There are a Java and a C++ reference implementation linked from the Bidi TR.
The Java one is straightforward (and slow), written so that you can read each rule in the TR and see in the source that it works as specified.
The C++ code is verified to produce the same results as the Java code.
ICU's Bidi implementation is faster than either and also verified to produce the same results.
(Doug Felt tested that after I wrote the implementation. He had also contributed to the Java reference.)
Bernard Miller wrote:
> --I did
> however ask if you could verify that no implementation of the Unicode
> bidirectional algorithm works with the Unicode 3.20 compatible “from 4 to
> null” logic (below). It should be easy for you to do this at least for ICU,
> but you were strangely silent on this question. Bidirectional users might
> like to know that there are no bugs in their <= 3.20 implementations.
This archive was generated by hypermail 2.1.2 : Fri May 31 2002 - 16:28:48 EDT