The scenario:

Let’s say we have a database table ‘TABLE’ with entities identified by ‘ID’. Each entity has several features stored in columns (‘COLUMN_1’ to ‘COLUMN_N’). There is an additional feature of the entities ‘FEATURE_X’ that is not in the database.

An SQL query with a WHERE clause filters the entities based on conditions involving the columns and ‘FEATURE_X’.

Now we can’t include ‘FEATURE_X’ in the actual condition of the query that goes to the database, because that feature is not represented by a column in the database, but let’s say is determined by a very slow algorithm.

The problem:

I want to split the logical condition into two parts. One that involves only columns in order to send to the database and one condition involving ‘FEATURE_X’.

I want to construct a query to retrieve entities that form a superset of the actual result set, so that I can start the very slow algorithm with an already prefiltered set of potential result entities.

Is there a logical operation that ensures that we always get a superset regardless on how the logical expression is?

An example:

Given an expression

Φ = (COLUMN_1<10) AND ((COLUMN_2=4) OR (COLUMN_3>5) OR NOT (COLUMN_4=0 AND FEATURE_X=1))

the SQL statement would look like this:

SELECT ... FROM ... WHERE Φ

How can I exclude ‘FEATURE_X’ without compromising the correctness of the result?

My pseudo code that doesn’t work:

1. Identify Φ_database
   Φ = (COLUMN_1<10) AND ((COLUMN_2=4) OR (COLUMN_3>5) OR NOT (COLUMN_4=0 AND TRUE))
2. Identify Φ_feature_x
   Φ = FEATURE_X=1
3. Construct SQL query
   SELECT "ID" FROM "TABLE" WHERE (COLUMN_1<10) AND ((COLUMN_2=4) OR (COLUMN_3>5) OR NOT (COLUMN_4=0 AND TRUE)))
4. Use the result as an input for the slow algorithm and get the result.

The error here is, that the first sql query does not always give me back a super set but sometimes also an empty set, which means that the logical expressions in the where clause are not preserved.