ENH: use the option framework to opt-out of auto-alignment

[jxrossel]

Feature Type

• Adding new functionality to pandas

• Changing existing functionality in pandas

• Removing existing functionality in pandas

Problem Description

As far as I know, auto-alignment of DataFrames and Series (e.g. with arithmetic operations) cannot be disabled. To work in "strict" mode, the user must first check alignment and then make a computation:

a = pd.DataFrame(...)
b = pd.DataFrame(...)

c = pd.Series(...)
d = pd.Series(...)

if not a.index.equals(b.index) or not a.columns.equals(b.columns):
    raise ...
k = a + b

if not c.index.equals(d.index):
    raise ...
m = c + d

Since alignment must also be checked internally by pandas, this is not only cumbersome but inefficient.

Feature Description

Use the pandas option framework to opt-out of auto-alignment. One could even go one step further to allow deactivation of alignment check altogether for data that is known to be aligned (to gain some computation time). So the option would look like: mode.on_misaligned:

• coerce: default and current behaviour
• raise: strict mode, raise an error if indices are not aligned
• ignore: skip alignment checking and proceed with the computation assuming data is already aligned
  The user would then set the option as he sees fit, e.g. with a context manager:

with pd.option_context("mode.on_misaligned", "raise"):
    k = a + b
    m = c + d

Alternative Solutions

The alternative solution is the problem

Additional Context

No response