That doesn't really matter if you're not using it as an expression, but the last statement in a lambda always is.

What would be the best way here? Two separate lambdas?

if (isExisting) {
            context.database.use {
                update(RadarItem.TABLE, values, "id = ?", arrayOf(radar.id.toString()))
            }
        } else {
            context.database.use {
                insert(RadarItem.TABLE, RadarItem.C_TITLE, values)
            }
        }

Looks pretty bad though

Well there's no real problem, it's just a triggerhappy idea warning.

There is a suppression flag, but I ended up with casting the update value to long.

update(RadarItem.TABLE, values, "id = ?", arrayOf(radar.id.toString())).toLong()

Right, problem is that it is used: it's the return value of the lambda.

but if the lambda result is not assigned to anything? (am I understanding correctly, that the lambda is the use{} block?)

Well of course we know it isn't used, but that would require the compiler to look into the

use

function and even further to the surrounding code, which it currently doesn't do.