Jerky For Dogs Solution¶

Here we build the plot, then make it better and better in subsequent "drafts".

Draft 1Draft 2Draft 3Draft 4

import matplotlib.pyplot as plt

# Instantiate a figure with 4 Axes (2 rows, 2 columns)
# Use tuple unpacking to get each Axes object
fig, ((ax1, ax2), (ax3, ax4)) = plt.subplots(nrows=2, ncols=2, figsize=(8,4))
fig.suptitle('Sales by date')

ax1.plot(dates[flavors == 'traditional'], sales[flavors == 'traditional'])
ax1.set_title('traditional')

ax2.plot(dates[flavors == 'salsa'], sales[flavors == 'salsa'])
ax2.set_title('salsa')

ax3.plot(dates[flavors == 'sweet orange'], sales[flavors == 'sweet orange'])
ax3.set_title('sweet orange')

ax4.plot(dates[flavors == 'smokey'], sales[flavors == 'smokey'])
ax4.set_title('smokey')

Instantiate a figure with a 2x2 grid of subplots. (See pyplot.subplots.)
```
import matplotlib.pyplot as plt

fig, ((ax1, ax2), (ax3, ax4)) = plt.subplots(nrows=2, ncols=2, figsize=(8,4))
```
plt.subplots(2, 2, figsize=(8,4)) returns two things:
- a figure object
- a 2x2 numpy array of Axes objects.
We "unpack" all five of these objects using fig and a tuple of tuples of ax variables.
Subplot arrangement

The axes in this example are arranged in a grid like this:
```
ax1 ax2
ax3 ax4
```
Add a title to the entire figure.
```
fig.suptitle('Sales by date')
```
Plot the data on each Axes.
```
ax1.plot(dates[flavors == 'traditional'], sales[flavors == 'traditional']) # (1)!
ax1.set_title('traditional') # (2)!
```
1. Subset dates and sales by flavors == 'traditional', then make a simple line plot from this data using Axes.plot().
2. Give the Axes a title with Axes.set_title().
Repeat this for each (flavor, Axes) pair.

Issue
The y axes for traditional jerky ranges from 0 to 150, but the y axis for salsa jerky ranges from 0 to 1500! It's unfair to compare these flavors until they're on a consistent y scale. (The x axes suffers same issue.) We'll fix this in the next draft.

import matplotlib.pyplot as plt

# Instantiate a figure with 4 Axes (2 rows, 2 columns)
# Use tuple unpacking to get each Axes object
fig, ((ax1, ax2), (ax3, ax4)) = plt.subplots(2, 2, figsize=(8,4), sharex=True, sharey=True)
fig.suptitle('Sales by date')

ax1.plot(dates[flavors == 'traditional'], sales[flavors == 'traditional'])
ax1.set_title('traditional')

ax2.plot(dates[flavors == 'salsa'], sales[flavors == 'salsa'])
ax2.set_title('salsa')

ax3.plot(dates[flavors == 'sweet orange'], sales[flavors == 'sweet orange'])
ax3.set_title('sweet orange')

ax4.plot(dates[flavors == 'smokey'], sales[flavors == 'smokey'])
ax4.set_title('smokey')

We add sharex=True, sharey=True parameters to plt.subplots(). This tells Matplotlib to use the same x scale and y scale for every Axes object.

Issue
The layout's a bit messy and the x tick labels are unreadable. We'll fix this in the next draft.

import matplotlib.pyplot as plt

# Instantiate a figure with 4 Axes (2 rows, 2 columns)
# Use tuple unpacking to get each Axes object
fig, ((ax1, ax2), (ax3, ax4)) = plt.subplots(2, 2, figsize=(8,4), sharex=True, sharey=True, layout='tight')
fig.suptitle('Sales by date')

ax1.plot(dates[flavors == 'traditional'], sales[flavors == 'traditional'])
ax1.set_title('traditional')

ax2.plot(dates[flavors == 'salsa'], sales[flavors == 'salsa'])
ax2.set_title('salsa')

ax3.plot(dates[flavors == 'sweet orange'], sales[flavors == 'sweet orange'])
ax3.set_title('sweet orange')

ax4.plot(dates[flavors == 'smokey'], sales[flavors == 'smokey'])
ax4.set_title('smokey')

# Adjust the x label formats
fig.autofmt_xdate()

We add layout='tight' to plt.subplots() to reduce the whitespace around the grid of Axes.
We add fig.autofmt_xdate() to automatically rotate the x tick labels so they're not overlapping.

Issue
There's still a lot of manual code repetition. We should follow the Don't Repeat Yourself (DRY) principle of coding.

import matplotlib.pyplot as plt

# Instantiate a figure with 4 Axes (2 rows, 2 columns)
fig, axs = plt.subplots(2, 2, figsize=(8,4), sharex=True, sharey=True, layout='tight')
fig.suptitle('Sales by date')

# Loop over Axes, flavors
unique_flavors = np.unique(flavors)
for flav, ax in zip(unique_flavors, axs.ravel()):
    ax.plot(dates[flavors == flav], sales[flavors == flav])
    ax.set_title(flav)

# Adjust the x label formats
fig.autofmt_xdate()

Instead of unpacking the Axes objects into four variables (ax1, ax2, ax3, ax4), we store all four Axes in a single variable (axs). Keep in mind, axs is a NumPy ndarray.
```
type(axs)
# <class 'numpy.ndarray'>

axs
# array([[<AxesSubplot:>, <AxesSubplot:>],
#        [<AxesSubplot:>, <AxesSubplot:>]], dtype=object)
```

Next we get the unique flavors from the flavors array.

unique_flavors = np.unique(flavors)

print(unique_flavors)
# ['salsa' 'smokey' 'sweet orange' 'traditional']

Then we iterate through each (flavor, Axes) pair, drawing the line plot for that flavor and setting the Axes title.

for flav, ax in zip(unique_flavors, axs.ravel()): # (1)!
    ax.plot(dates[flavors == flav], sales[flavors == flav])
    ax.set_title(flav)

axs.ravel() flattens the (2,2) array into a (4,) array.

print(axs)
# [[<AxesSubplot:> <AxesSubplot:>]
#  [<AxesSubplot:> <AxesSubplot:>]]

print(axs.ravel())
# [<AxesSubplot:> <AxesSubplot:> <AxesSubplot:> <AxesSubplot:>]

We could also use axs.flatten(), but flatten() returns a copy of the array whereas ravel() returns a view of the array.

See the problem