Would like the ability to use the __time column in filters (not intervals)

[vogievetsky]

I would like to issue filters on the __time column just like any other dimension column.

This is to support the usecase described here: https://groups.google.com/forum/#!topic/imply-user-group/EQ_9oi29kgg

Essentially I want to be able to compute (in PlyQL):

SUM(`revenue_in_dollars` WHERE __time < '2016-04-04T00:00:00Z') +
SUM(`revenue_in_cents` WHERE '2016-04-04T00:00:00Z' <= __time) / 100

by expressing it as two filtered aggs that I add in a postAgg.

I would like, for example, the following query to work:

{
  "queryType": "timeseries",
  "dataSource": "wikipedia",
  "intervals": "2015-09-12/2015-09-13",
  "granularity": "all",
  "context": {
    "timeout": 10000
  },
  "aggregations": [
    {
      "type": "filtered",
      "name": "!T_0",
      "filter": {
        "type": "bound",
        "dimension": "__time",
        "alphaNumeric": true,
        "upper": 1442080800000,
        "upperStrict": true
      },
      "aggregator": {
        "name": "!T_0",
        "type": "doubleSum",
        "fieldName": "rev1"
      }
    },
    {
      "type": "filtered",
      "name": "!T_1",
      "filter": {
        "type": "bound",
        "dimension": "__time",
        "alphaNumeric": true,
        "lower": 1442080800000
      },
      "aggregator": {
        "name": "!T_1",
        "type": "doubleSum",
        "fieldName": "rev2"
      }
    }
  ],
  "postAggregations": [
    {
      "type": "arithmetic",
      "fn": "+",
      "fields": [
        {
          "type": "fieldAccess",
          "fieldName": "!T_0"
        },
        {
          "type": "fieldAccess",
          "fieldName": "!T_1"
        }
      ],
      "name": "Revenue"
    }
  ]
}

[vogievetsky]

Note this is possible to achieve with now (credit @gianm for the idea) using a JS aggregator that can access __time and the two columns as is documented in detail here: https://github.com/implydata/pivot/blob/master/docs/configuration.md#switching-metric-columns

[vogievetsky]

Also note that this query:

{
  "queryType": "timeseries",
  "aggregations": [
    {
      "name": "Count",
      "type": "count"
    }
  ],
  "dataSource": "wikipedia",
  "filter": {
    "type": "javascript",
    "dimension": "__time",
    "function": "function(s) {return s === null;}"
  },
  "granularity": "all",
  "intervals": [
    "2010/2020",
  ]
}

always matches everything because __time (s) is always passed in as null