Lets say a table had a column called price defined as N(11, 4 ) Now it is
found that the number of different prices is much less then 10^11, but
something more like 10,000 ( Or more specifically much less then the limit
of VFPs integer's type ). Because this column has repeated values that are
not a foreign key is not following good normalization ( OK OK note the title
Fanatical Normalization ). Therefore the column could be changed to use an
integer F-Key to a lookup table of prices, and the bonus is that the column
would be using 4/11 of the space it did before.
Now in the real world I would only CONSIDER doing this if space and/or
network transfer speeds were a major concern. ( Yes I know PKZIP etc. could
help also ) Well as I type this I'm thinking maybe I would do it if the
number of prices was tiny <<100 so that I could use 1 or 2 character codes.
In the real world I can see the following problems:
-Makes reading the data difficult, e.g. one has to do something like a
SQL-Select to join the 2 tables.
-Too Complex for the benefit
-The savings are too small to matter
-This is taking data normalization too far, use some common sense.
I am sure other real world disadvantages can be seen.
My question is from a fanatics point of view is there anything wrong with
this suggestion?
Anthony
--- StripMime Report -- processed MIME parts ---
multipart/alternative
text/plain (text body -- kept)
text/html
---
©2004 Anthony Testi |
