However ... that isn't because one call to xmltable is producing four rows
and recomputing the column_expression each time.

It's because that xmltable is the RHS of a LATERAL, and the LHS of the
LATERAL is producing four tuples with different values of columns the RHS
depends on, so the RHS (xmltable) is being called four different times,
producing one row each time, still with XmlTableSetColumnFilter being called
only during initialization.


> sure - using any expression in PATH clause should to demonstrate this
> functionality.

Well, there seem to be two distinct features touched on in the docs:

1. The column_expression is allowed to be an expression, not restricted
   to a string literal as it is in the standard (and Oracle).

2. Not only is it an expression, but it's an expression whose evaluation
   is deferred and can happen more than once in the same xmltable call.

The example in the regression tests certainly demonstrates (1). Without (1),
it would be a syntax error.

I think the same example would produce the same output even with feature (2)
absent. It's LATERAL doing the magic there. So I am doubtful that it
demonstrates (2).

I put some effort last night into trying to even construct any query that
would demonstrate (2), and I came up short, but that could be my lack of
imagination. (Somewhere in that effort I happened to notice that xmltable
doesn't seem to be parsed successfully inside a ROWS FROM (...) construct,
which might be another issue for another time....)

So, if you have a way to build a query that demonstrates (2), my aha! moment
might then arrive.

Regards,
-Chap