如何在熊猫中多次提取字符串中存在的单个模式

Question

I have following data

Description
4 GB+ 2 GB Night 3G/2G Data
Unlimited Local & STD Calls + 500 MB 3G/2G Data (T&C apply)
9GB + 8GB night data+ 6GB next night data
8 GB data 4G

What I want is to extract amount of data(4GB etc) and merge them in a single column

df2=df['Description'].str.extract('([0-9]+(\.[0-9][0-9]?)?\s?GB|[0-9]+(\.[0-9][0-9]?)?\s?MB)')

I have used pandas function extractall() too but both extract and extractall() giving me result like this

0     1    2
4GB   Nan  Nan     #2 gb is missing
500MB Nan  Nan   
9GB   Nan  Nan     # 8gb 6 gb is missing
8Gb   Nan  Nan

Where i am wrong? Also when combining the rows with df.fillna(' ') i am getting an output like this

     0 
    4GB,2GB, 
    500MB, , 
    9GB,8GB,6GB
    8GB, , 

though what i want is

0
4GB,2GB
500MB
9GB,8GB,6GB
8GB

I dont want spaces.Is there any way in pandas to get the data in the above format? I am a beginner in python,don't know how to achieve this.If there is another way please mention.

EDIT:

this the full code:

df2=df['Description'].str.extractall('([0-9]+(\.[0-9][0-9]?)?\s?GB|[0-     9]+(\.[0-9][0-9]?)?\s?MB)')
#print df2
df2[1].fillna("",inplace=True);
df2[2].fillna("",inplace=True)
print df2
df3=df2[0]+','+df2[1]+','+df2[2];
print df3

Answer 1

Using extractall should work like below:

df.Description.str\
  .extractall('(\d*\s?[GM]B)').groupby(level=0)\
  .apply(lambda x: ','.join(x[0])\
  .replace(' ',''))
Out[75]: 
0        4GB,2GB
1          500MB
2    9GB,8GB,6GB
3            8GB
dtype: object

Answer 2

df2=df['Description'].str.extractall('(\d*\.\d+|\d+\s?GB|\d*\.\d+|\d+\s?MB)').reset_index()
df2 = pd.pivot_table(df2, index='level_0', columns="match", values=0, aggfunc='last').reset_index(drop=True)
df2 = df2.apply(lambda row: ','.join(row.dropna()), axis=1)

Try this code for your expected output.

Answer 3

You can get the data as columns like this

df2=df['Description'].str.extractall('([0-9]+(\.[0-9][0-9]?)?\s?GB|[0-9]+(\.[0-9][0-9]?)?\s?MB)')
df2.reset_index().groupby('match')[0].apply(lambda x: "{%s}" % ', '.join(x)).apply(lambda x:x.replace(" ",""))

output :

match
0    {4GB,500MB,9GB,8GB}
1              {2GB,8GB}
2                  {6GB}
Name: 0, dtype: object