Filter by string measure doesn't work (equals operator)

[Anger79]

Describe the bug
I have a string measure (it's supported https://cube.dev/docs/reference/data-model/types-and-formats#string) in my schema and I'm trying to run a query which contains a filter by this measure ("equals" operator). Cubejs returns following error message:
Query failed: Error: Cannot parse string 'device_A' as Float64: syntax error at begin of string. Note: there are toFloat64OrZero and toFloat64OrNull functions, which returns zero/NULL instead of throwing exception.: While processing argMax(device_name, max_timestamp) = toFloat64('device_A'). ; query id: d9ca2185-8df2-4638-8470-7b919d33bd5b

Query I'm running:

{
  "dimensions": [
    "Cube.id"
  ],
  "measures": [
    "Cube.latest_device_name"
  ],
  "filters": [
    {
      "member": "Cube.latest_device_name",
      "operator": "equals",
      "values": [
        "device_A"
      ]
    }
  ]
}

Noticed also that "operator": "contains" works as expected.

Expected behavior
Filter by string measure works.

Screenshots

Minimally reproducible Cube Schema

cube(`Cube`, {
    measures: {
        latest_device_name: {
            sql: `argMax(device_name, timestamp)`,
            type: `string`
        },
    },
    dimensions: {
        device_name: {
            sql: `device_name`,
            type: `string`
        },
        id: {
            sql: `id`,
            type: `string`
        },
        timestamp: {
            sql: `timestamp`,
            type: `time`
        }
    }
});

Version:
[e.g. 1.3.16]

Additional context
Previosly I had a similar issue with time measures and it has been fixed: #9435

[igorlukanin]

I guess the database is ClickHouse, judging by argMax.

I was able to reproduce it with v1.3.23. With the following data model...

cube(`Cube`, {
  sql: `
    SELECT 1 AS id, 'device_A' AS device_name, '2025-01-01'::TIMESTAMP AS timestamp UNION ALL
    SELECT 1 AS id, 'device_B' AS device_name, '2025-01-02'::TIMESTAMP AS timestamp UNION ALL
    SELECT 2 AS id, 'device_C' AS device_name, '2025-02-10'::TIMESTAMP AS timestamp UNION ALL
    SELECT 2 AS id, 'device_D' AS device_name, '2025-02-11'::TIMESTAMP AS timestamp
  `,

    measures: {
        latest_device_name: {
            sql: `argMax(device_name, timestamp)`,
            type: `string`
        },
    },
    dimensions: {
        device_name: {
            sql: `device_name`,
            type: `string`
        },
        id: {
            sql: `id`,
            type: `string`
        },
        timestamp: {
            sql: `timestamp`,
            type: `time`
        }
    }
});

...an incorrect SQL in the HAVING clause is generated:

SELECT
  `cube`.id `cube__id`,
  argMax(device_name, timestamp) `cube__latest_device_name`
FROM
  (
    SELECT
      1 AS id,
      'device_A' AS device_name,
      '2025-01-01' :: TIMESTAMP AS timestamp
    UNION ALL
    SELECT
      1 AS id,
      'device_B' AS device_name,
      '2025-01-02' :: TIMESTAMP AS timestamp
    UNION ALL
    SELECT
      2 AS id,
      'device_C' AS device_name,
      '2025-02-10' :: TIMESTAMP AS timestamp
    UNION ALL
    SELECT
      2 AS id,
      'device_D' AS device_name,
      '2025-02-11' :: TIMESTAMP AS timestamp
  ) AS `cube`
GROUP BY
  `cube__id`
HAVING
  (argMax(device_name, timestamp) = toFloat64(?))
ORDER BY
  `cube__latest_device_name` DESC
LIMIT
  10000

Leads to the same error as shown above.

[Anger79]

@igorlukanin that's right, database is ClickHouse.