Skip to content

Latest commit

 

History

History
87 lines (61 loc) · 3.51 KB

README.md

File metadata and controls

87 lines (61 loc) · 3.51 KB
Raw

umich_solr_library_filters

This package contains a number of solr analysis filters to transform library data at index and/or query time in a solr index.

By putting transformations into your solr install, you can guarantee that the same changes happen at both index and query time, so the same input will always match. Otherwise, you're stuck trying to make sure you're using identical normailzation routines in your indexing code and your search-collecting application code, which can be problematic.

For an example of a configuration that uses these, check out my solr6 config.

Using and/or Building

To use the .jar in releases, just stick it into a place where solr will find it, and create appropriate fieldType definitions, as shown below.

The jarfile provided is built on/against solr 6.x. To build it youself, simply check out this repo inside of the solr/contrib/ directory and run (from solr/) ant dist-contrib. You'll still need to copy the jarfile to wherever you need it.

LCCN Normalizer

The LCCNNormalizer will, as you might expect, normalize LCCNs (typically found in MARC field 010) according to the algorithm on the Library of Congress site.

This filter presumes that whatever you're sending it is supposed to be an LCCN; if you give it something else, what comes out the other end may or may not make any sense. (This is primarily because it's hard to just look at a string and determine if it's supposed to be an LCCN).

	<fieldType name="lccn_normalizer" class="solr.TextField">
		<analyzer>
			<tokenizer class="solr.KeywordTokenizerFactory"/>
			<filter class="edu.umich.lib.solr_filters.LCCNNormalizerFilterFactory"/>
		</analyzer>
	</fieldType>

ISBN Normalizer

The ISBNNormalizer will take a token, attempt to extract something that looks like a valid ISBN out of it, turn it into an ISBN-13 if need be, and index the resulting 13-digit string.

Anything that doesn't seem to contain an ISBN will not index anything at all, so if you want to be able to look for random strings that come out of your 020 fields, you'll need to use an additional field for that.

Note how we limit to strings of exactly length 13, since that's all we should get.

<!-- In this example, allow multiple comma- or semi-colon-separated values -->
	<fieldType name="isbn" class="solr.TextField">
	  <analyzer>
	    <tokenizer class="solr.PatternTokenizerFactory" pattern="[;,]\s*" />
	    <filter class="edu.umich.lib.solr_filters.ISBNNormalizerFilterFactory"/>
	    <filter class="solr.RemoveDuplicatesTokenFilterFactory"/>
            <filter class="solr.LengthFilterFactory" min="13" max="13" />
          </analyzer>
	</fieldType>

LC Shelf Key

This is a simple wrapper around the solrmarc code for generating a sortable shelfkey from an LC Classification Number. It's useful as a sort key and (at least a bit) for string matching, although that last part is notoriously iffy.

You can see the original solrmarc-centric examples in the code; I'm just wrapping it up as a filter.

	<fieldType name="lc_callnumber_sortable" class="solr.TextField">
		<analyzer>
			<tokenizer class="solr.KeywordTokenizerFactory"/>
			<filter class="edu.umich.lib.solr_filters.LCCallNumberShelfKeyFilterFactory"/>
		</analyzer>
	</fieldType>