值 Pandas Dataframe 列中的项目计数，其中包含字符串列表作为值

Question

I want to count the occurrence of items inside list present in column of a dataset. I have my tags column in the dataset. My dataset consists data in following format

tags
-----------
['symfony' 'assestic]
['java' 'containers' 'kubernetes']
['python' 'pelican']
['python' 'api' 'oath' 'python-requests']
['google-api' 'google-cloud-storage']

The list seems to be in string format too. I am not being able to convert the string into list without concatenating all the item inside the list.

#Checking the type of first 5 rows tags
for i,l in enumerate(df.tags):
    print('list',i,'is class', type(l) )
    if i ==4:
        break

Output will be

list 0 is class <class 'str'>
list 1 is class <class 'str'>
list 2 is class <class 'str'>
list 3 is class <class 'str'>
list 4 is class <class 'str'>

I tried two methods for it Method 1:

def clean_tags_list(list_):
    list_ = list_.replace("\"['" , '[')
    list_ = list_.replace("']\"", ']')
    list_ = list_.replace("'","")
    return list_
df['tags'] = df['tags'].apply(clean_tags_list)

Output will be

   tags                              
   ----------------------------------
   [symfony assestic]                 
   [java containers kubernetes]      
   [python pelican]                  
   [pyton api oath python-requests]   
   [google-api google-cloud-storage]  

But The Value counts doesnt work with the above Series. Value Counts will give following output

[symfony assestic]                 1                
[java containers kubernetes]       1      
[python pelican]                   1                 
[pyton api oath python-requests]   1   
[google-api google-cloud-storage]  1

Method 2: I tried using replace, strip, asl.literal_eval().

Question How to achieve output in following format?

python 2
symfony 1
assestic 1

Answer 1

You can flatten the column so that each list element is in a separate row, then just use .value_counts() . However since the data is actually strings that look like lists, you'll have to convert them to actual lists first.

Here's an example:

import ast

df = pd.DataFrame({
    "tags": [
        "['symfony', 'assestic']",
        "['java', 'containers', 'kubernetes']",
        "['python', 'pelican']",
        "['python', 'api', 'oath', 'python-requests']",
        "['google-api', 'google-cloud-storage']",
    ]
})

df["tags"]\
    .apply(ast.literal_eval)\ # convert strings to lists
    .apply(lambda x: pd.Series(x))\ # convert lists to series
    .stack()\ # flatten the multiple series into a single series
    .value_counts() # get value counts

With result:

python                  2
java                    1
oath                    1
google-cloud-storage    1
api                     1
assestic                1
kubernetes              1
pelican                 1
symfony                 1
python-requests         1
google-api              1
containers              1

Note that if the data you're working with is composed of lists rather than strings that look like lists, the approach is the same without the .apply(ast.literal_eval) line.