CODEWITHSUNDEEP
pythonpandas
panda
panda

Reputation: 625

How to merge csv files with different headers but same data on condition

I have the following data set. https://drive.google.com/drive/folders/1NRelNsXQJ7MTNKcm-T69N6r5ZsOyFmTS?usp=sharing

For merging all together if column name is same with sheet name as a separate column following is the code

import pandas as pd
import glob
import os
#file directory that contains the csv files
files = glob.glob('/Users/user/Desktop/demo/*.csv')
dfs = [pd.read_csv(fp).assign(SheetName=os.path.basename(fp).split('.')[0]) for fp in files]
data = pd.concat(dfs, ignore_index=True)


data.columns = data.columns.str.lower()

data=data.rename(columns={'sheetname':'Source'})

merged_data = data

data after running the above code

merged_data

id  user    product price[78]   price[79]       Source
105 dummya  egg         22      28.0            sheet1
119 dummy1  soya        67      NaN             sheet1
567 dummya  spinach     22      28.0            sheet2
897 dummy1  rose        67      99.0            sheet2
345 dummya  egg         87      98.0            sheet3
121 dummy1  potato      98      99.0            sheet3
​

How to merge the files on condition? Condition.

Sheet   ID  price1_col1 price1_col2 price1         price2_col1 price2_col2 price2                      sheetname
sheet1  yes     78                   price1_col1     78                    price2_col1                  yes
sheet2  yes     78        79         price1_col1+    78         79         price2_col1+                 yes
                                         price1_col2                           price2_col2
sheet3  yes     78        79         max(price1_col1, 79        78         min(price2_col1,price2_col2) no 
                                         price1_col2)

price 1 on the above snippets points to sheet1 with column name that contains int 78 . if 78+79 means sum the those columns and give name as price1.

output

id  product price1      price2  sheetname
105 egg     22             28       sheet1
119 soya    67                      sheet1
567 spinach 50            28        sheet2
897 rose    166           99        sheet2
345 egg     98            87    
121 potato  99             98

Upvotes: 1

Views: 545

Answers (1)

jezrael
jezrael

Reputation: 862641

Use:

print (merged_data)
    id    user  product  price[78]  price[79]  Source
0  105  dummya      egg         22       28.0  sheet1
1  119  dummy1     soya         67        NaN  sheet1
2  567  dummya  spinach         22       28.0  sheet2
3  897  dummy1     rose         67       99.0  sheet2
4  345  dummya      egg         87       98.0  sheet3
5  121  dummy1   potato         98       99.0  sheet3

print (Condition)
    Sheet   ID  price1_col1  price1_col2                    price1_out  \
0  sheet1  yes           78          NaN                   price1_col1   
1  sheet2  yes           78         79.0       price1_col1+price1_col2   
2  sheet3  yes           78         79.0  max(price1_col1,price1_col2)   

   price2_col1  price2_col2                    price2_out sheetname  
0           78          NaN                   price2_col1       yes  
1           78         79.0       price2_col1+price2_col2       yes  
2           79         78.0  min(price2_col1,price2_col2)        no  

#merge data together by left join    
df = merged_data.merge(Condition.rename(columns={'Sheet':'Source'}), on='Source', how='left')
#replace columns to empty strings, remove sheetname and ID columns
df['Source'] = np.where(df.pop('sheetname') == 'yes', df['Source'], '')
df['id'] = np.where(df.pop('ID') == 'yes', df['id'], '')

#filter integers between [] to ned DataFrame 
df1 = df.filter(regex='\[\d+\]').copy()
#filter all columns with price, exclude df1 
df2 = df[df.filter(regex='price').columns.difference(df1.columns)].copy()
#convert column to integers
df1.columns = df1.columns.str.extract('\[(\d+)\]', expand=False).astype(int)
#helper column for match missing values
df1['a'] = np.nan
#filter columns without/with _out
mask = df2.columns.str.endswith(('_col1','_col2'))
final_cols = df2.columns[ ~mask]
removed_cols = df2.columns[mask]
#replace columns by match values from df2
for c in removed_cols:
    df2[c] = df1.lookup(df1.index, df2[c].fillna('a'))

print (df2)

   price1_col1  price1_col2                    price1_out  price2_col1  \
0           22          NaN                   price1_col1         22.0   
1           67          NaN                   price1_col1         67.0   
2           22         28.0       price1_col1+price1_col2         22.0   
3           67         99.0       price1_col1+price1_col2         67.0   
4           87         98.0  max(price1_col1,price1_col2)         98.0   
5           98         99.0  max(price1_col1,price1_col2)         99.0   

   price2_col2                    price2_out  
0          NaN                   price2_col1  
1          NaN                   price2_col1  
2         28.0       price2_col1+price2_col2  
3         99.0       price2_col1+price2_col2  
4         87.0  min(price2_col1,price2_col2)  
5         98.0  min(price2_col1,price2_col2)

#create MultiIndex for separate eah price groups
df2.columns = df2.columns.str.split('_', expand=True)

def f(x):
    #remove first level
    x.columns = x.columns.droplevel(0)
    out = []
    #loop each row
    for v in x.itertuples(index=False):
        #remove prefix
        t = v.out.replace(x.name+'_', '')
        #loop each namedtuple and replace values
        for k1, v1 in v._asdict().items():
            t = t.replace(k1, str(v1))
        #pd.eval cannot working with min, max, so handled different
        if t.startswith('min'):
            out.append(min(pd.eval(t[3:])))
        elif t.startswith('max'):
            out.append(max(pd.eval(t[3:])))
        #handled +-*/
        else:
            out.append(pd.eval(t))
    #return back
    return pd.Series(out)

#overwrite original columns
df[final_cols] = df2.groupby(level=0, axis=1).apply(f).add_suffix('_out')
#if necessary remove helpers
df = df.drop(removed_cols, axis=1)

print (df)
    id    user  product  price[78]  price[79]  Source  price1_out  price2_out
0  105  dummya      egg         22       28.0  sheet1        22.0        22.0
1  119  dummy1     soya         67        NaN  sheet1        67.0        67.0
2  567  dummya  spinach         22       28.0  sheet2        50.0        50.0
3  897  dummy1     rose         67       99.0  sheet2       166.0       166.0
4  345  dummya      egg         87       98.0                98.0        87.0
5  121  dummy1   potato         98       99.0                99.0        98.0

Upvotes: 1

Related Questions