using (or converting to code) strings for np.select

Question

I have the following code that works perfect:

df2 = pd.DataFrame({'TEXT': ['add', 'bede', 'agdd', 'bbbb', 'aaaa'],
                       'PRICE': [622, 200, 100, 459, 250]})
temp=df2['TEXT']
col         = 'TEXT'
conditions  = [ temp.str.contains('a'), temp.str.contains('b'), temp.str.contains('c') ]
choices     = [ "contains a", 'contains b', 'contains c' ]

df2["what_contains"] = np.select(conditions, choices, default=np.nan)

the thing is, the contents of conditions have to be read from a csv, which of course means that they will be strings. I have tried the following:

conditions=the_csv['cond'].apply(compile,filename='<string>',mode='eval')

but I get an error:invalid entry 0 in condlist: should be boolean ndarray

the .csv looks like this:

thanks!!

Answer 1

If your conditions are always of the form .str.contains, you can avoid using eval - which will be safer and more clear.

# get the conditions as a list of string conditions
given_conds = the_csv['cond'].tolist()
# get the string inside each condition
searching_for = [c.split("('")[-1].split("')")[0] for c in given_conds]
# form the real boolean conditions
conditions = [temp.str.contains(c) for c in searching_for]
# choices text
choices = [f"contains {c}" for c in searching_for]

df2["what_contains"] = np.select(conditions, choices, default=np.nan)

Answer 2

You could use eval. Whether you should is up to you.

import pandas as pd
import numpy as np

df2 = pd.DataFrame({'TEXT': ['add', 'bede', 'agdd', 'bbbb', 'aaaa'],
                       'PRICE': [622, 200, 100, 459, 250]})
temp=df2['TEXT']
col         = 'TEXT'
conditions  = [ eval("temp.str.contains('a')"), eval("temp.str.contains('b')"), eval("temp.str.contains('c')") ]
choices     = [ "contains a", 'contains b', 'contains c' ]

df2["what_contains"] = np.select(conditions, choices, default=np.nan)