"""Manipulation timestep dataframe."""
import numpy as np
import pandas as pd
[docs]def _timestep_single(array, index=0, timesteps=2, keep_first=True):
"""Add timesteps array for single column array.
Parameters
----------
array : array
Single column two-dimensional array.
index : int, optional
Index column, by default 0
timesteps : int, optional
Number of timesteps, by default 2
keep_first : bool, optional
Include original array if set True, by default True
Returns
-------
array
Return 2D array with timesteps.
"""
if array.ndim == 1:
array = array.reshape(-1, 1)
x = []
for i in range(timesteps + 1):
start = timesteps - i
end = -i if i > 0 else None
x.append(array[start:end, index])
if not keep_first:
x.pop(0)
return np.array(x).transpose()
[docs]def _timestep_multi(array, index=None, timesteps=2, keep_first=True):
"""Add timesteps array for multiple column array.
Parameters
----------
array : array
Multiple numeric column two-dimensional array.
index : list of int, optional
List of columns index, by default None
timesteps : int, optional
Number of timesteps, by default 2
keep_first : bool, optional
Include original column if set True, by default True
Returns
-------
array
Return 2D array with timesteps.
"""
_, col = array.shape
col = range(col)
x = []
for c in col:
arr = array[:, c].reshape(-1, 1)
if index is None:
x.append(_timestep_single(arr, timesteps=timesteps,
keep_first=keep_first))
elif c in index:
x.append(_timestep_single(arr, timesteps=timesteps,
keep_first=keep_first))
else:
x.append(arr[timesteps:])
return np.concatenate(x, axis=1)
[docs]def timestep_table(dataframe,
columns=None,
timesteps=2, keep_first=True,
template="{column}_tmin{i}"):
"""Generate timesteps directly from DataFrame.
Parameters
----------
dataframe : DataFrame
Dataframe consist of numeric-column only
columns : list of str, optional
List of columns name to generate, by default None
timesteps : int, optional
Number of timesteps, by default 2
keep_first : bool, optional
Column _tmin0 will be included if set True, by default True
template : str
Format column name, by default "{column}_tmin{i}"
Returns
-------
DataFrame
DataFrame with additional timesteps columns.
"""
# dataframe parameter
columns_name = list(dataframe.columns)
new_index = dataframe.index[timesteps:]
# Generate new column name
new_columns_name = []
for column in columns_name:
if columns is None:
for i in range(timesteps + 1):
new_columns_name.append(template.format(column=column, i=i))
elif column in columns:
for i in range(timesteps + 1):
new_columns_name.append(template.format(column=column, i=i))
else:
new_columns_name.append(column)
if keep_first is False:
for column in new_columns_name:
if column.endswith("_tmin0"):
new_columns_name.remove(column)
# Convert name columns to index
index_columns = []
if columns is not None:
for column in columns:
index_columns.append(columns_name.index(column))
else:
index_columns = None
# Generate Dataframe column with timesteps
columns_values = dataframe.values
columns_values = _timestep_multi(
columns_values,
index=index_columns,
keep_first=keep_first,
timesteps=timesteps
)
timestep_dataframe = pd.DataFrame(
data=columns_values,
index=new_index,
columns=new_columns_name
)
return timestep_dataframe
