Re: XMLSerialize: version and explicit XML declaration

Hi Tom

On 25.09.24 18:02, Tom Lane wrote:
> AFAICS, all we do with an embedded XML version string is pass it to
> libxml2's xmlNewDoc(), which is the authority on whether it means
> anything. I'd be inclined to do the same here.

Thanks. I used xml_is_document(), which calls xmlNewDoc(), to check if
the returned document is valid or not. It then decides if an unexpected
version deserves an error or just a warning.

Attached v1 with the first attempt to implement these features.

==== INCLUDING / EXCLUDING XMLDECLARATION (SQL/XML X078) ====

The flags INCLUDING XMLDECLARATION and EXCLUDING XMLDECLARATION include
or remove the XML declaration in the XMLSerialize output of the given
DOCUMENT or CONTENT, respectively.

SELECT
  xmlserialize(
    DOCUMENT '<foo><bar>42</bar></foo>'::xml AS text
    INCLUDING XMLDECLARATION);

                         xmlserialize
---------------------------------------------------------------
 <?xml version="1.0" encoding="UTF8"?><foo><bar>42</bar></foo>
(1 row)

SELECT
  xmlserialize(
    DOCUMENT '<?xml version="1.0"
encoding="UTF-8"?><foo><bar>42</bar></foo>'::xml AS text
    EXCLUDING XMLDECLARATION);

       xmlserialize
--------------------------
 <foo><bar>42</bar></foo>
(1 row)

If omitted, the output will contain an XML declaration only if the given
XML value had one.

SELECT
  xmlserialize(
    DOCUMENT '<?xml version="1.0"
encoding="UTF-8"?><foo><bar>42</bar></foo>'::xml AS text);

                          xmlserialize                          
----------------------------------------------------------------
 <?xml version="1.0" encoding="UTF-8"?><foo><bar>42</bar></foo>
(1 row)

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

==== VERSION (SQL/XML X076)====

VERSION can be used to specify the version in the XML declaration of the
serialized DOCUMENT or CONTENT.

SELECT
  xmlserialize(
    DOCUMENT '<foo><bar>42</bar></foo>'::xml AS text
    VERSION '1.0'
    INCLUDING XMLDECLARATION);
    
                         xmlserialize                          
---------------------------------------------------------------
 <?xml version="1.0" encoding="UTF8"?><foo><bar>42</bar></foo>
(1 row)

In case of XML values of type DOCUMENT, the version will be validated by
libxml2's xmlNewDoc(), which will raise an error for invalid
versions or a warning for unsupported ones. For CONTENT values no
validation is performed.

SELECT
  xmlserialize(
    DOCUMENT '<foo><bar>42</bar></foo>'::xml AS text
    VERSION '1.1'
    INCLUDING XMLDECLARATION);
    
WARNING:  line 1: Unsupported version '1.1'
<?xml version="1.1" encoding="UTF8"?><foo><bar>42</bar></foo>
                   ^
                         xmlserialize
---------------------------------------------------------------
 <?xml version="1.1" encoding="UTF8"?><foo><bar>42</bar></foo>
(1 row)

SELECT
  xmlserialize(
    DOCUMENT '<foo><bar>42</bar></foo>'::xml AS text
    VERSION '2.0'
    INCLUDING XMLDECLARATION);

ERROR:  Invalid XML declaration: VERSION '2.0'

SELECT
  xmlserialize(
    CONTENT '<foo><bar>42</bar></foo>'::xml AS text
    VERSION '2.0'
    INCLUDING XMLDECLARATION);

                         xmlserialize
---------------------------------------------------------------
 <?xml version="2.0" encoding="UTF8"?><foo><bar>42</bar></foo>
(1 row)

This option is ignored if the XML value had no XML declaration and
INCLUDING XMLDECLARATION was not used.

SELECT
  xmlserialize(
    CONTENT '<foo><bar>42</bar></foo>'::xml AS text
    VERSION '1111');

       xmlserialize
--------------------------
 <foo><bar>42</bar></foo>
(1 row)

Best, Jim