With many DBMSes, we need to repeat the formula / expression in the where clause, i.e.
SELECT table_id, radians( 25 ) AS rad
FROM test_table
WHERE radians( 25 ) < 5
ORDER BY radians( 25 ) ASC
However, in this case, since the calculated column is a constant, the query itself doesn't make much sense. Was there maybe a missing part, as in say radians (25 * myColumn) or something like that ?
Edit (following info about true nature of formula etc.)
You seem disappointed, because the formula needs to be repeated... A few comments on that:
The fact that the formula needs to be explicitly spelled-out rather than aliased may make the query less readable, less fun to write etc. (more on this below), but the more important factor to consider is that the formula being used in the WHERE clause causes the DBMS to calculate this value for potentially all of the records in the underlying table!!!
This in turns hurts performance in several ways:
- SQL may not be able use some indexes, and instead have to scan the table (or parts thereof)
- if the formula is heavy, it both makes for slow response and for a less scalable server
The situation is not quite as bad if additional predicates in the WHERE clause allow SQL to filter out [a significant amount of] records that would otherwise be processed. Such additional search criteria may be driven by the application (for example in addition to this condition on radiant, the [unrelated] altitude of the location is required to be below 6,000 ft), or such criteria may be added "artificially" to help with the query (for example you may know of a rough heuristic which is insufficient to calculate the "radian" value within acceptable precision, but may yet be good enough to filter-out 70% of the records, only keeping these which have a chance of satisfying the exact range desired for the "radian".
Now a few tricks regarding the formula itself, in an attempt to make it faster:
- remember that you may not need to run 100% of the textbook formula.
I'm not sure which part of the great circle math is relevant to this radian calculation, but speaking in generic terms, some formulas include an expensive step, such as a square root extraction, a call to a trig function etc. In some cases it may be possible to simplify the formula (which has to be run for many records/values), by applying the reverse step to the other side of the predicate (which, typically, only needs to be evaluated once). For example if say the search condition predicate is "WHERE SQRT((x1-x2)^2 + (y1-y2)^2) > 5"
. Since the calculation of distance involves finding the square root (of the sum of the squared differences), one may decide to remove the square root and instead compare the results of this modified formula with the square of the distance value origninally, i.e. "WHERE ((x1-x2)^2 + (y1-y2)^2) > (5^2)"
- Depending on your SQL/DBMS system, it may be possible to implement the formula in a custom-defined function, which would make it both more efficient (because "pre-compiled", maybe written in a better language etc.) and shorter to reference, in the SQL query itself (event though it would require being listed twice, as said)
- Depending on situation, it may also be possible to alter the database schema and underlying application, to have the formula (or parts thereof) but pre-computed, and indexed, saving the DBMS this lengthy resolution of the function-based predicate.