[
https://issues.apache.org/jira/browse/SOLR-1760?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13292200#comment-13292200
]
Lance Norskog commented on SOLR-1760:
-------------------------------------
This issue is obsolete.
> convert synonymsfilter to new tokenstream API
> ---------------------------------------------
>
> Key: SOLR-1760
> URL: https://issues.apache.org/jira/browse/SOLR-1760
> Project: Solr
> Issue Type: Task
> Components: Schema and Analysis
> Reporter: Robert Muir
>
> This is the other non-trival tokenstream to convert to the new API. I looked
> at this again today, and think I have a design where it will be nice and
> efficient.
> if you have ideas or are already looking at it, please comment!! I havent
> started coding and we shouldn't duplicate any efforts.
> here is my current design:
> * add a variable 'maximumContext' to SynonymMap. This is simply the maximum
> singleMatch.size(), its the maximum number of tokens lookahead that is ever
> needed.
> * save/restoreState/cloning can be minimized by using a stack (fixed array of
> maximumContext) of references to the SynonymMap submaps. This way we can
> backtrack efficiently for multiword matches without save/restoreState and
> less comparisons.
> * two queues (can be fixed arrays of maximumContext) are needed still for
> placing state objects. the first is those that have been evaluated (always
> empty in the case of !preserveOriginal), and the second is those that havent
> yet been evaluated, but are queued due to lookahead.
> i plan on coding this up soon, if you have a better idea or have started
> work, please comment.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
