Add column_dtypes to forcibly convert the type

[grieve54706]

Description

We will convert types in the data frame by input dtypes.
We also format datetime to a formatted string.

HTTP example

POST /v2/ibis/postgres/query
{
  "connectionInfo": {...},
  "manifestStr": "base64 string of manifest",
  "sql": 'SELECT * FROM "Orders" LIMIT 1',
  "columnDtypes": {
      "totalprice": "float",
      "orderdate": "datetime64",
      "timestamp": "datetime64",
      "timestamptz": "datetime64"
  }
}

Response difference

{
    "columns": [
        "orderkey",
        "custkey",
        "orderstatus",
        "totalprice",
        "orderdate",
        "order_cust_key",
        "timestamp",
        "timestamptz"
    ],
    "data": [
        [
            1,
            370,
            "O",
            "172799.49",
-           820540800000,
+           "1996-01-02 00:00:00.000000",
            "1_370",
-           1704153599000,
+           "2024-01-01 23:59:59.000000",
-           1704153599000
+           "2024-01-01 23:59:59.000000 UTC"
        ]
    ],
    "dtypes": {
        "orderkey": "int32",
        "custkey": "int32",
        "orderstatus": "object",
-       "totalprice": "object",
+       "totalprice": "float64",
        "orderdate": "object",
        "order_cust_key": "object",
-       "timestamp": "datetime64[ns]",
+       "timestamp": "object",
-       "timestamptz": "datetime64[ns, UTC]"
+       "timestamptz": "object"
    }
}

Additional information

Pandas dtypes https://pandas.pydata.org/docs/user_guide/basics.html#dtypes

[goldmedal]

Thanks @grieve54706. I have one question for this.

[goldmedal]

Curiously, why are those object after setting the specific column type?

[grieve54706]

Because we format the datetime to string. String in the dtype is objcet. The data type is mapping the data value. If we want to keep the type is datetime, we should not format the value. But we need the value format.

[goldmedal]

I see. Thanks. It's good for now.

[goldmedal]

LGTM