Closed
Description
A new module core.align
would hold a couple of generic alignment functions, one of which will be align_multiple
or possibly align2
(taking 2 pandas objects) and then align
would take a sequence of pandas objects to align.
Need to think about the API a bit more.
Psuedocode-ish:
df = DataFrame(...)
df2 = DataFrame(...) # possibly different index and columns than df
s = Series(...)
result = pd.align([df, df2, s], how='outer') # compute the union of the indexes
# smoke test 1
result.columns == df.columns.join(df2.columns, how='outer').join(s.index, how='outer')
result.index == df.index.join(df2.index, how='outer')
result2 = pd.align([df, df2, s], how='inner') # intersection
# smoke test 2
result2.columns == df.columns.join(df2.columns).join(s.index)
result2.index == df.index.join(df2.index)
input with Index
types would return the union/intersection of the objects. not clear what would happen if aligning an Index
and another non-Index
pandas object