Datetime in Trino row_condition is bound as VARCHAR, causing TIMESTAMP type mismatch

[joshua-stauffer]

Discussed in #12010

^{Originally posted by hsanslavillepro July 29, 2026}

Trino row condition binds Python datetime as VARCHAR

Hi everyone,

I am using Great Expectations with Trino and a TableAsset.

When I use a Python datetime in a row_condition, the value is sent to Trino as a VARCHAR instead of a TIMESTAMP.

Environment

• Great Expectations: 1.18.2
• Python: 3.11
• Database: Trino
• Expectation: ExpectTableRowCountToBeBetween

Code

from datetime import datetime

import great_expectations as gx
from great_expectations.expectations.row_conditions import Column


min_allowed_datetime = datetime.now()

row_condition = (
    Column("date_insertion_clean")
    >= min_allowed_datetime
)

expectation = gx.expectations.ExpectTableRowCountToBeBetween(
    min_value=0,
    max_value=100,
    row_condition=row_condition,
)

The Trino column is:

date_insertion_clean TIMESTAMP(3)

Error

TrinoUserError(
    type=USER_ERROR,
    name=TYPE_MISMATCH,
    message="Cannot apply operator: timestamp(3) <= varchar(26)"
)

The generated SQL contains:

WHERE date_insertion_clean >= ?

The parameter is:

'2026-07-27T12:16:25.146982'

Although I pass a Python datetime, the parameter appears to be serialized and bound as a VARCHAR.

Expected behavior

The parameter should be bound as a SQL TIMESTAMP, or the generated SQL should use an explicit cast:

WHERE date_insertion_clean >= CAST(? AS TIMESTAMP)

Questions

1. Is this a known issue with Trino row conditions?
2. Is it fixed in Great Expectations 1.19.1 or a later version?
3. Is there a supported way to specify the SQL type or apply a CAST in a row_condition?

Thank you for your time ⌚ 🐛 🙏

[joshua-stauffer]

This bug has the same root cause as #11566. During execution (like here) it shows up as a SQL type mismatch, and on persist it becomes a JSON round-trip mismatch. I'm working on a design to address both cases. Until that design is finalized this issue is blocked, and any PRs targeting it will be closed.