Ways of accessing DataFrame elements, rows, columns, and subsets

[maxim-belkin]

This isn't really an issue, but rather me trying to summarize ways one can access various subsets of a DataFrame in Pandas. Perhaps, it would be of benefit to someone or inspire someone to improve the lesson... :)))

1. Access a single column

   # by name
   data["col_name"]   # as a Series
   data[["col_name"]] # as a DataFrame
   
   # by name using .loc
   data.T.loc["col_name"]  # as a Series
   data.T.loc[["col_name"]].T  # as a DataFrame
   
   # Dot notation (Series)
   data.col_name
   
   # by index (iloc)
   data.iloc[:, col_index]   # as a Series
   data.iloc[:, [col_index]] # as a DataFrame
   
   # using a mask
   data.T[data.T.index == "col_name"].T

2. Access a single row

   # by name using .loc
   data.loc["row_name"] # as a Series
   data.loc[["row_name"]] # as a DataFrame
   
   # by name
   data.T["row_name"] # as a Series
   data.T[["row_name"]].T as a DataFrame
   
   # by index
   data.iloc[row_index]   # as a Series
   data.iloc[[row_index]]   # as a DataFrame
   
   # using mask
   data[data.index == "row_name"]

3. Access an individual DataFrame element

   # by column/row names
   data["column_name"]["row_name"]         # as a Series
   
   data[["col_name"]].loc["row_name"]  # as a Series
   data[["col_name"]].loc[["row_name"]]  # as a DataFrame
   
   data.loc["row_name"]["col_name"]  # as a value
   data.loc[["row_name"]]["col_name"]  # as a Series
   data.loc[["row_name"]][["col_name"]]  # as a DataFrame
   
   data.loc["row_name", "col_name"]  # as a value
   data.loc[["row_name"], "col_name"]  # as a Series. Preserves index. Column name is moved to `.name`.
   data.loc["row_name", ["col_name"]]  # as a Series. Index is moved to `.name.` Sets index to column name.
   data.loc[["row_name"], ["col_name"]]  # as a DataFrame (preserves original index and column name)
   
   # by column/row names: Dot notation
   data.col_name.row_name
   
   # by column/row indices
   data.iloc[row_index, col_index] # as a value
   data.iloc[[row_index], col_index] # as a Series. Preserves index. Column name is moved to `.name` 
   data.iloc[row_index, [col_index]] # as a Series. Index is moved to `.name.` Sets index to column name. 
   data.iloc[[row_index], [col_index]] # as a DataFrame (preserves original index and column name)
   
   # column name + row index
   data["col_name"][row_index]
   data.col_name[row_index]
   data["col_name"].iloc[row_index]
   
   # column index + row name
   data.iloc[:, [col_index]].loc["row_name"]  # as a Series
   data.iloc[:, [col_index]].loc[["row_name"]]  # as a DataFrame
   
   # using masks
   data[data.index == "row_name"].T[data.T.index == "col_name"].T

4. Access several columns

   # by name
   data[["col1", "col2", "col3"]]
   data.loc[:, ["col1", "col2", "col3"]]
   
   # by index
   data.iloc[:, [col1_index, col2_index, col3_index]]

5. Access several rows

   # by name
   data.loc[["row1", "row2", "row3"]]
   
   # by index
   data.iloc[[row1_index, row2_index, row3_index]]

6. Access a subset of specific rows and columns

   # by names
   data.loc[["row1", "row2", "row3"], ["col1", "col2", "col3"]]
   
   # by indices
   data.iloc[[row1_index, row2_index, row3_index], [col1_index, col2_index, col3_index]]
   
   # column names + row indices
   data[["col1", "col2", "col3"]].iloc[[row1_index, row2_index, row3_index]]
   
   # column indices + row names
   data.iloc[:, [col1_index, col2_index, col3_index]].loc[["row1", "row2", "row3"]]

7. Access a subset of row and column ranges

   # by name
   data.loc["row1":"row2", "col1":"col2"]
   
   # by index
   data.iloc[row1_index:row2_index, col1_index:col2_index]
   
   # column names + row indices
   data.loc[:, "col1_name":"col2_name"].iloc[row1_index:row2_index]
   
   # column indices + row names
   data.iloc[:, col1_index:col2_index].loc["row1":"row2"]

[vahtras]

Thanks, these are useful notes!