We had a soft update to RedLightGreen last week -- the afternoon of 22 March. One of the changes was to switch from using the Recommind MindServer for the ISBN search to using a DB2 SQL query.
This means intercepting every query to see if it matches an ISBN pattern. We're using a loose definition that specifies that, from the keyword search box, a single token that may or may not have hyphens, of 10 or 13 characters long not counting any hyphens. If the token is 10 char long, the first 9 char must be digits; the 10th may be an x or digits. If the token is 13 char long -- all char must be digits.
When there's an ISBN match, now, the system displays the edition view of the specific edition -- not the default edition for that title-cluster. Ideally, there's only one match to an ISBN, so the system now directly takes the user to the matching edition.
But it's not an ideal world. One case is where we have two edition records for the same ISBN in the same work. This can occur because one record was entered with a misspelling or abbreviation of a publisher's name, and it appeared algorithmically to be a different edition. In that case, the system displays one of the two records, preferring the most recent if there is any difference in the dates.
And then there was the report from the programmer, "I picked 10000 records in [the identifier] table: 192 [ISBNs] have more than one edition, some editions also belong to different [title clusters]."
How could they belong to different title clusters? An occasional typo, perhaps? (I would have thought the presence of the checksum digit would fix that.) It turns out that in some cases, too, ISBNs are issued for series.
Ah.
Notes:
Title: The Australian encyclopaedia of forms and precedents. Volume 3, Companies. -- ISBN for both the set and each individual volume.
Six Chinese titles 9620407725 -- 22 in Eureka -- appear to be a series ISBN. (962 is the ISBN country code for Hong Kong.)
==Comments closed due to spam. Please contact tunabreath at the grey-cat domain with your comments and they will be posted ==
Posted by judielaine at March 30, 2004 05:47 PM | TrackBack