Question

I have a query like this:

select count(*) from people where text_field is not null;

It’s slow. Is it kosher to put an index on mediumtext field? What’s the best practice for this?

Answer 1

This is not null query predicate is notorious for being impossible to index. That’s because null comparisons in SQL work non-intuitively.

Your best bet is to figure out some other way to find these non-empty text_field values. If this query needs to run often, you’ll need to store some kind of flag value.

For example, add an INT column called text_field_length and populate it with the length of the text_field, then select where text_field_length > 0

Edit eight years later

Newer versions of MySQL (5.7 +) and MariaDB now have computed columns. Those computed columns may be indexed.

So you may add a computed column, maybe called text_field_not_null, to your people table, and an index on that column.

ALTER TABLE people add column text_field_not_null INT
        AS (IF(text_field IS NULL,0,1)) STORED;
CREATE INDEX text_field_not_null_dex ON people(text_field_not_null);

Then, this query will exploit that new index.

 SELECT COUNT(*) FROM people WHERE text_field_not_null = 1;

And, the value of this new column gets handled automatically.

(MySQL has almost caught up to the pricey table servers in this respect.)