feat: match_to_schema

[KronosTheLate]

Is your feature request related to a problem?

When combining a new table into an existing table, it is often useful to allow some degree of flexibily for e.g. casting dates to strings or opposite, or up/downcasting the integer and floating columns. However, instead of upcasting to the most general type, it is really useful to be able to say "try to cast the existing dataframe to the new schema", if a date-column was previously cast to a string, and is not correctly cast to a date.

What is the motivation behind your request?

No response

Describe the solution you'd like

The function match_to_schema from polars is more or less exactly what I want. It, or something like it, would allow more flexible schema matching without sacrificing the control, in the sense that the new schema should have priority over the old one.

What version of ibis are you running?

None

What backend(s) are you using, if any?

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[KronosTheLate]

I ended up writing an implementation for this:

def match_to_schema(df_left, df_right, raise_on_extra_columns:bool = False, raise_on_missing_columns:bool = False):
    """
    Return a version of `df_left`, but matched to the schema of `df_right`
    
    Required arguments:
        - df_left: The table to return with altered schema
        - df_right: The table to infer the schema from
        
    Optional arguments:
        - raise_on_extra_columns: If True, raise an error if df_right has a columns that df_left does not. Default: False
        - raise_on_missing_columns:  If True, raise an error if df_left has a columns that df_right does not. Default: False
    """
    exprs = []
    if raise_on_missing_columns:
        for c in df_left.columns:
            if c not in df_right.columns:
                raise ValueError(f"Column '{c}' was found in df_left, but not in df_right, and `raise_on_missing_columns` is set to True.")
    for target_name, target_type in df_right.schema().items():
        if target_name in df_left.columns:
            exprs.append(df_left[target_name].cast(target_type).name(target_name))
        else:
            if raise_on_extra_columns:
                raise ValueError(f"Column '{target_name}' was found in df_right, but not in df_left, and `raise_on_extra_columns` is set to True.")
            exprs.append(ibis.literal(None).name(target_name))
    return df_left.select(exprs)

In the same vein as the functions suggested in #11378, I think this could be quite naturally added as a method to ibis.Table.