Let's consider a table with two columns of counter type: cqlsh:ks> CREATE TABLE t(p int PRIMARY KEY, c1 COUNTER, c2 COUNTER); cqlsh:ks> UPDATE t SET c1 = c1 + 1 WHERE p = 1;
We have one row now: p | c1 | c2---+----+------ 1 | 1 | null My question is: should we treat the null counter (`c2` here) as zero? E.g. this query returns an empty result set: cqlsh:ks> SELECT * FROM t WHERE c2 = 0 ALLOW FILTERING; p | c1 | c2---+----+---- (0 rows) so NULL counters are not comparable to regular integers. Makes sense in general, but counters are specific: one could argue that since c1 := c1 + 1 ended up as 1, its previous value must have been 0 (?). On the other hand, a counter 'nullified' by DELETE should not appear as 0... Is this a desired (and designed) behavior? Do you think it won't change in the future and can be relied on? Cheers, JS
