[Data-modeling] removal of /book/author/book_editions_published

Brian Karlak zenkat at metaweb.com
Fri Jan 30 23:35:33 UTC 2009


Hello All --

Normally, the publishing domain links authors (/book/author) to books  
(/book/written_work + /book/book), which in turn have editions (/book/ 
book_edition).

However, the /book/author type also has a property called /book/author/ 
book_editions_published, which links directly from an author to an  
edition.  This property was useful for early mass data loads of book  
editions, where clean data on reconciled titles was not always  
available.

At this point, however, this denormalization is no longer necessary,  
and seems counterproductive.  Only ~400 editions link directly to an  
author without also linking to a book.  The /book/author/ 
book_editions_published property forces user to enter data in two  
places, with the opportunity to have inconsistent data.  It also burns  
an unnecessary link primitive.

We would like to refactor /book/author to remove this  
denormalization.  Specifically:

1) The /book/author/book_editions_published property will be removed  
from the /book author schema, and its reverse property author_editor  
will be removed from /book/book_edition.  Editions will be connected  
to an author solely through the /book/book/edition.

2) New /book/book entries will be created for the ~400 editions that  
are only linked to authors via /book/author/book_editions_published.

3) After the schema modification, the current /book/author/ 
book_editions_published links will be marked for later deletion from  
OTG.

Please note that this will break any existing applications that rely  
on the /book/author/book_editions_published property.

Questions, comments and concerns all welcome, especially if you use / 
book/author/book_editions_published -- please follow up in this thread.

Brian
(aka /user/zenkat)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.freebase.com/pipermail/data-modeling/attachments/20090130/01667d94/attachment.htm 


More information about the Data-modeling mailing list