💰 The Comparison Operator assumes elementwise comparisons

[joocer]

trafficstars

We spend time and memory building lists to compare against, and then discard all except the first entry.

Example:

    elif operator == "NotLike":
        # MODIFIED FOR OPTERYX - see comment above
        _check_type("NOT LIKE", identifier_type, (TOKEN_TYPES.VARCHAR))
        matches = compute.match_like(arr, value[0])
        matches = compute.fill_null(matches, True)
        return numpy.invert(matches)

The value list only uses the first item. LIKE, ILIKE etc are never elementwise, so we don't need to waste time and memory building the list.

[joocer]

I think the solution should be for the expression engine to not expand LITERALS to the number of rows in the table, this will have two consequences:

• if it is being used to add a column to a table, this should detect scalars and expand them
• if being used in functions that are elementwise, this will need to expand the list