Python Pandas DataFrame cell changes disappear

Question

I'm new to python and pandas and I'm trying to manipulate a csv data file. I load two dataframes one contains a column with keywords and the other is a "bagOfWords" with "id" and "word" columns. What i whant to do is to add a column to the first dataframe with the ids of the keywords in a "list string" like so "[1,2,8,99 ...]".

This is what i have come up with so far

websitesAlchData = pd.io.parsers.read_csv('websitesAlchData.csv', sep=';', index_col='referer', encoding="utf-8")

bagOfWords = pd.io.parsers.read_csv('bagOfWords.csv', sep=';', header=0, names=["id","words","count"], encoding="utf-8")
a = set(bagOfWords['words'])
websitesAlchData['keywordIds'] = "[]"
for i in websitesAlchData.index
    keywords = websitesAlchData.loc[i,'keywords']
    try:
        keywordsSet = set([ s.lower() for s in keywords.split(",") ])
    except:
        keywordsSet = set()
    existingWords = a & keywordsSet
    lista = []
    for i in bagOfWords.index:
        if bagOfWords.loc[i,'words'] in existingWords:
            lista.append(bagOfWords.loc[i,'id'])

    websitesAlchData.loc[i,'keywordIds'] = str(lista)
    print(str(lista))
    print(websitesAlchData.loc[i,'keywordIds'])
websitesAlchData.reset_index(inplace=True)
websitesAlchData.to_csv(path_or_buf = 'websitesAlchDataKeywordCode.csv', index=False, sep=";", encoding="utf-8")

The two prints at the end of the for loop give the excpected results but when I try to print the whole dataframe "websitesAlchData" the column "keywordIds" is still "[]" and so it is in the resulting .csv as well.

My guess would be that i create a copy somewhere but i can't se where.

Any ideas what is wrong here or how to do the same thing diffrently? Thanks!

UPDATE:

The websitesAlchData.cvs looks like this

referer;category;keywords
url;int;word0,word2,word3
url;int;word1,word3
...

And the bag of words cvc.

id;index;count
0;word0;11
1;word1;14
2;word2;14
3;word3;14
...

Expected output

referer;category;keywords;keywordIds
url;int;word0,word2,word3;[0,2,3]
url;int;word1,word3;[1,3]

Answer 1

there's definitely something wrong with using i for both for loops. change that and see if that helps.

Answer 2

I'd try something like this. You'll want to profile the performance on the larger dataset.

In [146]: df1
Out[146]: 
  referer category           keywords
0     url      int  word0,word2,word3
1     url      int        word1,word3

[2 rows x 3 columns]

In [147]: df2
Out[147]: 
       id  count
index           
word0   0     11
word1   1     14
word2   2     14
word3   3     14

[4 rows x 2 columns]

Split the keywords column into a list of words. Generally storing lists in DataFrames is a bad idea performance wise, but this is the most straightforward way for now.

In [148]: vals = df1.keywords.str.split(',')

In [149]: vals
Out[149]: 
0    [word0, word2, word3]
1           [word1, word3]
Name: keywords, dtype: object

Then apply a lookup from df2 to each element of the lists in vals :

In [151]: ids = vals.apply(lambda x: [df2.loc[y, 'id'] for y in x])

In [152]: ids
Out[152]: 
0    [0, 2, 3]
1       [1, 3]
Name: keywords, dtype: object

Finally concat:

In [154]: df = pd.concat([df1, ids], axis=1)

In [155]: df
Out[155]: 
  referer category           keywords   keywords
0     url      int  word0,word2,word3  [0, 2, 3]
1     url      int        word1,word3     [1, 3]

[2 rows x 4 columns]