The following review has been posted through the commitfest application:
make installcheck-world: tested, passed
Implements feature: tested, passed
Spec compliant: not tested
Documentation: not tested
As the reporter of the issues raised in this email thread, I've reviewed the first patch
and contributed the second and third.
WHAT THE PATCHES DO:
xmltable-xpath-result-processing-bugfix-5.patch contains code changes correcting
a subset of the issues that were raised in this email thread.
xmltable-xmlexists-passing-mechanisms-2.patch adjusts the grammar to allow the XML
parameter passing mechanism BY VALUE as well as BY REF. Both are ignored, but
formerly BY VALUE was a syntax error, which was unintuitive considering that BY VALUE
is the passing mechanism PostgreSQL implements (XML node identities are not preserved).
xml-functions-type-docfix-1.patch conforms the documentation to reflect the changes in
this patch set and the limitations identified in this thread.
WHAT I HAVE REVIEWED:
I have applied all three patches over 18c0da8 and confirmed that installcheck-world passes
and that the code changes resolve the issues they set out to resolve.
I've made no entry for "spec compliant" because the question is moot; the spec is written
in terms of the XQuery language, types, and concepts, and these facilities in PG are
implemented on XPath 1.0, which doesn't have those. But the changes in this patch set
do make the PG behaviors more, well, closely analogous to the way the spec compliant
functions would behave.
WHAT I HAVE NOT REVIEWED:
The passing-mechanisms and docfix patches are my own work, so there should be another
reviewer who is not me. I've looked closely at the technical, SQL/XML behavior aspects already,
but a reviewer with an eye for documentation would be welcome.
I'll venture my opinion that this is ready-for-committer to the extent of my own review, but will
leave the status at needs-review for a not-me reviewer to update.