Simplified query language

[FrankChen021]

{
        "dataSource": "exception-metrics",
        "type": "timeseries",
        "fields": [
          "eventCount"
        ],
        "bucketCount": 30
      }

It's too verbose. Can be simplied by one expression such as:

-- time serieis
from exception-metrics where appName = 'test' select eventCount group by timestamp
from exception-metrics select eventCount group by timestamp(30)
from exception-metrics select eventCount group by timestamp(30), exceptionClass
from exception-metrics where eventCount group by timestamp(30), exceptionClass

-- groupBy
from exception-metrics select exceptionClass, eventCount by exceptionClass order by exceptionClass asc limit 10

--list
from exception-metrics select timestamp, instanceName, thread, message order by timestamp asc

There other two alternatives or references:

1. https://prql-lang.org/
2. PromSQL

[FrankChen021]

for timeseries,

group by interval(timestamp)
group by interval(timestamp, 30) -- field = timestamp, bucke count = 30
group by interval(timestamp, 30, 60) -- field = timestamp, bucket count = 30, minimal length is 60 seconds

for group by, the grouped field can be omitted in the SELECT clause

from exception-metrics select eventCount group by exceptionClass order by exceptionClass

[FrankChen021]

The filter can be in the last of the statement so that it simplifies the processing of user input

from exception-metrics select eventCount group by exceptionClass order by exceptionClass WHERE appName = 'test'

[FrankChen021]

Other things to do

1. ResultFormat supports columar format
2. The response of timeseries data follows the general response format to be more simple and flexible
3. the zero-fil can be optional at the client side

[FrankChen021]

To simply the processing at the FE, the server can turn this expression into original JSON object format.

[FrankChen021]

https://promlabs.com/promql-cheat-sheet/

[FrankChen021]

Extension:

1. percentage
2. size
3. duration resolution Now we support auto resolution and flexible resolution by given fixed bucket count. rate(http_requests_total[5m])[30m:1m]