Confused with data formats

I am using pandas and scikit to explore machine learning. I am experimenting with reading historical information about ships and attempting to classify the type of ship. I am very confused with the proper format of the data.

I have a data set that looks like the following in a pandas dataframe

     mmsi vessel_type        vessel_name  length  width  flag_country                  
412417097     Unknown     ZHELINGYU52052      36      7         China       
412422090       Cargo        JIE BANG 17      50      8         China    
413975458       Other  GUIPINGNANHUO2626     308     14         China      
412456888       Other  ZHE CANG YU 72777     194     14         China       
413808364     Unknown    WAN SHUN FA1818      55     10         China       
215380000       Cargo          GIORGOS B     186     30         Malta

my initial experiment is to see if i can predict the vessel_type by using a set of trivial features (length, width, flag_country).

Question 1

What is the correct format of the data for passing to scikit? I know I need a list of features but am not sure what the content should be. see below:

features
[36,7,China]
[50,8,China]
[308,14,China]

or do i need to map the flag_country to an integer value? FI so is there an easy way to do this in scikit such that when I run metrics the country is actually displayed instead of the corresponding integer code.

likewise with labels i am not sure how to represent the labels. SHoudl I have only one value represented for each feature row, or do I need all labels for every feature row with all but the single label associated to the row set to 0?

Labels
['Unknown']
['Cargo']
['Other']

-or-

Unknown Cargo  Other
   1      0      0
   0      1      0
   0      0      1

Question 2

In addition to the content of the files I am having a hard time getting the pandas dataframe type to be accepted by scikit. This is new. In the past I did some textbase classification using pandas and scikit and everything juts worked like magic. No whenever I try and pass pandas dataframe objects to scikit i get errors like ValueError: Array contains NaN or infinity. I went back to my previous experiments and tried to run the same code that worked about 9 months ago and get the same error. I have upgraded both pandas and scikit since the successful experiment, but did not expect to encounter such significant incompatibilities. Below is my base attempt to passing data to scikit. Note I have tried several explicit conversions to mp.arrays, using dataframe.values, but nothing seems to work.

labels = df_train["vessel_type"]
features = df_train[featurelist]
clf = LogisticRegression()
clf.fit(features,labels)

sorry for the long post, but i am pretty much stuck on this after having spend over 10 hours to try and get the 1st experiment to work. Any help or advice would be greatly appreciated.