[LUCENE-3971] MappingCharFilter rarely has wrong correctOffset (for finalOffset)

• Type: Bug
• Status: Patch Available
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 4.0-ALPHA, 3.6.1
• Component/s: modules/analysis
• Lucene Fields: New
• Assignee: Dawid Weiss
• Reporter: Robert Muir
• Created: 11/Apr/12 14:03
• Updated: 15/Sep/24 22:24
• Resolved: 14/Apr/12 07:31

Description

Found this bug over on LUCENE-3969, but I'm currently tracking a ton of bugs, so
I figure I would open an issue and see if this one is obvious to anyone:

Consider this input string: "gzw f quaxot" (length = 12) with a WhitespaceTokenizer.
If i have mapping rules like this, then it works!:

"t" => ""

But if I have mapping rules like this:

"t" => ""
"tmakdbl" => "c"

Then it will compute final offset wrong:

    [junit] junit.framework.AssertionFailedError: finalOffset  expected:<12> but was:<11>

Looks like some logic/recursion bug in the correctOffset method? The second rule is not even "used" for this string,
it just happens to also start with 't'

Issue Links

• relates to: LUCENE-3830 MappingCharFilter could be improved by switching to an FST.