Replace partial matches behavior

[TomAugspurger]

I ran into some unexpected behavior with replace today.

I want to replace the answer with a numeric value according to the dictionary weights

In [33]: df.answer.head()
Out[33]: 
0       Strongly Agree
1                Agree
2              Neutral
3             Disagree
4    Strongly Disagree
Name: answer, dtype: object

In [34]: weights
Out[34]: 
{'Agree': 4,
 'Disagree': 2,
 'Neutral': 3,
 'Strongly Agree': 5,
 'Strongly Disagree': 1}

In [35]: df.answer.replace(weights).head()
Out[35]: 
0    4
1    4
2    3
3    2
4    2
dtype: int64

It looks like replace matches on the first part, Agree, or Disagree and doesn't make it through the dict to the Stronglys. Am I just being a noob with how regexes work, or is this a bug?

[cpcloud]

ok yep ... should possibly add a 'search' or 'match' parameter or something like match=False

[cpcloud]

right now it does search

[cpcloud]

in this case instead of using replace you can do

Series(weights)[answer]

this is probably faster for larger series as well

[cpcloud]

@jreback @jtratner what do think ?

was thinking:

mode='search' (default) or mode='match'

or

match=False

[cpcloud]

really trivial fix after deciding on the API

[jreback]

he wants an exact match....I thought you do that UNLESS you see a regex?

[cpcloud]

hm yes should be that ... let me look

[cpcloud]

yep just needed to accept tuples