Using kind INCREMENTAL_BY_TIME_RANGE and running sqlmesh plan give 'ConfigError: Time column '' not found in model', snowflake normalisation= case_insensitive

[SophieLee-Hoareau]

SQLMesh Bug Report

Title: INCREMENTAL_BY_TIME_RANGE: Time column not found when using normalization_strategy=case_insensitive with Snowflake

Repo: https://github.com/TobikoData/sqlmesh/issues

---

Bug Description

When using INCREMENTAL_BY_TIME_RANGE models with a Snowflake connection and normalization_strategy=case_insensitive in the dialect, sqlmesh plan fails with:

ConfigError: Time column '' not found in model 'my_schema.my_model'

Root Cause

In sqlmesh/core/model/definition.py, the convert_to_time_column method (line ~831) does a case-sensitive Python dict lookup:

if self.time_column.column.name not in columns_to_types:

• normalization_strategy=case_insensitive lowercases the model's time_column.column.name → createdatutc
• engine_adapter.columns() returns Snowflake's native UPPERCASE column names → CREATEDATUTC
• The case-sensitive not in check always fails

Reproduction

config.yaml:

gateways:
  snowflake_prod:
    model_defaults:
      dialect: "tsql,normalization_strategy=case_insensitive"
    connection:
      type: snowflake
      # ...

model.sql:

MODEL (
    name my_schema.my_model,
    kind INCREMENTAL_BY_TIME_RANGE (
        time_column CreatedAtUtc
    )
);

SELECT
    src.CreatedAtUtc,
    src.Id
FROM upstream.source AS src
WHERE src.CreatedAtUtc BETWEEN @start_date AND @end_date

Run sqlmesh --gateway snowflake_prod plan → fails with Time column '' not found.

Suggested Fix

Normalize the column names from engine_adapter.columns() before comparison, or make the lookup in convert_to_time_column case-insensitive:

# Before
if self.time_column.column.name not in columns_to_types:

# After
columns_to_types_lower = {k.lower(): v for k, v in columns_to_types.items()}
if self.time_column.column.name.lower() not in columns_to_types_lower:

A more robust fix might normalize column names at the point they're returned from engine_adapter.columns() (line ~501) to match the model's normalization strategy.

Environment

• SQLMesh version: 0.230.1
• Python: 3.12
• Connection: Snowflake
• Dialect: tsql,normalization_strategy=case_insensitive

[SophieLee-Hoareau]

model example (project dialects is t-sql, normalisation=case_insensitive)

`MODEL (
name analytics_usersessions.usersession_details,
kind INCREMENTAL_BY_TIME_RANGE (
time_column CreatedAtUtc
),
tags 'usersessions',
grain UserSessionId,
cron '@daily',
audits (
NUMBER_OF_ROWS(threshold := 1),
UNIQUE_VALUES(columns := (
UserSessionId
)),
NOT_NULL(columns := (
UserSessionId
))
)
);

SELECT
us.StartedAtUtc AS CreatedAtUtc,
us.UserSessionId,
us.UserId,
us.OrganisationId,
us.Device_Brand,
us.Device_Family,
us.Device_Model,
COALESCE(SmartEnum_FormFactor.[Name], 'Unknown') AS Device_FormFactor,
us.OS_Family,
us.OS_Major,
us.OS_Minor,
us.Browser_Family,
us.Browser_Major,
us.Browser_Minor,
us.Location_City AS IpLocation_City,
us.Location_CountryCode AS IpLocation_CountryCode,
us.Location_Latitude AS IpLocation_Latitude,
us.Location_Longitude AS IpLocation_Longitude
FROM orgs.UserSessions AS us
WHERE
us.StartedAtUtc BETWEEN @start_dt AND @end_dt
LEFT JOIN [orgs].[SmartEnum_FormFactor] AS SmartEnum_FormFactor
ON us.Device_FormFactor = SmartEnum_FormFactor.Id`