Re: Help with search engine-like behavior
Posted by:
Rick James
Date: December 18, 2010 12:05AM
"Normalization" has a different meaning in database technology.
18in, 18", eighteen, etc -> to some canonical representation is something else. And it is something important. And very difficult (or at least labor intensive).
Not every product has a color. Or a length. Or a weight. Or a pattern. Or...
Not every product description has all the information it should. (A dog collar might not list its color.)
A product might come in 7 different colors; what then? 7 products?
And with 1M products, searches can be slow. Suggest you build a product 'tree' down to some level, then throw the info at the user.
Jewelry > Necklaces > Gold > list of such.
Jewelry > Necklaces > With Stone(s) > list of such.
Pet > Dog > Collar > Flea Collar > list of such.
You might let them search for "necklace", but not for "18in". Instead, you might let them sort on length. After all, 17in, 18in, 19in might all be "good enough" for the buyer; why make the the 19in ones hard to find?
That advice goes for anything in a 'continuum' -- length, weight, color, shape, etc. OTOH, flash thumb drives come in 1G, 2G, 4G, 8G, 16G, 32G, etc sizes; this feels more 'discrete', than 'continuous'. And, if you aren't careful, 32G will sort between 2G and 4G.
Also, consider "auto complete".
You have a big task; don't try to do it all. But do try to do "enough".