DRY and "Do One Thing"

Writing Functions in Python

Shayne Miel

Software Architect @ Duo Security

Don't repeat yourself (DRY)

train = pd.read_csv('train.csv')
train_y = train['labels'].values
train_X = train[col for col in train.columns if col != 'labels'].values
train_pca = PCA(n_components=2).fit_transform(train_X)
plt.scatter(train_pca[:,0], train_pca[:,1])

val = pd.read_csv('validation.csv')
val_y = val['labels'].values
val_X = val[col for col in val.columns if col != 'labels'].values
val_pca = PCA(n_components=2).fit_transform(val_X)
plt.scatter(val_pca[:,0], val_pca[:,1])

test = pd.read_csv('test.csv')
test_y = test['labels'].values
test_X = test[col for col in test.columns if col != 'labels'].values
test_pca = PCA(n_components=2).fit_transform(train_X)
plt.scatter(test_pca[:,0], test_pca[:,1])

The problem with repeating yourself

train = pd.read_csv('train.csv')
train_y = train['labels'].values
train_X = train[col for col in train.columns if col != 'labels'].values
train_pca = PCA(n_components=2).fit_transform(train_X)
plt.scatter(train_pca[:,0], train_pca[:,1])

val = pd.read_csv('validation.csv')
val_y = val['labels'].values
val_X = val[col for col in val.columns if col != 'labels'].values
val_pca = PCA(n_components=2).fit_transform(val_X)
plt.scatter(val_pca[:,0], val_pca[:,1])

test = pd.read_csv('test.csv')
test_y = test['labels'].values
test_X = test[col for col in test.columns if col != 'labels'].values
test_pca = PCA(n_components=2).fit_transform(train_X)  ### yikes! ###
plt.scatter(test_pca[:,0], test_pca[:,1])

Another problem with repeating yourself

train = pd.read_csv('train.csv')
train_y = train['labels'].values  ### <- there and there --v ### 
train_X = train[col for col in train.columns if col != 'labels'].values
train_pca = PCA(n_components=2).fit_transform(train_X)
plt.scatter(train_pca[:,0], train_pca[:,1])

val = pd.read_csv('validation.csv')
val_y = val['labels'].values  ### <- there and there --v ### 
val_X = val[col for col in val.columns if col != 'labels'].values
val_pca = PCA(n_components=2).fit_transform(val_X)
plt.scatter(val_pca[:,0], val_pca[:,1])

test = pd.read_csv('test.csv')
test_y = test['labels'].values  ### <- there and there --v ### 
test_X = test[col for col in test.columns if col != 'labels'].values
test_pca = PCA(n_components=2).fit_transform(test_X)
plt.scatter(test_pca[:,0], test_pca[:,1])

Use functions to avoid repetition

def load_and_plot(path):
  """Load a dataset and plot the first two principal components.

  Args:
    path (str): The location of a CSV file.

  Returns:
    tuple of ndarray: (features, labels)
  """
  data = pd.read_csv(path)
  y = data['label'].values
  X = data[col for col in data.columns if col != 'label'].values
  pca = PCA(n_components=2).fit_transform(X)
  plt.scatter(pca[:,0], pca[:,1])
  return X, y

train_X, train_y = load_and_plot('train.csv')

val_X, val_y = load_and_plot('validation.csv')

test_X, test_y = load_and_plot('test.csv')

def load_and_plot(path):
  """Load a dataset and plot the first two principal components.

  Args:
    path (str): The location of a CSV file.

  Returns:
    tuple of ndarray: (features, labels)
  """
  data = pd.read_csv(path)
  y = data['label'].values
  X = data[col for col in data.columns if col != 'label'].values

  pca = PCA(n_components=2).fit_transform(X)
  plt.scatter(pca[:,0], pca[:,1])

  return X, y

def load_and_plot(path):
  """Load a dataset and plot the first two principal components.

  Args:
    path (str): The location of a CSV file.

  Returns:
    tuple of ndarray: (features, labels)
  """
  # load the data
  data = pd.read_csv(path)
  y = data['label'].values
  X = data[col for col in data.columns if col != 'label'].values

  pca = PCA(n_components=2).fit_transform(X)
  plt.scatter(pca[:,0], pca[:,1])

  return X, y

def load_and_plot(path):
  """Load a dataset and plot the first two principal components.

  Args:
    path (str): The location of a CSV file.

  Returns:
    tuple of ndarray: (features, labels)
  """
  # load the data
  data = pd.read_csv(path)
  y = data['label'].values
  X = data[col for col in data.columns if col != 'label'].values

  # plot the first two principal components
  pca = PCA(n_components=2).fit_transform(X)
  plt.scatter(pca[:,0], pca[:,1])

  return X, y

def load_and_plot(path):
  """Load a dataset and plot the first two principal components.

  Args:
    path (str): The location of a CSV file.

  Returns:
    tuple of ndarray: (features, labels)
  """
  # load the data
  data = pd.read_csv(path)
  y = data['label'].values
  X = data[col for col in data.columns if col != 'label'].values

  # plot the first two principle components
  pca = PCA(n_components=2).fit_transform(X)
  plt.scatter(pca[:,0], pca[:,1])

  # return loaded data
  return X, y

Do One Thing

def load_data(path):
  """Load a dataset.

  Args:
    path (str): The location of a CSV file.

  Returns:
    tuple of ndarray: (features, labels)
  """
  data = pd.read_csv(path)
  y = data['labels'].values
  X = data[col for col in data.columns 
           if col != 'labels'].values
  return X, y

def plot_data(X):
  """Plot the first two principal components of a matrix.

  Args:
    X (numpy.ndarray): The data to plot.
  """
  pca = PCA(n_components=2).fit_transform(X)
  plt.scatter(pca[:,0], pca[:,1])

Advantages of doing one thing

The code becomes:

More flexible
More easily understood
Simpler to test
Simpler to debug
Easier to change

Code smells and refactoring

"Any fool can write code that a computer can understand. Good programmers write code that humans can understand." - Martin Fowler (1999)

"Refactoring" by Martin Fowler

Let's practice!

Writing Functions in Python