BUG #19486: Regression in SQL-language functions using XML values and IS DOCUMENT

The following bug has been logged on the website:

Bug reference: 19486
Logged by: Artem Zarubin
Email address: a(dot)prototype7(at)gmail(dot)com
PostgreSQL version: 18.4
Operating system: Debian 13
Description:

Hello, I found a regression in SQL-language functions using XML values and
`IS DOCUMENT`.

Tested commits:

bad: 98dd6c2046965e51da015681e81c20109be46d71, PostgreSQL 18.4
bad: 5107398e6d5ecad96f3d1c0efcfc9aa02b9cdff9, PostgreSQL 19devel
good: e9e7b66044c9e3dfa76fd1599d5703acd3e4a3f5, parent of 0dca5d68
PostgreSQL 18devel before SQL-function plan cache changes

The server was configured with:

./configure --enable-tap-tests --enable-debug --enable-cassert
--with-libxml

SQL-script to reproduce:

CREATE OR REPLACE FUNCTION xml_to_text_no_inline(pXml xml) RETURNS text
LANGUAGE sql
IMMUTABLE
SET search_path = pg_catalog
AS $$
SELECT CASE WHEN pXml IS DOCUMENT
THEN (xpath('/*/text()', pXml))[1]::text
ELSE pXml::text
END;
$$;

SELECT xml_to_text_no_inline(XMLPARSE(CONTENT '2019-12-16T00:00:00.000'));

Expected result:

2019-12-16T00:00:00.000

Actual result:

ERROR: could not parse XML document
DETAIL: line 1: Start tag expected, '<' not found
2019-12-16T00:00:00.000
^
CONTEXT: SQL function "xpath" statement 1
SQL function "xml_to_text_no_inline" statement 1

---
Best regards,
Artem Zarubin
Postgres Professional: https://postgrespro.com/