I have a tables (simplified) like this:
Lender
Id (PK)
AgencyCode1
AgencyCode2
AgencyCode3...
AgencyCode20
Agency
AgencyCode
AgencyName
The Lender table is denormalized data coming from a view for reporting. I need one row in my output for each Lender and I need to join the agency codes to an agency table to get the display name for each agency.
The resulting output I am looking for in each row is:
Lend开发者_如何转开发erId, AgencyCode1, AgencyName1, ... AgencyCode20, AgencyName20
What is the pattern that strikes the best balance between performance and (developer) productivity to query something like this?
--Edit, Agency Code wasn't a primary key, as I had originally indicated.
Echoing @Martin Smith's comment, unpivot, join, and pivot would probably be the most efficient, performance wise -- and, through working out the intricacies of how to unpivot and pivot, your developers would be more expereienced and thus more productive over time.
Alternatively, if there are 20 and always 20 denormalized columns, you could just write out 20 left outer joins (under the assumption that ever Lender is not related with 20 Agencies). This is ugly code, and would require SQL to process 20 joins... but if the Agency table is small (where I'd call 8 pages/1 extent small), then the overall processing time might be relatively short. Testing would be called for, to determine which performs best.
As regards to developer productivity overall, keep track of how long everything takes (including posting to SO). When it's done, tally everything up, and use it to show how much time was wasted by your developers having to work with poorly designed tables.
精彩评论