Recommended way to create columns based on conditions in pandas flavor

[gygabyte017]

Hi, suppose I have a CASTable object with a string column col1. I want to create, using pandas flavor, a new column col2 such that for example

if col1 starts with 'A' => col2 = 0
else if col1 starts with 'B' => col2 = 1
else col2 = 2

I'm trying something like df["col2"] = df[df.col1.str.startswith('A')] ... ? but I'm not going far.
I saw that in pandas usually something like numpy.select is used but clearly numpy doesn't make sense here.

So what would you recommend here? I know I might use eval or a datastep but I would like to keep it as python as possible to improve debugging and standardization.

Just to set a context, in the application I'm writing I will need to apply to tables some long conditional rules coming from config files and calculate new columns, so I'd prefer to just write conditions with python syntax and apply them to CASTables letting swat doing the hard work (i.e. calculating all the computedvars), instead of writing the datastep as a string which would be harder to debug.

Thank you

[kesmit13]

If it was just 0 or 1, I think you could use just the Python methods to do this since the boolean result would get mapped to an integer, but I think the addition of a third branch might mean having to use some sort of SAS-specific functions. I can't say that I'm coming up with any ideas that wouldn't require something along the lines of what you have in your example.