1. First, take note that the state names are in the index of the DataFrame. These states are correctly placed vertically and do not need to be restructured. It is the column names that are the problem. The stack method takes all of the column names and reshapes them to be vertical as a single index level:

>>> state_fruit.stack()
Texas    Apple      12
         Orange     10
         Banana     40
Arizona  Apple       9
         Orange      7
         Banana     12
Florida  Apple       0
         Orange     14
         Banana    190
dtype: int64

2. Notice that we now have a Series with a MultiIndex. There are now two levels in the index. The original index has been pushed to the left to make room for the old column names. With this one command, we now essentially have tidy data. Each variable, state, fruit, and weight is vertical. Let's use the reset_index method to turn the result into a DataFrame:

>>> state_fruit_tidy = state_fruit.stack().reset_index()
>>> state_fruit_tidy

3. Our structure is now correct, but the column names are meaningless. Let's replace them with proper identifiers:

>>> state_fruit_tidy.columns = ['state', 'fruit', 'weight']
>>> state_fruit_tidy

4. Instead of directly changing the columns attribute, it's possible to use the lesser-known Series method rename_axis to set the names of the index levels before using reset_index:

>>> state_fruit.stack()
               .rename_axis(['state', 'fruit'])

state    fruit 
Texas    Apple      12
         Orange     10
         Banana     40
Arizona  Apple       9
         Orange      7
         Banana     12
Florida  Apple       0
         Orange     14
         Banana    190
dtype: int64

5. From here, we can simply chain the reset_index method with the name parameter to reproduce the output from step 3:

>>> state_fruit.stack()
               .rename_axis(['state', 'fruit'])
               .reset_index(name='weight')