如何避免熊猫的 concate 和 to_csv 函数中的空集？

Question

I have a dictionary to be stored in csv through pandas:

df = pd.concat([pd.Series(node_dict[k], name=k) for k in HEADERS], 1)
df.to_csv(os.path.join(abspath, outputfile), sep='\t', index=False)

The keys correspond to the columns in the CSV or pandas frame, and the values are a list of sets. Each set is the current row's values. Let's see if I have two columns:

   names                     companies                      
{'john', 'smith', 'mary'}   {'ms', 'fb'} 
 set()                      {'ms', 'fb', 'tw', 'g', 'lk'}
 ...                         ...

Some rows's values are empty, as indicated by the set() printout in the file. I hope there is a way to modify this line:

[pd.Series(node_dict[k], name=k) for k in HEADERS]

to write the invisible '' into the file, instead of the string 'set()'.

Sample of the dict:

node_dict['names'] = [{'john', 'smith', 'mary'}, {}]
node_dict['companies'] = [{'ms', 'fb'}, {'ms', 'fb', 'tw', 'g', 'lk'} ]

Of course the actual lists are much longer in the dictionary.

Answer 1

I think you can do something like:

node_dict = {k: [x if x else "invisible" for x in v] for k,v in node_dict.items()}

prior to doing [pd.Series(node_dict[k], name=k) for k in HEADERS]

Answer 2

You can just drop all the {} . Convert the dict to a string , drop and re-evaluate as dictionary. Done.

df = pd.concat([pd.Series(eval(str(node_dict[k]).replace('{}',' ')), name=k) for k in HEADERS], 1)

df
                 names            companies
0  {john, mary, smith}             {fb, ms}
1                  NaN  {g, ms, lk, fb, tw}

Even works with trailing , in the dictionary. df.to_csv() evaluates the NaN automatically as empty string DataFrame.to_csv(path, sep: str = ',', na_rep: str = ''...)