Rank encode a dataframe column

rank_encode_dataframe.py

"""
USAGE:

>>> replace_with_rank_indexes(dataframe, column_name)

or

>>> df.pipe(replace_with_rank_indexes, column='column_name')

Returns a copy of the dataframe with the designated column replaced by 
the respective rank values of each numeric value in that column

"""

import pandas as pd
import numpy as np
import numba

@numba.njit
def _replace_over_vector(rank_map):
    ranks = np.empty(rank_map.shape[0], np.int32)
    
    # initialize to lowest value
    prev_val = rank_map[0][1]
    current_rank_idx = 1
    
    for i in range(ranks.size):
        
        if rank_map[i][1] > prev_val:
            prev_val = rank_map[i][1]
            current_rank_idx += 1
            
        ranks[int(rank_map[i][0])] = current_rank_idx   
    return ranks

def replace_with_rank_indexes(df, column=None):
    """
    -------------------------------------------------
    :param df: a pandas dataframe
    :param column: the columm to replace with rank values. MUST be numeric

    :return: a copy dataframe with the respective rank 
        values of each numeric value in the designated column
    __________________________________________________
    """
    
    if column is None:
        raise ValueError('method requires column value')
    df = df.copy()
    df_col = df[column]
    sorted_df_col = df_col.sort_values()
    rank_map = np.array((sorted_df_col.index + 0.0001, sorted_df_col.values)).T
    df[column] = _replace_over_vector(rank_map)
    return df