Thank you for the PR @maliozer !

I would prefer this be defined during the "Using the Iris dataset" section in line 131:

    >>> from sklearn.datasets import load_iris
    >>> from sklearn import tree
    >>> iris = load_iris()
    >>> X, y = iris.data, iris.target
    >>> clf = tree.DecisionTreeClassifier()
    >>> clf = clf.fit(X, y)

The reason I chose not to add it at the Using the Iris dataset section, is because it would be meaningless why the iris datasets is loaded twice until you don't apply the graphviz example. The above example demonstrates the return_X_y parameter usage. I tried to leave the example as it is.

For this specific case, I think it is nice to show the complete usage of the load_iris() Bunch object. The first part

    >>> from sklearn.datasets import load_iris
    >>> from sklearn import tree
    >>> iris = load_iris()
    >>> X, y = iris.data, iris.target
    >>> clf = tree.DecisionTreeClassifier()
    >>> clf = clf.fit(X, y)

shows how to extract the data and target from the iris Bunch object. Later, the graphviz example part shows how to get the feature names from the iris Bunch object.

Side note: Most of the user guide uses load_iris(return_X_y=True), so having an example with the default return_X_y=False would be beneficial.