dplyr vs. pandas

Here’s a markdown table comparing dplyr (R) and pandas (Python) for various data manipulation tasks:

Data frame verbs

1. Rows

   Feature	dplyr (R)	pandas (Python)
   Arrange	arrange(df, col)	df.sort_values(by='col', ascending=True)
   Distinct	distinct(df, col)	df.drop_duplicates(subset='col')
   Filter	filter(df, condition)	df[df['condition']]
   Slice	slice(df, rows)	df.iloc[rows]

2. Columns

   Feature	dplyr (R)	pandas (Python)
   Glimpse	glimpse(df)	df.info()
   Mutate	mutate(df, new_col = func(old_col))	df['new_col'] = df['old_col'].apply(func)
   Pull	pull(df, col)	df['col']
   Rename	rename(df, new_col_name = old_col_name)	df.rename(columns={'old_col_name': 'new_col_name'}, inplace=True)
   Select	select(df, col1, col2, ...)	df[['col1', 'col2', ...]]

3. Groups

   Feature	dplyr (R)	pandas (Python)
   Group By	group_by(df, col)	df.groupby('col')
   Summarise	summarise(df, new_col = func(col))	df.agg({'col': func})

4. Data frames

   Feature	dplyr (R)	pandas (Python)
   bind_cols	bind_cols(df1, df2)	pd.concat([df1, df2], axis=1)
   bind_rows	bind_rows(df1, df2)	pd.concat([df1, df2], axis=0)
   Inner Join	inner_join(df1, df2, by = "key_column")	pd.merge(df1, df2, on='key_column', how='inner')
   Left Join	left_join(df1, df2, by = "key_column")	pd.merge(df1, df2, on='key_column', how='left')
   Right Join	right_join(df1, df2, by = "key_column")	pd.merge(df1, df2, on='key_column', how='right')
   Full Join	full_join(df1, df2, by = "key_column")	pd.merge(df1, df2, on='key_column', how='outer')
   Semi Join	semi_join(df1, df2, by = "key_column")	Not directly available; can use merge and isin together for similar effect.
   Anti Join	anti_join(df1, df2, by = "key_column")	Not directly available; can use merge and isin together for similar effect.

5. Vector functions

   Feature	dplyr (R)	pandas (Python)
   if_else	mutate(df, new_col = if_else(condition, true_val, false_val))	df['new_col'] = np.where(condition, true_val, false_val)
   na_if	mutate(df, col = na_if(col, value))	df['col'].replace(value, np.nan)
   n_distinct	n_distinct(df, col)	df['col'].nunique()
   sample_n	sample_n(df, n)	df.sample(n=n)
   sample_frac	sample_frac(df, fraction)	df.sample(frac=fraction)
   case_when	mutate(df, new_col = case_when(condition1 ~ value1, condition2 ~ value2, ...))	df['new_col'] = np.select([condition1, condition2, ...], [value1, value2, ...], default=default_value)
   cummean	mutate(df, new_col = cummean(col))	df['new_col'] = df['col'].expanding().mean()
   row_number	mutate(df, row_num = row_number())	df['row_num'] = range(1, len(df)+1)
   min_rank	mutate(df, rank = min_rank(col))	df['rank'] = df['col'].rank(method='min')
   dense_rank	mutate(df, rank = dense_rank(col))	df['rank'] = df['col'].rank(method='dense')

   Note that while many functionalities are directly available in both dplyr and pandas, some might require slight variations or custom functions to achieve the same result.