https://github.com/apache/jena/pull/820 makes a change to SPARQL queries
in the way that \u and \U escapes are handled.
\U is like \u but with 8 hex.
It applies to Syntax.ARQ ("with extensions") which is the default in
jena-arq and also used in Fuseki.
Syntax.SPARQL_11 remains untouched.
See discussion:
w3c/rdf-tests#67
w3c/sparql-12#77
tl;dr:
SPARQL and Turtle differ in the way they handle \u and \U escapes.
In SPARQL, they can appear anywhere (even for special characters like
"{" or "#" or in keywords "SEL\u0045CT") because it happens in the input
stream before parsing. A javacc feature that isn't present in other
parser tools.
ARQ has not supported \U up to now. RIOT does.
In Turtle, \u and \U can appear in strings and URIs, not in prefix
names, and nowhere else. This is a better design.
The general community consensus seems to be that the change should
happen and it was a mistake in SPARQL 1.0.
Personally, I don't recall having seen a query in the wild that uses \u
outside of strings or URIs. It is a confustication vector.
Andy
