Its just an example
'30002-1003' is just an example to test if the concept works!
Now that I know how to do it, the real code is looking up longitude and latitude for any zip+4 code in the state of Georgia and of course gets passed whatever zip+4 I happen to be working with. Since there is only one longitude/latitude for the centroid any given zip+4 (at least in theory!), it does not matter that I pull the first record since it should be the only record.
In fact the source data (USPS) does have a few errors in it such that some zip+4's have two entries, but it happens very rarely and is not really a big problem. Since I can now quickly search for any given zip+4 I can now write some routines to find the double entries and use some other methods to work out which is likely to be the more accurate one. I don't know why the USPS source data, which I think they got in turn from the US census data, should have inaccuracies, but there you go.
I think this is probably not an especially efficient way to do this since I am opening and closing the database for each search, but it is many many times quicker than searching record by record through an unindexed database and works great for what I want to do.