Default values and sentinel values

A map pairs keys with values. You give it a key of one type, and it hands back the value of a second type that is associated with that key. The source material makes this concrete with a phone book: a name goes in, a telephone number comes out. This article looks at the single moment you ask a map for a value by writing a key in square brackets, an operation called subscripting. The goal is to understand exactly what happens inside that one line of code, including the case where the key you asked about has never been stored. That case is where many learners' expectations go wrong, so it gets the most attention.

Picture the map as a set of pairs. Each key sits next to exactly one value. The diagram below shows two illustrative entries. A lookup starts from the key you supply, finds the pair that holds it, and reads off the value on the other side. The sample names and numbers are invented for illustration; the source describes the structure, not these particular entries.

The function below is the entire lookup described in the source. It takes a string, uses that string as a key into a map called phone_book, and returns whatever value the map gives back. Notice that it does not loop, search, or compare anything by hand. The subscript operation performs the entire search, and the function is a named wrapper around that one operation. This is why the source describes subscripting as essentially the lookup itself.

The subscript has two branches, and the choice between them depends entirely on whether the key is already in the map at the moment the subscript runs. The flowchart below traces both paths from the same starting point. In the first branch the map already holds the key, so nothing is added and the stored value comes back. In the second branch the key is absent, so the map grows by one entry before the default value is returned.

The branch that matters most for surprises is the second one. A learner who expects a missing key to produce only a notification of absence is following a model the source does not describe. Instead, the map creates an entry for the missing key and gives it a default value, and that default is what the lookup returns.

The process of inserting a missing key with a default value is called default value initialization. The default value depends on the type of the value. The source gives the integer case: the default for an integer type is 0. Because the subscript both inserts and returns a value, a lookup of a missing key changes the map as well as producing a result. The diagram below compares the phone book before and after looking up the missing name Ben. The sample entries are illustrative.

When a lookup returns something surprising, the fastest way to find out why is to trace it. The source points to two questions that locate the divergence: was the key present when the subscript ran, and what was stored under that key at that moment? Checking those two facts usually explains the returned value. The worked example below applies both questions to a lookup that a learner might not expect.

A default value that the program uses to mean that no real answer exists is one kind of sentinel value. The source's example is 0 used for an invalid telephone number, a placeholder that signals the program's design rather than a genuine stored fact. The danger is that a sentinel and a real stored value can have the same content. The table below compares them. Notice that the value alone may not reveal which one you are looking at.