I built up a very simple and stupid test table.
It looks like this:
>SELECT * FROM mytesttable
name: mytesttable
time Status async deppTag myTag subldubl value
---- ------ ----- ------- ----- -------- -----
1970-01-01T00:00:00Z OK true woot 30
1970-01-01T00:00:00Z OK Blub 0.1
1970-01-01T00:00:00.000000001Z BAD Blub 0.7
1970-01-01T00:00:00.001Z BAD true subl 11.11
1970-01-01T00:00:00.002Z OK true woot 10
2017-06-28T13:15:12.285171247Z OK Blub 0.5
All of the columns are tags except for value which is a field. I can verify this by looking at the series (And seeing that abviously time 0 exists twice, so the timestamps are in different series):
> SHOW SERIES
key
---
mytesttable,Status=BAD,async=true,deppTag=subl
mytesttable,Status=BAD,myTag=Blub
mytesttable,Status=OK,async=true,deppTag=woot
mytesttable,Status=OK,async=true,subldubl=woot
mytesttable,Status=OK,myTag=Blub
If I now query the entire measurement, I would expect from the REST API (And Java Client and so on), to split my response in multiple Series. The result, however, looks different. For a query like the following, I receive all of my data within one Series:
curl -G 'http://localhost:8086/query?pretty=true' --data-urlencode "db=test" --data-urlencode "q=SELECT * FROM \"mytesttable\""
{
"results": [
{
"statement_id": 0,
"series": [
{
"name": "mytesttable",
"columns": [
"time",
"Status",
"async",
"deppTag",
"myTag",
"subldubl",
"value"
],
"values": [
[
"1970-01-01T00:00:00Z",
"OK",
"true",
"woot",
null,
null,
30
],
[
"1970-01-01T00:00:00Z",
"OK",
null,
null,
"Blub",
null,
0.1
],
[
"1970-01-01T00:00:00.000000001Z",
"BAD",
null,
null,
"Blub",
null,
0.7
],
[
"1970-01-01T00:00:00.001Z",
"BAD",
"true",
"subl",
null,
null,
11.11
],
[
"1970-01-01T00:00:00.002Z",
"OK",
"true",
null,
null,
"woot",
10
],
[
"2017-06-28T13:15:12.285171247Z",
"OK",
null,
null,
"Blub",
null,
0.5
]
]
}
]
}
]
}
Obviously, no tags are set and all tags are treated in the same way as fields are treated. The result consists out of a single series, even though the data is stored in multiple series.
To make my point clear: I really like this behavior. As I already mentioned in another question, I thought it to be quite uncomfortable if my data in a single measurement wouldn’t be queryable sorted “globally” over time. So I do want to query my measurement with time ascending order even if that measurement is split up in multiple series (And not only time ascending per series).
The question is: Can I rely on getting the data from a measurement sorted over time, even if it is huge and contains a lot of series? And can I always assume that for this kind of query, I always receive one result and one series only? (And get multiple series only if I do GROUP BY or such?!)
Best regards
