Pandas¶

Overview¶

Pandas provides a timeseries and stack-of-timeseries objects. They seem heavily geared towards financial data. Despite the fact of being an ndarray, Pandas objects seem to be a specialized alternative to ndarrays rather than an augmentation of them.

Features

Is an ndarray
axes are not named
Is dict-like, with respect to its indices (ticks)
If ticks are indices, semantics of indexing are ambiguous
Separate objects from 1D and 2D, no support for n>2

Indexing¶

Point-indexing syntax can use ticks or integer indices. Range indexing only works with integers, but uses the same syntax

Semantic Ambiguity¶

Integer tick values interfere with integer indexing, for example:

>>> t = pandas.Series.fromValue(1.0, range(5,0,-1), 'i')
>>> t[:] = np.random.randint(100, size=5)
>>> t
5    23
4    62
3    66
2    91
1    91
>>> t[2] = 0
>>> t
5    23
4    62
3    66
2    0
1    91

Binary Operations¶

Alignment¶

If data is partially aligned, missing data is filled with NaNs. This introduces a union with respect to the “range” of the data. This also will cast the data to floating point.:

>>> t.dtype
 dtype('int32')
>>> t - t[:3]
5    0.0
4    0.0
3    0.0
2    NaN
1    NaN

Table Of Contents

Pandas
- Overview
  - Indexing
    - Semantic Ambiguity
  - Binary Operations
    - Alignment

Previous topic

Larray (aka Larry)

Next topic

This Page