[
https://issues.apache.org/jira/browse/LUCENE-5781?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14039907#comment-14039907
]
Michael McCandless commented on LUCENE-5781:
--------------------------------------------
bq. Exactly: I am just very suspicious that the bug is in the python code.
I am too.
I measured the dead states for the N=2 case and length 6 string: 80 of 210
states are dead, and 640 out of 1336 transitions.
Also, those are just the "obviously dead" states (they cycle to themselves).
There are further non-obvious dead states: once I remove those the automaton
has 73 states (down from 130) and 279 transitions (down from 499, after
reducing the 696 transitions created by LevA).
We never noticed this before because the previous automaton impl couldn't "see"
states created that are unreachable from the initial node.
> don't create unused states in LevenshteinAutomata
> -------------------------------------------------
>
> Key: LUCENE-5781
> URL: https://issues.apache.org/jira/browse/LUCENE-5781
> Project: Lucene - Core
> Issue Type: Bug
> Reporter: Michael McCandless
> Assignee: Michael McCandless
> Fix For: 5.0, 4.10
>
> Attachments: LUCENE-5781.patch
>
>
> While working on LUCENE-5752 I discovered that LevenshteinAutomata creates
> unused states with cycles. I think they are basically harmless but we still
> shouldn't create them?
--
This message was sent by Atlassian JIRA
(v6.2#6252)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
