MARCgrep lets you find all the records in a library collection matching some criteria. It searches MARC records in the dumbest way possible: parsing them one at a time and returning any that match.
Input and output are designed to be pluggable, so it should be possible to extend MARCgrep to read your MARC records from any source and write them out in whatever format you like. So far I've got it reading records from a file or a Lucene index, and writing to either MARC21 or plain text, but I'd be happy to lend a hand if you have grand ideas (like reading records from your ILMS system).
Some examples:
-
Find records with the keyword "laugh" in the 245$a and "microform" in the 245$h
-
Find records with the string "laugh" in the 245$a (matches "laugh", "laughed", "laughing")
-
Find records where a 1XX field whose first indicator is '1' contains the keyword "Smith"
-
Find records containing more than one 600 field whose 008 field matches the regular expression "^.{35}eng"
-
Find records where the keyword "spatula" appears in any field
These sort of search criteria can be combined with the standard AND/OR/NOT boolean operators to produce more complex queries.
Fields to match against can be specified with as much detail as is useful. For example, a field specification might match:
-
Any field (*)
-
Any subfield in a 2XX field (2*)
-
Any subfield in a 245 field (245)
-
Any subfield in a 245 field with a particular indicator value (245!1#)
-
Subfields a, b, c in a 245 field with a particular indicator value (245!1#$abc)
Don't let the syntax scare you off: the simple cases are simple (just type the tag of the field you want to search), and there's inline documentation for when you need something more complicated.
Not as slow as you might think. By pre-compiling your queries into fairly efficient Java bytecode and splitting the MARC record parsing and matching across a couple of CPUs, MARCgrep's naïve approach actually performs pretty well. Limiting MARCgrep to 4 cores of a 3ghz Intel Xeon box, I can still check between ten and fifteen-thousand records per second.
Everything in MARCgrep happens on a single screen: you add one or more jobs by completing the form at the top, then you run your jobs by clicking the "Run all jobs" button. MARCgrep batches jobs together so running multiple jobs is just as fast (or slow, if you like) as running one, so there's no harm in running lots of jobs at once.
The quickest way to get running with MARCgrep is to run it straight from the command-line, using its built in web server. To do this:
-
Get Leiningen from http://github.com/technomancy/leiningen and put the 'lein' script somewhere in your $PATH.
-
From marcgrep's root directory, run 'lein uberjar'. Lein will grab all required dependencies and produce a 'target/marcgrep-1.0.4-standalone.jar'.
-
Copy resources/config.clj.example to resources/config.clj and edit as appropriate.
-
Run the jar from your MARCgrep directory, for example:
java -cp resources:target/marcgrep-1.0.4-standalone.jar marcgrep.core
-
Point your browser at http://localhost:9095/
-
Get Leiningen from http://github.com/technomancy/leiningen and put the 'lein' script somewhere in your $PATH.
-
Copy resources/config.clj.example to resources/config.clj and edit as appropriate.
-
From marcgrep's root directory, run 'lein deps' and then 'lein ring uberwar'. Lein will grab all required dependencies and produce a 'target/marcgrep-1.0.4-standalone.war'.
-
Deploy this WAR file in your servlet container of choice (Jetty, Tomcat, ...)
Parts of the MARCgrep development were funded by the National Library of Australia, who have generously contributed their modifications back to the open source community. Hurrah!
Stefano Bargioni independently spun up a project called MARCgrep.pl at around the same time I was doing this. If you're looking for a handy way of filtering a file of MARC records from the command line, this may be the tool for you.