python - Numpy select returning boolean error message

Question

I would like to find matching strings in a path and use np.select to create a new column with labels dependant on the matches I found.

This is what I have written

import numpy as np
conditions  = [a["properties_path"].str.contains('blog'),
               a["properties_path"].str.contains('credit-card-readers/|machines|poss|team|transaction_fees'),
               a["properties_path"].str.contains('signup|sign-up|create-account|continue|checkout'),
               a["properties_path"].str.contains('complete'),
               a["properties_path"] == '/za/|/',
              a["properties_path"].str.contains('promo')]
choices     = [ "blog","info_pages","signup","completed","home_page","promo"]
a["page_type"] = np.select(conditions, choices, default=np.nan)

However, when I run this code, I get this error message:

ValueError: invalid entry 0 in condlist: should be boolean ndarray

Here is a sample of my data

3124465                                       /blog/ts-st...
3124466                                       /card-machines
3124467                                       /card-machines
3124468                                       /card-machines
3124469                               /promo/our-gift-to-you
3124470                                   /create-account/v1
3124471                                          /za/signup/
3124472                                   /create-account/v1
3124473                                             /sign-up
3124474                                                 /za/
3124475                                        /sign-up/cart
3124476                                           /checkout/
3124477                                            /complete
3124478                                       /card-machines
3124479                                       /continue
3124480                             /blog/article/get-car...
3124481                             /blog/article/get-car...
3124482                                          /za/signup/
3124483                                 /credit-card-readers
3124484                                          /signup
3124485                                 /credit-card-readers
3124486                                   /create-account/v1
3124487                                 /credit-card-readers
3124488                                   /point-of-sale-app
3124489                                   /create-account/v1
3124490                                   /point-of-sale-app
3124491                                 /credit-card-readers

Answer 1

The .str methods operate on object columns. It's possible to have non-string values in such columns, and as a result pandas returns NaN for these rows instead of False. np then complains because this is not a Boolean.

Luckily, there's an argument to handle this: na=False

a["properties_path"].str.contains('blog', na=False)

Alternatively, you could change your conditions to:

a["properties_path"].str.contains('blog') == True
#or
a["properties_path"].str.contains('blog').fillna(False)

---

Sample

import pandas as pd
import numpy as np

df = pd.DataFrame({'a': [1, 'foo', 'bar']})
conds = df.a.str.contains('f')
#0      NaN
#1     True
#2    False
#Name: a, dtype: object

np.select([conds], ['XX'])
#ValueError: invalid entry 0 in condlist: should be boolean ndarray

conds = df.a.str.contains('f', na=False)
#0    False
#1     True
#2    False
#Name: a, dtype: bool

np.select([conds], ['XX'])
#array(['0', 'XX', '0'], dtype='<U11')