MIN / MAX Time aggregation function represented as numeric

[pauldheinrichs]

Describe the bug
I've noticed when attempting to utilize a time aggregation in a few different ways i'm getting odd behavrious in each scenario.

Scenario A)

type: min
sql: time

• GraphQL: represented as numeric that breaks the query
• SQL API: Represented as numeric that always results in null
• Pre aggregations work... but not in any sort of useful functionality as i can't query the measure

Okay - that's fine let's try something else

Scenario B)

• Let's do a bit of coercion to make things work

type: time
sql: MIN(time)

• PreAggregations: no longer are applicable when requesting the measure in my query.
• Graphql: Represented as string that also returns unqueryable data
• SQL API: represented as string (this actually works)

Tested:

• ON redshift driver
• On Druid driver

No difference

Expected behavior

I would anticipate in scenario 1 to be able to query the metric
in scenario 2 i would expect the pre-aggregation to apply
Screenshots
If applicable, add screenshots to help explain your problem.
scenario A)

Scenario B)

Minimally reproducible Cube Schema
In case your bug report is data modelling related please put your minimally reproducible Cube Schema here.
You can use selects without tables in order to achieve that as follows.

cube(`test`, {
  sql: `
    SELECT 1 as userid, '2021-01-01' as signupdate
    UNION ALL
    SELECT 2 as userid, '2021-01-02' as signupdate
    UNION ALL
    SELECT 3 as userid, '2021-01-03' as signupdate
    UNION ALL
    SELECT 4 as userid, '2021-01-04' as signupdate
  `,
  description: `table`,
  // datasource: `default`,

  measures: {
    // does not work with pre-aggregations
    min_time_broken_pre_aggs: {
      sql: `MIN(signupdate)`,
      type: `time`,
    },

    // is a numeric value unqueryable
    min_time_unqueryable: {
      sql: `signupdate`,
      type: `min`,
    },
  },
 
  dimensions: {
    userid: {
      description: `id generated when user is created`,
      sql: `userid`,
      type: `number`,
      primary_key: true,
    },

    signupdate: {
      description: `date when the user is created`,
      sql: `signupdate`,
      type: `time`,
    },
  },

  preAggregations: {
    test_rollup: {
      type: 'rollup',
      measureReferences: [min_time_broken_pre_aggs],
      dimensionReferences: [userid],
      timeDimensionReference: signupdate,
      granularity: 'day',
      partitionGranularity: 'month',
      refreshKey: {
        every: '1 hour',
        incremental: true,
        updateWindow: '1 hour'
      }
    },

    test_rollup_2: {
      type: 'rollup',
      measureReferences: [min_time_unqueryable],
      dimensionReferences: [userid],
      timeDimensionReference: signupdate,
      granularity: 'day',
      partitionGranularity: 'month',
      refreshKey: {
        every: '1 hour',
        incremental: true,
        updateWindow: '1 hour'
      }
    }
  }
});

Version:
[0.35.12]

Additional context
I've noticed the same behaviour is represented by max as welll

error logs in scenario 1 where min is represented by numeric

│ 2024-04-16 13:19:05,191 WARN  [cubesql::compile::engine::df::scan] Unable to parse value as f64: invalid float literal                                                                                       │
│ 2024-04-16 13:19:05,191 WARN  [cubesql::compile::engine::df::scan] Unable to parse value as f64: invalid float literal