Python select columns with similar names

Question

My data frame has many columns, but for some, they start with similar words as column names:

>>> df.columns
Index([u'VESSELTYPE', u'VESSELNAME', u'PERIODSTART', u'PERIODEND', u'ESTREDELI', u'HIRE', u'DAYS Mar-14', u'DAYS Q2 2014', u'DAYS Q3 2014', u'DAYS Q4 2014', u'DAYS Q1 2015', u'DAYS Q2 2015', u'DAYS Q3 2015', u'DAYS Q4 2015', u'NUMDAYS'], dtype='object')
>>> 

See here, I want to select all with 'DAYS' in front, is there any thing like in SQL, I can use * to represent everything after 'DAYS'?

Answer 1

Try to adapt this to your code.

l = ([u'VESSELTYPE', u'VESSELNAME', u'PERIODSTART', u'PERIODEND', u'ESTREDELI', u'HIRE', u'DAYS Mar-14', u'DAYS Q2 2014', u'DAYS Q3 2014', u'DAYS Q4 2014', u'DAYS Q1 2015', u'DAYS Q2 2015', u'DAYS Q3 2015', u'DAYS Q4 2015', u'NUMDAYS'], 'AnotherItem')

days = [s for s in l[0] if s.startswith('DAYS')]
print days

You can use the "startswith()" method available to all strings and make some logic with it.

If you don't like list-comprehensions, here's its equivalent in a for-loop:

days = []
for s in l[0]:
    if s.startswith('DAYS'):
        days.append(s)

print days

Answer 2

pandas indexes (and dataframes) are numpy arrays, so you can use Numpy's index tricks - in particular, indexing with a vector of bools can is helpful here. To test whether any particular entry starts with "DAYS" is a matter of Python's standard string methods; put this in a list comprehension to get your list of bools:

df.columns[[x.startswith("DAYS") for x in df.columns]]

or you can use the map method on arrays to avoid the double brackets:

df.columns(df.columns.map(lambda x: x.startswith("DAYS"))

Answer 3

Yes, you can try something like this:

'DAYS' == x[:len('DAYS')]

Where you are comparing your target string to the string first n characters. Being n the length of your target string.

You can use this condition in your selection.

Answer 4

To get only the days columns of your dataframe do:

print df[[x for x in df.columns if "DAY" in x]]