[0.9.4.1] DROP SERIES with a field in the WHERE clause always drops all points in the measurement

[simcap]

I have got a series with values equal to 0. Since we decided not to push any value that is 0 in our InfluxDB, I want to clean up this serie and remove the existing values equal to 0. Here is what I did:

> select count(value) from p2p
name: p2p
---------
time      count
1970-01-01T00:00:00Z  286216

> select count(value) from p2p where value=0
name: p2p
---------
time      count
1970-01-01T00:00:00Z  113630

> drop series from p2p where value=0
> select count(value) from p2p where value=0
> select count(value) from p2p
name: p2p
---------
time      count
1970-01-01T00:00:00Z  3

I am surprised that is removed everything (the 3 entries is because the DB get populated at the same time by live clients).

Is the DROP series for a specific value supported?

[simcap]

(i have just updated the main description above)

[beckettsean]

I believe the WHERE clause for DROP SERIES only operates on tags. For it to work on fields (like value) it would have to scan every point in the series, as field values are unindexed.

[simcap]

Ok thanks for the info.

[beckettsean]

@simcap I am closing this in favor of another issue open on DROP SERIES (#4276). It appears the WHERE clause is not being properly evaluated

[beckettsean]

Actually, looking more closely these are two different issues. I'm re-opening this one to focus on the field matching not working. (I don't think it's supposed to work, but it should error or do nothing, not match all points instead.)

[beckettsean]

It appears that a WHERE clause containing a field key will match every point in the series, regardless:

> SELECT * FROM h2o_feet limit 4
name: h2o_feet
--------------
time                location    status!             water_level
1970-01-01T00:00:01.439856Z santa_monica    below three feet        2.064
1970-01-01T00:00:01.439856Z coyote_creek    between six and nine feet   8.12
1970-01-01T00:00:01.43985636Z   coyote_creek    between six and nine feet   8.005
1970-01-01T00:00:01.43985636Z   santa_monica    below three feet        2.116

> drop series from h2o_feet where water_level = 'foo'
> SELECT * FROM h2o_feet limit 4
> SELECT * FROM h2o_feet 
> 

Note that the WHERE water_level = 'foo' cannot ever be true, since water_level has floats and foo is a string. However, the DROP SERIES result suggests that the WHERE clause evaluated to true for every point in the measurement.

[beckettsean]

See also #4276

[beckettsean]

Might be related somehow to this generic WHERE field = bug: #4306