Solr has a lot of extensive features

1
Solr Integration and Enhancements

• Solr has a lot of extensive features

Todd Hatcher
2
What is Solr?

• Solr offers advanced, optimized, scalable
  searching capabilities
• Communicate with Solr using XML, JSON and HTTP
• Includes a HTML admin interface
• Solr is built on top of Lucene
• Rich features of Lucene can be leveraged when
  using Solr
• Solr is very configurable

3
Integration with ColdFusion

• Very little direct integration with ColdFusion
• ColdFusion communicates with Solr using HTTP
• Solr runs in its own JVM, does not share with
  ColdFusion
• Using ColdFusion installation, Solr runs in a
  jetty servlet container on port 8983
  (http//localhost8983/solr)
• Solr is exposed in production by default
• Important files located C\ColdFusion9\solr\multic
  ore
• Solr offers a lot more than what is available
  using cfindex cfcollection cfsearch

4
Solr

• What is a core? its like a verity collection
  (a searchable data group)
• Single Core (one index) vs Multicore (multiple
  isolated configurations/schemas/indexes using
  same Solr instance)
• C\ColdFusion9\solr\multicore\solr.xml is the
  central file that points to locations of the Solr
  cores configuration and data (this what CF
  administrator reads/writes to when creating and
  using Solr collections)
• You can put your Solr cores under you project
  directory and keep them in source control

5
core/conf/solrconfig.xml

• Main configuration for solr core
• ltqueryResponseWriter namejson /gt determines
  the format of the results. ColdFusion uses xslt
  by default
• You can return JSON, XML, python, ruby, php
• Multiple query response writers can be
  configured, one can be set as default others can
  be specified by passing parameter wtname (eg.
  wtjson)
• cfsearch type of methods will not work if the
  response writer is not what ColdFusion is
  expecting

6
core/conf/schema.xml

• Field Types maps custom types to the solr/lucene
  type
• type solr.TextField allows for analyzers
• Analyzers can be run at index time or query time
• They allow for manipulations of the data
  (typically filtering)
• The order in which filters are declared is the
  order processed
• StopFilterFactory removes common words that do
  not help the search results
• WordDelimiterFilterFactory can adds words like
  WiFi, Wi, Fi by splitting the original into
  subwords

7
core/conf/schema.xml cont.

• EnglishPorterFilterFactory determines root word
  using word variations like -ing determines root
  word and adds to index
• SynonymFilterFactory treats words as same
• DoubleMetaphoneFilterFactory for phonetic logic
  (better than Soundex which Verity uses)
• TextSpell/TextSpellPhrase feedback did you mean
• ltcopyField sourcefieldName destd/gt dest
  fieldtype can run different analyzers on source
  field and store result
• wiki.apache.org/solr/AnalyzersTokenizersTokenFilte
  rs
• Adobe adds quite a bit to the file to create
  fieldtypes to be compatible with what was in
  verity

8
core/conf/schema.xml cont.

• Similar to creating a database table. Maps field
  names to types using ltfield /gt
• Gives you the ability to store additional data
• Field can be indexed (searchable)
• Field can be stored (referenced and returned with
  results)
• Field can be required
• ltuniqueKeygtfield namelt/uniqueKeygt
• ltsolrQueryParse defaultOperatorOR /gt

9
Indexing

• Data is sent using api - HTTP POST to Solr as
  XML/JSON/Binary
• Commit is an intensive task. Do bulk adds first
  then call commit
• ltcfindex /gt calls commit after each index
  (confirmed?)
• Commit after each would noticeably increase index
  time
• Efficient Process add data (queue), commit,
  optimize

10
Search Syntax

• fieldterm ( returns everything)
• A score is generated at query time, the value
  itself doesnt have any meaning, the scores are
  relevant only when relative to each other (a
  scale)
• fq can filter query based on some supplied
  condition
• wt is the return type of the results (xml,json,
  etc.)
• qt is the request handler used to process the
  request (default is standard)
• fl is the list of fields to return (field must be
  stored)
• q is the query string
• You can specify the start value and maxrows

11
DisMaxRequestHandler

• Declared in solrconfig.xml
• Allows simplified searching without strict syntax
• Can be configured with default weighted
  parameters (which can be overriden)
• Causes the q parameters to be parsed differently

12
Resources

• Lucene In Action
• http//wiki.apache.org/solr/
• http//cfadminsearcher.riaforge.org/
• http//cfsolrlib.riaforge.org/
• CF Solr Lib written by Shannon Hicks Wrapper
  for Solr functionality