[xquery-talk] BumbleBee XQuery Test Harness

Jason Hunter jhunter at xquery.com
Wed Oct 15 15:09:10 PDT 2003


Sander Bos wrote:

> I could write and test an adapter for our product (X-Hive/DB) in about
> 1.5 hours, including a little 'hacking' (turns out we read over the
> document -> doc function renaming in the spec), and query adjusting (we
> implement the later 22 August version of the spec). So congratulations
> on such an easy to use product! 

Thanks for the props!  I'm glad BumbleBee helped you discover a spec 
compliance issue in its first day.  We'll be adding August tests in the 
2003-08 directory before long.

> I will post my adapter class on this
> list once I can run it against a non-obfuscated version of Adapter.

The open Adapter classes are available in the 1.0.1 release, now online.

> With respect to our success percentage, well we have a little work to
> do.... I did look into detail into the usecases tests (also to tune the
> adapter against the test-framework). What I saw is that we had some
> problems with doc(...) as mentioned, maybe we include incorrectly at
> some points (not sure whether that's a bug), so 800.0 instead of 800

Yes, some vendors serialize an xs:double without the .0 but 
http://www.w3.org/TR/xmlschema-2/#double says this is incorrect and the 
.0 is necessary.  In the 1.0.1 release we updated a few test results 
accordingly.

BTW, you can see what's new in the CHANGES.txt file with the distribution.

> we
> do not implement import schema yet, we include default attributes when
> importing into another constructed element (very probably a bug), and
> then there were a few test failures I could not directly explain (I am
> not one of the XQuery programmers here).

I look forward to everyone reaching 100% compliance on the use cases. 
Then we'll be able to expand beyond.  The Use Cases themselves cover 
just a portion of the full specification.

If people want to contribute tests, we have a "custom" area specifically 
for this.  You'll see some tests I've written in there.

> I think there is also a bug in the test-framework that causes hundreds
> of tests to fail (596 mentions in the log), we have a lot of results
> like:
> Query: 
> <results>
>        {fn:boolean(xs:integer("-99999999999999999999"))}
>       </results>
> Actual Result: 
> <results>true</results>
> Expected Possible Result 1: 
> true
> Failure Message 1: 
> Expected node type '3' but was '1' - comparing <BumbleBee_Result ...>
> true </BumbleBee_Result> at /BumbleBee_Result[1]/text()[1] to
> <results...> at /BumbleBee_Result[1]/results[1]
> 
> I think we produce the result in the right way, so surrounded by a
> results element? But BumbleBee does not like it.

Yes, you are correct.  That also is fixed in 1.0.1.

-jh-



More information about the talk mailing list