Skip to content

Instantly share code, notes, and snippets.

@osiemomatrixx
Forked from VankatPetr/ExcelCompare.py
Created September 30, 2021 19:41
Show Gist options
  • Save osiemomatrixx/26aaf408d469480ad679a12c10f707ef to your computer and use it in GitHub Desktop.
Save osiemomatrixx/26aaf408d469480ad679a12c10f707ef to your computer and use it in GitHub Desktop.
Compare 2 Excel files and create an Excel diff using Python
import pandas as pd
from pathlib import Path
#define parameters
#path to files
path_old=Path(r'C:\Users\owner\Documents\old.xlsx')
path_new=Path(r'C:\Users\owner\Documents\new.xlsx')
#list of key column(s)
key=['id']
#sheets to read in
sheet='Sheet1'
# Read in the two excel files and fill NA
old = pd.read_excel(path_old).fillna(0)
new = pd.read_excel(path_new).fillna(0)
#set index
old=old.set_index(key)
new=new.set_index(key)
#identify dropped rows and added (new) rows
dropped_rows = set(old.index) - set(new.index)
added_rows = set(new.index) - set(old.index)
#combine data
df_all_changes = pd.concat([old, new], axis='columns', keys=['old','new'], join='inner')
#prepare functio for comparing old values and new values
def report_diff(x):
return x[0] if x[0] == x[1] else '{} ---> {}'.format(*x)
#swap column indexes
df_all_changes = df_all_changes.swaplevel(axis='columns')[new.columns[0:]]
#apply the report_diff function
df_changed = df_all_changes.groupby(level=0, axis=1).apply(lambda frame: frame.apply(report_diff, axis=1))
#create a list of text columns (int columns do not have '{} ---> {}')
df_changed_text_columns = df_changed.select_dtypes(include='object')
#create 3 datasets:
#diff - contains the differences
#dropped - contains the dropped rows
#added - contains the added rows
diff = df_changed_text_columns[df_changed_text_columns.apply(lambda x: x.str.contains("--->") == True, axis=1)]
dropped = old.loc[dropped_rows]
added = new.loc[added_rows]
#create a name for the output excel file
fname = '{} vs {}.xlsx'.format(path_old.stem, path_new.stem)
#write dataframe to excel
writer=pd.ExcelWriter(fname, engine='xlsxwriter')
diff.to_excel(writer, sheet_name='diff', index=True)
dropped.to_excel(writer, sheet_name='dropped', index=True)
added.to_excel(writer, sheet_name='added', index=True)
#get xlswriter objects
workbook = writer.book
worksheet = writer.sheets['diff']
worksheet.hide_gridlines(2)
worksheet.set_default_row(15)
#get number of rows of the df diff
row_count_str=str(len(diff.index)+1)
#define and apply formats
highligt_fmt = workbook.add_format({'font_color': '#FF0000', 'bg_color':'#B1B3B3'})
worksheet.conditional_format('A1:ZZ'+row_count_str, {'type':'text', 'criteria':'containing', 'value':'--->',
'format':highligt_fmt})
#save the output
writer.save()
print ('\nDone.\n')
@osiemomatrixx
Copy link
Author

this is awesome

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment