Python-Numpy-Pandas Compatible Functions

calcpy provides some utility functions to provide a consistent interface to scaler, np.ndarray, pd.Series, and pd.DataFrame.

Properties

calcpy.shape(arg)[source]

Get the shape of an argument.

Parameters:

arg

Return type:

tuple

Examples

>>> shape(1)
()
>>> shape([1, 2, 3])
(3,)
>>> shape(np.array([1, 2, 3]))
(3,)
>>> shape(pd.Series([1, 2, 3]))
(3,)
>>> shape(pd.DataFrame({"A": 1, "B": 2}, index=[0]))
(1, 2)
calcpy.ndim(arg)[source]

Get the number of dimensions of an argument.

Parameters:

arg

Return type:

int

Examples

>>> ndim(1)
0
>>> ndim([1, 2, 3])
1
>>> ndim(np.array([1, 2, 3]))
1
>>> ndim(pd.Series([1, 2, 3]))
1
>>> ndim(pd.DataFrame({"A": 1, "B": 2}, index=[0]))
2
calcpy.size(arg)[source]

Get the size of an argument.

Parameters:

arg

Return type:

int

Examples

>>> size(1)
1
>>> size([1, 2, 3])
3
>>> size(np.array([1, 2, 3]))
3
>>> size(pd.Series([1, 2, 3]))
3
>>> size(pd.DataFrame({"A": 1, "B": 2}, index=[0]))
2
calcpy.empty(arg)[source]

Check whether it is empty.

Parameters:

arg

Return type:

bool

Examples

>>> empty(1)
False
>>> empty([])
True
>>> empty([1, 2, 3])
False
>>> empty(np.array([1, 2, 3]))
False
>>> empty(pd.Series([1, 2, 3]))
False
>>> empty(pd.DataFrame({"A": 1, "B": 2}, index=[0]))
False

Create and Compute

calcpy.broadcast_first(fun)[source]

Decorator for supporting np.ndarray, pd.Series, and pd.DataFrame.

Parameters:

fun (callable) – Callable that applies to a single element in its first argument.

Returns:

Callable that applies to a single element or a list, tuple, np.ndarray,

pd.Series, or pd.DataFrame.

Return type:

callable

Examples

>>> @broadcast_first
... def add(x, y):
...     return x + y
>>> add(1, 2)
3
>>> add([1, 2, 3], 2)
[3, 4, 5]
>>> add(np.array([1, 2, 3]), 2)
array([3, 4, 5])
>>> add(pd.Series([1, 2, 3]), 2)
0    3
1    4
2    5
dtype: int64
>>> add(pd.DataFrame({"A": 1, "B": 2}, index=[0]), 2)
      A  B
0     3  4
calcpy.full_like(template, fill_value, **kwargs)[source]

Create a np.array or pd.Series or pd.DataFrame with the same shape as template.

Parameters:

  • template (list | tuple | np.ndarray | pd.Series | pd.DataFrame)

  • fill_value – Value to populate.

  • **kwargs – Keyword arguments for np.full_alike(), pd.Series(), or pd.DataFrame().

Return type:

list | tuple | np.ndarray | pd.Series | pd.DataFrame

Raises:

TypeError –

Examples

Create list and tuple.

>>> full_like([1, 2, 3], 0)
[0, 0, 0]
>>> full_like((1, 2, 3), 0)
(0, 0, 0)

Create np.array.

>>> full_like(np.array([1, 2, 3]), 0)
array([0, 0, 0])

Create pd.Series and pd.DataFrame.

>>> full_like(pd.Series([1, 2, 3]), 0)
0    0
1    0
2    0
dtype: int64
>>> full_like(pd.DataFrame({"A": 1, "B": 2}, index=[0]), 0)
      A  B
0     0  0
calcpy.mapi(inputs, f)[source]

Apply a function on every input element and its index.

Parameters:

  • inputs (list | tuple | np.ndarray | pd.Series | pd.DataFrame) – Input data to transform

  • f (callable) – Transformation function with some positional arguments: - For list and tuple: f(value, index) -> new_value - For ndarray: f(value, index_0, index_1, …, index_(ndim-1)) - For DataFrame: f(value, index, column) -> new_value - For Series: f(value, index, name) -> new_value

Returns:

Transformed data with same shape/index/columns

Return type:

list | tuple | np.ndarray | pd.Series | pd.DataFrame

Examples

Transform a list.

>>> def printall(*args):
...    return ":".join(str(arg) for arg in args)
>>> mapi([1, 2, 3], printall)
['1:0', '2:1', '3:2']

Transform a ndarray.

>>> from calcpy import add
>>> a = np.ones(shape=(2, 3, 4))
>>> mapi(a, add)
array([[[1., 2., 3., 4.],
        [2., 3., 4., 5.],
        [3., 4., 5., 6.]],
    [[2., 3., 4., 5.],
        [3., 4., 5., 6.],
        [4., 5., 6., 7.]]])

Transform a Series.

>>> s = pd.Series("value", index=range(3))
>>> mapi(s, printall)
0    value:0:None
1    value:1:None
2    value:2:None
dtype: object

Transform a Series with datetime index and Series name.

>>> tindex = pd.date_range("2000-01-01", "2000-01-03")
>>> s = pd.Series("value", index=tindex, name="name")
>>> mapi(s, printall)
2000-01-01    value:2000-01-01 00:00:00:name
2000-01-02    value:2000-01-02 00:00:00:name
2000-01-03    value:2000-01-03 00:00:00:name
Freq: D, Name: name, dtype: object

Transform a Series with multi-level index.

>>> mindex = pd.DataFrame({"app": "X", "date": tindex}).set_index(["app", "date"]).index
>>> s = pd.Series("value", index=mindex)
>>> mapi(s, printall)
app  date
X    2000-01-01    value:('X', Timestamp('2000-01-01 00:00:00')):...
     2000-01-02    value:('X', Timestamp('2000-01-02 00:00:00')):...
     2000-01-03    value:('X', Timestamp('2000-01-03 00:00:00')):...
dtype: object

Transform a Series to another datatype

>>> def sumlen(*args):
...     return sum(len(arg) for arg in args)
>>> s = pd.Series("value", index=["a", "b"], name="name")
>>> mapi(s, sumlen)  
a    10
b    10
Name: name, dtype: int...

Transform a DataFrame.

>>> df = pd.DataFrame('hello', index=range(4), columns=range(3))
>>> mapi(df, printall)
           0          1          2
0  hello:0:0  hello:0:1  hello:0:2
1  hello:1:0  hello:1:1  hello:1:2
2  hello:2:0  hello:2:1  hello:2:2
3  hello:3:0  hello:3:1  hello:3:2

Transform a DataFrame with multi-level index.

>>> df = pd.DataFrame("value", index=mindex, columns=["a"])
>>> print(mapi(df, printall))
                                                              a
app date
X   2000-01-01  value:('X', Timestamp('2000-01-01 00:00:00')):a
    2000-01-02  value:('X', Timestamp('2000-01-02 00:00:00')):a
    2000-01-03  value:('X', Timestamp('2000-01-03 00:00:00')):a

Handle empty input.

>>> s = pd.Series(dtype=object, name='empty')
>>> mapi(s, printall)
Series([], Name: empty, dtype: object)

Create a DataFrame whose elements are all the same as index.

>>> from calcpy import arggetter
>>> index = pd.date_range("2000-01-01", "2000-01-03")
>>> df = pd.DataFrame(index=index, columns=["A", "B"])
>>> mapi(df, arggetter(1))
                    A          B
2000-01-01 2000-01-01 2000-01-01
2000-01-02 2000-01-02 2000-01-02
2000-01-03 2000-01-03 2000-01-03

Compare

calcpy.overall_equal(loper, roper)[source]

Check whether two operands are exactly equal as a whole.

It behaves like np.array_equal for np.ndarray, and loper.equals(roper) for pd.Series and pd.DataFrame.

Parameters:

  • loper (number | list | tuple | np.ndarray | pd.Series | pd.DataFrame)

  • roper (number | list | tuple | np.ndarray | pd.Series | pd.DataFrame)

Return type:

bool

Examples

Compare lists.

>>> overall_equal([1, 2, 3], [1, 2, 3])
True

Compare pd.DataFrame.

>>> import pandas as pd
>>> df = pd.DataFrame({"A": 1, "B": 2}, index=[0])
>>> overall_equal(df, df+0)
True
>>> overall_equal(df, df+1)
False
>>> overall_equal(df, df.iloc[:, 0])
False
>>> overall_equal(df["A"], df["B"]-1)
True