Re: Bidi reordering test results

From: Jonathan Rosenne (
Date: Sat Dec 04 1999 - 18:47:36 EST

I have compared the results with IE5. See

Tests 7, 22, 23, 24 differ with the reference implementation. I don't
understand test 7.

Test 14 has a difference around the /.


At 07:52 04/12/99 -0800, Mark Leisher wrote:
>I am including all of the message for those who might not be on the Unicode
> Asmus> At 10:08 AM 12/3/99 -0800, Mark Leisher wrote:
> >> I have a web page showing the results from tests of two (soon to be
> >> three) freely available bidi reordering packages for Unicode text.
> >> text I used for the tests as well as the test programs I wrote for all
> >> three are
> >> available as well:
> >>
> >>
> >>
> >> Please feel free to donate additional tests/code, point out problems,
> >> or make comments.
> Asmus> The really interesting question is how do these algorithms compare
> Asmus> with the official bidi reference implementations published by the
> Asmus> Unicode Consortium.
> Asmus> The Unicode sample code is availabe at
> Asmus>
> Asmus>
> Asmus> The Technical report states:
> Asmus> 1 Reference Code
> Asmus> There are two versions of BIDI reference code available. Both have
> Asmus> been tested to produce identical results. One version is
written in
> Asmus> Java, while the other is written in C++. The Java version is
> Asmus> designed to closely follow the steps of the algorithm as described
> Asmus> in this report. The C++ code is designed to show one of the
> Asmus> optimization methods that can be applied to the algorithm, using a
> Asmus> state table for one phase.
> Asmus> It is very important to note that any implementations that don't
> Asmus> match the published algorithm, as evidenced by yielding different
> Asmus> results than the reference implementations, are *not conformant*.
> Asmus> Since the whole purpose of the bidi algorithm is to allow
> Asmus> to predict the ordering of their text on the *readers'* screen,
> Asmus> adherence to the conformance requirements is crucial.
> Asmus> The published reference implementations were tested
exhaustively up
> Asmus> to sequences of length 6 against each other, and stochastically
> Asmus> much longer sequences, with special prefixes used to force the
> Asmus> implemenations to hit the maximum nesting level during some
> Asmus> of the tests.
> Asmus> I would encourage Mark to test his algorithms exhaustively against
> Asmus> the reference code and report his results here.
> Asmus> For more details on bidi and the conformance requirements read the
> Asmus> report at:
> Asmus>
>Asmus is quite right. I have used the code he pointed out to add a column of
>reference results to
>Mark Leisher
>Computing Research Lab I have never made but one prayer to God,
>New Mexico State University a very short one:
>Box 30001, Dept. 3CRL "Oh Lord, make my enemies ridiculous."

>Las Cruces, NM 88003 And God granted it. -- Voltaire, letter

This archive was generated by hypermail 2.1.2 : Tue Jul 10 2001 - 17:20:56 EDT