Namespaces

Names are stored as written; their namespaces (Namespaces in XML 1.0) are resolved on demand, nothing is stored and parsing is unchanged. The namespace of a tag is declared by the nearest xmlns:prefix attribute (xmlns for an unprefixed name) on the tag or on its ancestors.

ids = coupled%find_all(name='{https://example.org/thermal}dt')   ! Clark notation: {namespace}local name
do t=1, coupled%tags_number()
   print '(A)', coupled%name(t)//' -> prefix "'//coupled%prefix(t)//'", local "'//coupled%local_name(t)// &
                '", namespace "'//coupled%namespace(t)//'"'
enddo

Method	Returns
prefix(id)	the name prefix, '' if none
local_name(id)	the name without prefix
namespace(id)	the namespace URI, '' for no namespace or an undeclared prefix

The xml prefix is bound to http://www.w3.org/XML/1998/namespace; xmlns="" undeclares the default namespace. Since a prefix cannot be declared empty, prefix(id)/='' with namespace(id)=='' means an undeclared prefix (not reported by parse).

Queries

print '(A)', 'thermal dt: '//coupled%content(id=ids(1))
t = coupled%find(path='f:coupled/f:case/f:dt', namespaces='xmlns:f="https://example.org/flow"')   ! your own prefixes
print '(A)', 'flow dt:    '//coupled%content(id=t)
t = coupled%find(name='m:conductivity', namespaces='xmlns:m="https://example.org/materials"')
print '(A)', 'k:          '//coupled%content(id=t)
print '(A,I0)', 'literal "dt" matches: ', size(coupled%find_all(name='dt'))       ! names without a map: literal

Name in find/find_all (and each path segment)	Matches
p:local, local	tags with this exact name (prefix included), whatever their namespace
{uri}local	tags with local name local in namespace uri, whatever their prefix; {}local: in no namespace
p:local with namespaces=	tags with local name local in the namespace that namespaces declares for p (the file's prefixes do not matter); an undeclared prefix matches nothing
local with namespaces=	tags with local name local in the default namespace of namespaces (xmlns="uri"), in no namespace if it has none

A / inside braces is not a path separator, so {uri} may hold a URL. Attribute names are always matched literally.

Next: Regular expressions.