Re: [PATCH] Add CANONICAL option to xmlserialize

On 26.08.24 16:59, Pavel Stehule wrote:
>
> 1. what about behaviour of NO INDENT - the implementation is not too
> old, so it can be changed if we want (I think), and it is better to do
> early than too late

While checking the feasibility of removing indentation with NO INDENT I
may have found a bug in XMLSERIALIZE ... INDENT.
xmlSaveToBuffer seems to ignore elements if there are whitespaces
between them:

SELECT xmlserialize(DOCUMENT '<foo><bar>42</bar></foo>' AS text INDENT);
  xmlserialize   
-----------------
 <foo>          +
   <bar>42</bar>+
 </foo>         +
 
(1 row)

SELECT xmlserialize(DOCUMENT '<foo> <bar>42</bar> </foo>'::xml AS text
INDENT);
        xmlserialize        
----------------------------
 <foo> <bar>42</bar> </foo>+
 
(1 row)

I'll take a look at it.

Regarding removing indentation: yes, it would be possible with libxml2.
The question is if it would be right to do so.
> 2. Are we able to implement SQL/XML syntax with libxml2?
>
> 3. Are we able to implement Oracle syntax with libxml2? And there are
> benefits other than higher possible compatibility?
I guess it would be beneficial if you're migrating from oracle to
postgres - or the other way around. It certainly wouldn't hurt, but so
far I personally had little use for the oracle's extra xmlserialize
features.
>
> 4. Can there be some possible collision (functionality, syntax) with
> CANONICAL?
I couldn't find anything in the SQL/XML spec that might refer to
canonocal xml.
>
> 5. SQL/XML XMLSERIALIZE supports other target types than varchar. I
> can imagine XMLSERIALIZE with CANONICAL to bytea (then we don't need
> to force database encoding). Does it make sense? Are the results
> comparable?
|
As of pg16 bytea is not supported. Currently type| can be |character|,
|character varying|, or |text - also their other flavours like 'name'.

|

--
Jim