Karl, your right, sorry. I used the current release of Solr, version 1.3.0, that's using Lucene 2.4.1. Solr 1.4 that will be released soon is using Lucene 2.9. For me, it seems that filter did not changed at the causing code lines. But I don't know, if this is the real root cause.
Now, I have tested this also with the current trunk of Solr that is already using Lucene 2.9. At first I tried a simple example with an analyzing workflow based on the WhitespaceTokenizer followed by the ShingleMatrixFilter and no problem occured.
Then, I tried the other field type configuration, that I had used at the former test and the exception was thrown.
Caused by: java.lang.IndexOutOfBoundsException: Index: 0, Size: 0
at java.util.ArrayList.RangeCheck(Unknown Source)
at java.util.ArrayList.get(Unknown Source)
To find the reason of it, I removed filter by filter. After a lot of tests, I found out that the problem was caused by the use of
that were used in that order. If I changed the positions of both filters, all seems to work okay.
This time, I tested this only with the field analysis view with different data
Also, it was really strange, that the exception only occured at the first analysis request, and extremely rarly a second time. But it was thrown at every first request.