[0.9.4.2] for WHERE time, Operator 'OR' behaves as 'AND'

[madshall]

The query:

select * from "application.heartbeat.count" WHERE time > '2015-10-10' LIMIT 10

returns:

time                            value
2015-10-13T13:33:46.257Z        1
2015-10-13T13:33:46.349Z        1
2015-10-13T13:33:46.611Z        1
2015-10-13T13:33:51.32Z         1
2015-10-13T13:33:51.415Z        1
2015-10-13T13:33:51.677Z        1
2015-10-13T13:33:56.382Z        1
2015-10-13T13:33:56.484Z        1
2015-10-13T13:33:56.742Z        1
2015-10-13T13:34:01.446Z        1

The query:

select * from "application.heartbeat.count" WHERE time < '2015-10-07' GROUP BY * LIMIT 10

returns:

time                            value
2015-10-06T17:44:38.689Z        1
2015-10-06T17:44:43.784Z        1
2015-10-06T17:44:48.877Z        1
2015-10-06T17:44:53.96Z         1
2015-10-06T17:44:59.036Z        1
2015-10-06T17:45:04.113Z        1
2015-10-06T17:45:09.188Z        1
2015-10-06T17:45:14.263Z        1
2015-10-06T17:45:19.334Z        1
2015-10-06T17:45:24.408Z        1

But the query:

select * from "application.heartbeat.count" WHERE time > '2015-10-07' OR time < '2015-10-10' GROUP BY * ORDER BY time DESC LIMIT 10

returns an intersection and not a conjunction:

time                            value
2015-10-09T20:55:21.206Z        1
2015-10-09T20:55:16.144Z        1
2015-10-09T20:55:11.08Z         1
2015-10-09T20:55:06.541Z        1
2015-10-09T20:55:01.477Z        1
2015-10-09T20:54:56.526Z        1
2015-10-09T20:54:51.463Z        1
2015-10-09T20:54:46.402Z        1
2015-10-09T20:54:41.343Z        1
2015-10-09T20:54:36.285Z        1

[beckettsean]

@madshall what version of InfluxDB are you running?

[madshall]

@beckettsean I'm running the latest 0.9.4.2 build.

[fg2it]

@madshall your query is a bit strange. From https://influxdb.com/docs/v0.9/query_language/functions.html it is expected that you use group by with an aggregate function. In addition, you are also expected to have a lower bound on time, which is not the case with your OR condition.
By the way, what is supposed to return a group by * statement ?

[madshall]

@fg2it that group by doesn't matter, I might leave it by mistake when compiled an example. Regarding your another question, GROUP BY * aggregates data by all tags.

[beckettsean]

@fg2it GROUP BY * produces results grouped by every tag set. For example:

> select count(value) from disk_total where time > now() - 1m 
name: disk_total
----------------
time                count
2015-10-14T22:19:02.805709299Z  32

> select count(value) from disk_total where time > now() - 1m group by *
name: disk_total
tags: fstype=autofs, host=catalyst, path=/home
time                count
----                -----
2015-10-14T22:18:58.925464583Z  9


name: disk_total
tags: fstype=autofs, host=catalyst, path=/net
time                count
----                -----
2015-10-14T22:18:58.925464583Z  9


name: disk_total
tags: fstype=devfs, host=catalyst, path=/dev
time                count
----                -----
2015-10-14T22:18:58.925464583Z  9


name: disk_total
tags: fstype=hfs, host=catalyst, path=/
time                count
----                -----
2015-10-14T22:18:58.925464583Z  9

[beckettsean]

Further characterization, when using relative offsets the "winner" of the OR is the one with the smaller result set (I think).

Testing with time > now() - xx

> select count(value) from disk_total where time > now() - 12m
name: disk_total
----------------
time                count
2015-10-14T22:11:47.377489Z 336

> select count(value) from disk_total where time > now() - 1m
name: disk_total
----------------
time                count
2015-10-14T22:22:49.986010247Z  36

> select count(value) from disk_total where time > now() - 12m or time > now() - 1m
name: disk_total
----------------
time                count
2015-10-14T22:22:53.794122118Z  32

> select count(value) from disk_total where time > now() - 1m or time > now() - 12m
name: disk_total
----------------
time                count
2015-10-14T22:22:58.537766Z 36

Testing with time < now() - xx

> select count(value) from disk_total where time < now() - 50d
name: disk_total
----------------
time            count
1970-01-01T00:00:00Z    290225

> select count(value) from disk_total where time < now() - 70d
name: disk_total
----------------
time            count
1970-01-01T00:00:00Z    152505

> select count(value) from disk_total where time < now() - 70d or time < now() - 50d
name: disk_total
----------------
time            count
1970-01-01T00:00:00Z    152505

> select count(value) from disk_total where time < now() - 50d or time < now() - 70d
name: disk_total
----------------
time            count
1970-01-01T00:00:00Z    152505

Another explanation is that the smaller result pool is returned

[beckettsean]

Same appears to be true for absolute time ranges:

> select count(value) from disk_total where time > '2015-10-13 12:00:01'
name: disk_total
----------------
time                count
2015-10-13T12:00:01.000000001Z  21290

> select count(value) from disk_total where time > '2015-10-14 12:00:01'
name: disk_total
----------------
time                count
2015-10-14T12:00:01.000000001Z  5644

> select count(value) from disk_total where time > '2015-10-13 12:00:01' or time > '2015-10-14 12:00:01'
name: disk_total
----------------
time                count
2015-10-14T12:00:01.000000001Z  5636

> select count(value) from disk_total where time > '2015-10-14 12:00:01' or time > '2015-10-13 12:00:01'
name: disk_total
----------------
time                count
2015-10-14T12:00:01.000000001Z  5676

> select count(value) from disk_total where time < '2015-08-13 12:00:01'
name: disk_total
----------------
time            count
1970-01-01T00:00:00Z    181657

> select count(value) from disk_total where time < '2015-08-01 12:00:01'
name: disk_total
----------------
time            count
1970-01-01T00:00:00Z    122598

> select count(value) from disk_total where time < '2015-08-13 12:00:01' or time < '2015-08-01 12:00:01'
name: disk_total
----------------
time            count
1970-01-01T00:00:00Z    122598

> select count(value) from disk_total where time < '2015-08-01 12:00:01' or time < '2015-08-13 12:00:01'
name: disk_total
----------------
time            count
1970-01-01T00:00:00Z    122598

[jsternberg]

This seems to be the same as #3290. The OR operator doesn't currently work as all time operations are assumed to be AND. I'm closing this in favor of #3290.