Re: Table with Text values

[Rick James]

Plan A:
* foo VARCHAR(128)
* UNIQUE(foo)
Yes, the index is bulkier, but that is not a big deal.
Note that Plan A is the simplest for you. (I think that matters.)
Only works if foo is under 767 bytes.

Plan B:
* hash BINARY(20), foo VARCHAR(128)/TEXT/VARCHAR(500)/...
* UNIQUE(hash)
This "works", and, hence, may be necessary in some situations (>767).
But it takes more code. And it is a 'random' index, which does not scale well.

Plan C:
* hash INT UNSIGNED, foo ...
* INDEX(hash)
32-bits is not enough to assure UNIQUEness.
You need (hash = ... AND foo = ...)

I like Plan A. Now that you have all(?) the relevant info, you can make your own choice.