Question

Is doing an exact match filter based on a text column conceptually slower than grabbing a set of rows based on a key and filtering using the programming language?

For example:

select columns from table where textcolumn='exactphrase';

vs

select columns from table where key='key';

for (results : resultset) { 
      if (resulsts.getString(textcolumn).equals(exactphrase)) { ... } }

I’m basically curious as to how MySQL (Innodb) deals with filtering text columns and what the performance pitfalls may be (if any).

Answer 1

tldr; There will be no performance difference for “finding” the record.

Since the (indexed) PK is being used then at most a single record will be returned. The server is smart enough to not perform a table scan on the text column, even if it is not indexed because of the 1-1 cardinality of the PK. (Query planners are smart.)

The differences are then:

1. The server might return a “useless” record to the client; this may waste a small amount of bandwidth¹ (and slightly more wasteful if the text isn’t required except for the test anyway), but more importantly it muddles semantics of the query.

2. The server supports different collation modes; it might therefor be case insensitive (e.g.) on the server and result in slightly different results than a client side filter.

---

¹ While very degenerate cases can be imagined, this should be take as “equivalent time” without an explicit use/performance case. It is, however, IMOHO, still sloppy to do this on the client side without further reason.