根据部分字符串匹配创建两个新的 pandas 列

Question

I have a dataframe of construction titles and names arranged in a random order (but someone's name is always in the cell to the right of their title) like so:

   contact_1_title contact_1_name contact_2_title contact_2_name contact_3_title contact_3_name      contact_4_title contact_4_name
0  owner_architect            joe    other_string   other_string    other_string   other_string         other_string   other_string
1     other_string   other_string       architect           jack    other_string   other_string         other_string   other_string
2     other_string   other_string    other_string   other_string    other_string   other_string  self_cert_architect           mary
3     other_string   other_string    other_string   other_string           owner           phil         other_string   other_string
4       contractor          sarah    other_string   other_string    other_string   other_string         other_string   other_string
5     other_string   other_string       expeditor           kate    other_string   other_string         other_string   other_string

I want to pull every title with the word "architect" in it and insert it into its own, new column. I also want to pull every name in the cell immediately to the right and insert it into its own column as well. My desired output:

        arch_title_col arch_name_col
0      owner_architect           joe
1            architect          jack
2  self_cert_architect          mary

I'm at a loss as to how to go about this. I tried working with iterrtuples() but I didn't get very far.

Answer 1

What you need is pd.wide_to_long , but I couldn't get the syntax right for how your columns are formatted. So here it is manually:

title = pd.concat([df[col] for col in df.filter(like='title')], axis=0)
name = pd.concat([df[col] for col in df.filter(like='name')], axis=0)
df = pd.concat([title, name], axis=1)
df.columns = ['title', 'name']

Now that we have things in a good format, it's a simple check:

out = df[df.title.str.contains('architect')]
print(out)

Output:

                 title  name
0      owner_architect   joe
1            architect  jack
2  self_cert_architect  mary

---

I promise you that 99% of the time, iter... is not what you want, and there is a far better panda's specific way to do whatever you want to do.