[ https://issues.apache.org/jira/browse/LUCENE-3094?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13033209#comment-13033209
]
Dawid Weiss commented on LUCENE-3094:
-------------------------------------
This looks good to me. And even if it doesn't affect performance it definitely should help
those poor souls wishing to actually understand this algorithm :)
> optimize lev automata construction
> ----------------------------------
>
> Key: LUCENE-3094
> URL: https://issues.apache.org/jira/browse/LUCENE-3094
> Project: Lucene - Java
> Issue Type: Improvement
> Reporter: Robert Muir
> Fix For: 4.0
>
> Attachments: LUCENE-3094.patch, after.png, before.png
>
>
> in our lev automata algorithm, we compute an upperbound of the maximum possible states
(not the true number), and
> create some "useless" unconnected states "floating around".
> this isn't harmful, in the original impl we did the Automaton is simply a pointer to
the initial state, and all algorithms
> traverse this list, so effectively the useless states were dropped immediately. But recently
we changed automaton to
> cache its numberedStates, and we set them here, so these useless states are being kept
around.
> it has no impact on performance, but can be really confusing if you are debugging (e.g.
toString). Thanks to Dawid Weiss
> for noticing this.
> at the same time, forcing an extra traversal is a bit scary, so i did some benchmarking
with really long strings and found
> that actually its helpful to reduce() the number of transitions (typically cuts them
in half) for these long strings, as it
> speeds up some later algorithms.
> won't see any speedup for short terms, but I think its easier to work with these simpler
automata anyway, and it eliminates
> the confusion of seeing the redundant states without slowing anything down.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org
|