Incorrect parsing of column a name when an identifier is used in Snowflake TIMEDIFF in the qualify method

[wezham]

As stated in the snowflake documentation the signature for timediff

TIMEDIFF( <date_or_time_part> , <date_or_time_expr1> , <date_or time_expr2> )

date_or_time_part
The unit of time. Must be one of the values listed in Supported Date and Time Parts (e.g. month). The value can be a string literal or can be unquoted (e.g. 'month' or month).

However, when using the unquoted argument like YEAR instead of 'YEAR', qualify sees the argument as a column and attempts to resolve it in the schema. For example you can see the code block below:

table_schema = {"sample_table": {"created_at": "timestamp_tz"}}
sql_query = "SELECT TIMEDIFF(YEAR, created_at, current_timestamp()) from sample_table;"
expression_ = qualify(
    parse_one(sql_query),
    dialect="snowflake",
    qualify_columns=True,
    validate_qualify_columns=True,
    expand_alias_refs=True,
    schema=table_schema,
)

Results in

    raise OptimizeError(
sqlglot.errors.OptimizeError: Column '"YEAR"' could not be resolved

[tobymao]

you need to specify dialect in parse_one

parse_one(sql, read="snowflake")

[tobymao]

separately we need to implement timediff

[wezham]

Yep that was my error when re-producing. Thanks