Assign Weights/Prepare

Currently this section only applies to FRTBDataSet. In the future python users will be able to define their own, custom prepare functions too.

Prepare is a fixed calculation which happens with every request for scans/db DataSources. For InMemory it happens only when you call .prepare().

from_config() call .prepare() depending on source_type.

A good usecase for that is to assign weights column to your data, as per FRTB regulation for example. Make sure you assign weights before doing any computations:

# TODO scan here to demonstrate prepare

import polars as pl
import ultibi as ul

# Note: sometimes Polars needs abit of help to determine the correct datatype for your
# columns. This is because Polars scans first N rows to determine correct dtype
exposures = pl.read_csv("./data/frtb/Delta.csv", dtypes={"SensitivitySpot": pl.Float64})
# Feel free to join other attributes such as Hierarchy here ...
ds = ul.FRTBDataSet.from_frame(exposures)

original = ds.frame()  # keep the old value for comparison

ds.prepare()

prepared = ds.frame()


# Let's see what happened
def diff(df1: pl.DataFrame, df2: pl.DataFrame) -> pl.DataFrame:
    """Columns in df1 which are not in df2

    Args:
        df1 (pl.DataFrame): First to compare
        df2 (pl.DataFrame): Second to compare

    Returns:
        pl.DataFrame: _description_
    """
    return df1.select([c for c in df1 if c.name not in df2])


diff_frame = diff(prepared, original)
print(diff_frame.columns)

You will get an error if you try to assign twice. Now, let's see what happened. We will need a little helper function:

At the time of writing this returns 6 new columns(names of the columns might change slightly but the meaning will always be the same):