Lucene+4.0+的灵活索引+API

《Lucene+4.0+的灵活索引+API》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、HeavyCommitting:FlexibleIndexinginLucene4UweSchindlerSDDataSolutionsGmbH,uschindler@sd-datasolutions.deMyBackground•IamcommitterandPMCmemberofApacheLuceneandSolr.MymainfocusisondevelopmentofLuceneJava.•Implementedfastnumericalsearchandmaintainingthenewattribute-basedtextana

2、lysisAPI.WellknownasGenericsandSophisticatedBackwardsCompatibilityPoliceman.•WorkingasconsultantandsoftwarearchitectforSDDataSolutionsGmbHinBremen,Germany.ThemaintaskismaintainingPANGAEA(PublishingNetworkforGeoscientific&EnvironmentalData)whereIimplementedtheportal'sgeo-spa

3、tialretrievalfunctionswithLuceneJava.•TalksaboutLuceneatvariousinternationalconferenceslikeLuceneRevolution,thepreviousLuceneEurocon,ApacheConEU/US,BerlinBuzzwords,andvariouslocalmeetups.3Agenda•Motivation•APIchangesingeneral•Codecs•Future•Wrapup4Luceneuptoversion3.4•Lucene

4、started>10yearsago•Lucene’svIntformatisoldandnotasfriendlyasnewcompressionalgorithmstoCPU’soptimizers(existssinceLucene1.0)•It’shardtoaddadditionalstatisticsforscoringtotheindex•IRresearchersdon’tuseLucenetotryoutnewalgorithms•Smallchangestoindexformatareoftenhugepatchescov

5、eringtonsoffiles•ExamplefromdaysofLucene2.4:finalomit-TFaPpatchofLUCENE-1340is~70KiBcovering~25files5WhyFlexibleIndexing?•ManynewcompressionapproachesreadytobeimplementedforLucene•Separateindexencodingfromterms/postingsenumerationAPI•Makeinnovationstothepostingsformateasier

6、TargetstomakeLuceneextensibleevenonthelowestlevel6Agenda•Motivation•APIchangesingeneral•Codecs•Future•Wrapup7QuickOverview•WillbeApacheLucene≥4.0only!•Allowsto•storenewinformationintotheindex•changethewayexistinginformationisstored•Underheavydevelopment•almoststableAPI,mayb

7、reaksuddenly•lotsofclassesstillinrefactoring•Replacesalotofexistingclassesandinterfaces→Lucene4.0willnotbebackwardscompatible(APIwise)8QuickOverview•Pre-4.0indexesarestillreadable,but<3.0isnolongersupported•IndexupgradetoolisavailablesinceLucene3.2•Supportsupgradeofpre-3.0i

8、ndexesin-place→two-stepmigrationto4.0(LUCENE-3082)9Architecture10New4-dimensionalE