From Survey Records to X and y

A garden centre asked its customers to fill in a survey. Every answer arrived as a dictionary (a record), for example {"age": 34, "visits": 5, "garden_m2": 120, "bought_tools": "yes"}. Before any model can learn from them, the records must become a feature table X (a list of rows) and a label list y.

Write to_xy(records, features, label) that returns the tuple (X, y):

• X[i] is the list of the values of the keys in features, in the order given by features;
• y[i] is the value of label for the same record;
• a record is skipped if any of those keys (a feature or the label) is missing from it or has the value None. Other keys in a record are ignored.

The records keep their original order, and records must not be changed.

Examples

Input:  records = [{"age": 34, "visits": 5, "bought": "yes"},
                   {"age": 51, "visits": None, "bought": "no"},
                   {"visits": 2, "age": 19, "bought": "no", "city": "Leeds"}]
        features = ["visits", "age"], label = "bought"
Output: ([[5, 34], [2, 19]], ["yes", "no"])
Explanation: the second record has no value for "visits", so it is skipped;
the columns follow the order of `features`, not of the record.

Input:  records = [{"a": 1}, {"a": 2, "y": 0}], features = ["a"], label = "y"
Output: ([[2]], [0])

Constraints

• 0 <= len(records) <= 10**4, 1 <= len(features) <= 20
• label is not one of features
• values are numbers or strings; 0, "" and False are real values, only a missing key or None makes a record incomplete

Goals

• Turn a list of records into a feature table X and a label list y
• Keep X and y in step: row i of X always belongs to label i of y
• Drop incomplete examples instead of guessing values