从熊猫中的1个CSV文件读取多个表

Question

suppose I have a csv file like this:

Name: Jack
Place: Binghampton
Age:27
Month,Sales,Revenue
Jan,51,$1000
Feb,20,$1050
Mar,100,$10000
### Blank File Space
### Blank File Space
Name: Jill
Place: Hamptonshire
Age: 49
Month,Sales,Revenue
Apr,11,$1000
May,55,$3000
Jun,23,$4600
### Blank File Space
### Blank File Space
...

And the contents of the file are evenly spaced as shown. I want to read each Month,Sales,Revenue portion in as its own df. I know I can do this manually by doing:

df_Jack = pd.read_csv('./sales.csv', skiprows=3, nrows=3)
df_Jill = pd.read_csv('./sales.csv', skiprows=12, nrows=3)

I'm not even super worried about the names of the df as I think I could do that on my own, I just don't really know how to iterate through the evenly spaced file to find sales records and store them as unique dfs.

Thanks for any help in advance!

Answer 1

How about create a list of dfs?

from io import StringIO

csvfile = StringIO("""Name: Jack
Place: Binghampton
Age:27
Month,Sales,Revenue
Jan,51,$1000
Feb,20,$1050
Mar,100,$10000
### Blank File Space
### Blank File Space
Name: Jill
Place: Hamptonshire
Age: 49
Month,Sales,Revenue
Apr,11,$1000
May,55,$3000
Jun,23,$4600
### Blank File Space
### Blank File Space""")

df = pd.read_csv(csvfile, sep=',', error_bad_lines=False, names=['Month','Sales','Revenue'])

df1 = df.dropna().loc[df.Month!='Month']

listofdf = [df1[i:i+3] for i in range(0,df1.shape[0],3)]

print(listofdf[0])

Output:

  Month Sales Revenue
4   Jan    51   $1000
5   Feb    20   $1050
6   Mar   100  $10000

print(listofdf[1])

Output:

   Month Sales Revenue
13   Apr    11   $1000
14   May    55   $3000
15   Jun    23   $4600

Answer 2

Obviously you could do this:

dfs = [pd.read_csv('./sales.csv', skiprows=i, nrows=3) for i in range(3, n, 9)]
# where n is your expected end line...

But another way is to read the csv yourself and pass the data back to pandas :

with open('./sales.csv', 'r') as file:
    streaming = True
    while streaming:
        name = file.readline().rstrip().replace('Name: ','')
        for _ in range(2): file.readline()
        headers = file.readline().rstrip().split(',')
        data = [file.readline().rstrip().split(',') for _ in range(3)]
        dfs[name] = pd.DataFrame.from_records(data, columns=headers)
        for _ in range(2):
            streaming = file.readline()

I'll concede it's quite brutish and inelegant compared to the other answer... but it works. And it actually gives you the DataFrame by name within a dictionary:

>>> dfs['Jack']

  Month Sales Revenue
0   Jan    51   $1000
1   Feb    20   $1050
2   Mar   100  $10000
>>> dfs['Jill']

  Month Sales Revenue
0   Apr    11   $1000
1   May    55   $3000
2   Jun    23   $4600