Unable to find column name when written in non-lowercasce quoted style

[rampage644]

Following query fails with 000200: DataFusion error: Schema error: No field named "User". Valid fields are source.user, source.email, source.crm_user_id, ...

create or replace transient table EMBUCKET.driveload.driveload_clari_export_forecast_net_iacv_source
    

    
    as (WITH source AS (

    SELECT *
    FROM EMBUCKET.driveload.clari_export_forecast_net_iacv

), renamed AS (

    SELECT
         "User"::VARCHAR            AS user,
         "Email"::VARCHAR           AS email,
         "CRM_User_ID"::VARCHAR     AS crm_user_id,
         "Role"::VARCHAR            AS sales_team_role,
         "Parent_Role"::VARCHAR     AS parent_role,
         "Timeframe"::VARCHAR       AS timeframe,
         "Field"::VARCHAR           AS field,
         "Week"::NUMBER             AS week,
         "Start_Day"::DATE          AS start_date,
         "End_Day"::DATE            AS end_date,
         "Data_Type"::VARCHAR       AS data_type,
         "Data_Value"::VARCHAR      AS data_value
    FROM source

)

SELECT *
FROM renamed
    )
;

[osipovartem]

The reason that we don't normalize ident if it was quoted

/// Normalize an owned identifier to a lowercase string, unless the identifier is quoted.
pub fn normalize_ident(id: Ident) -> String {
    match id.quote_style {
        Some(_) => id.value,
        None => id.value.to_ascii_lowercase(),
    }
}

So we have
Expr Cast(Cast { expr: Column(Column { relation: None, name: "User" }), data_type: Utf8 })
And
DFSchema { inner: Schema { fields: [Field { name: "user", data_type: Utf8, nullable: true, dict_id: 0, dict_is_ordered: false, metadata: {"PARQUET:field_id": "1"} }], metadata: {} }, field_qualifiers: [Some(Full { catalog: "snowplow", schema: "public", table: "source" })], functional_dependencies: FunctionalDependencies { deps: [] } }

which cannot pass

self.validate_schema_satisfies_exprs(schema, std::slice::from_ref(&expr))?;

[rampage644]

This is good to know. However, system should align with snowflake identifiers, not datafusion ones.