XSLT 2.0 and XPath 2.0 Programmer's Reference, 4th Edition (535 page)

, but not (and here's the surprise) when it is
co
.

Java could instead have standardized on the composed form of the character, but the accent-blind matching would then not work:
contains(“gar
ç
on”, “c”)
would be false.

Now let's look at a case where a pair of characters represents a single collation unit. Here we turn back to Spanish, where in older publications
ch
collates after
c
and
ll
collates after
l
. We can set this up in Java by defining a
RuleBaseCollator
using a rule that defines
c
<
ch
<
d
and
l
<
ll
<
m
. (Modern Spanish practice follows the English collating rules, so I had to set up these rules myself.)