Recursion

"Recursion is the root of computation since it trades description for time." --Alan Perlis (See https://www.cs.yale.edu/homes/perlis-alan/quotes.html.) Let's write some recursive functions!

Refresher

Anonymous functions and fixpoints

Recall our definition of pred, the predecessor function on nats.

In addition to the Check command to see an expression's type and the Compute command to see how it evaluates, there's a Print command to see something's definition. Let's try it.

Huh, weird! We defined pred with the Definition command, but Coq used a fun. It turns out that these notions are equivalent. That is, writing:
  Definition foo (x:T) (y:U) : Z := blah.

is just the same as:
  Definition foo : T → U → Z := fun (x:T) (y:U) : Z ⇒ blah.

which is just the same as:
  Definition foo (x:T) : U → Z := fun (y:U) ⇒ blah.

It's very useful to know equivalences in your programming language! It's also important to know the various forms of syntax. A Definition takes the form:
  Definition NAME ARGS× : RETURN TYPE := EXPRESSION.

where NAME is an identifier and ARGS× zero or more arguments (optionally, wrapped in parentheses and annotated with types). (It's conventional in computer science to use * to mean "zero or more". The name comes from the 20th Century theorist Stephen Cole Kleene.) The syntax for anonymous functions is similar:
  fun ARGS+ ⇒ EXPRESSION

Where the + means one or more. For Definitions return types are optional and for funs they are forbidden; a Definition need not have arguments, but anonymous functions need them. Finally---we introduced some of these concepts without such a nice description yesterday. Hopefully things are clearer now! What about recursive functions? I briefly showed you evenb'... let's look at the recursive double function again, in a bit more detail:

Rather than writing fun, it says fix double. Interesting! After that, it looks mostly like a fun, though there's a return type, too. Here's another equivalence, where ARGS+ means one or more arguments:
  Fixpoint NAME ARGS+ : RETURN TYPE := EXPRESSION.

is just the same as:
  Definition NAME := fix NAME ARGS+ : RETURN TYPE := EXPRESSION.

Exercise: 2 stars, standard (half)

Define a function half : nat → nat that divides a number by two, rounding down.

Exercise: 2 stars, standard (third)

Define a function third : nat → nat that divides a number by three, rounding down.

Case study: binary and trinary numbers

Exercise: 3 stars, standard (binary)

Consider a different, more efficient representation of natural numbers using a binary rather than unary system. That is, instead of saying that each natural number is either zero or the successor of a natural number, we can say that each binary number is either

• zero,
• twice a binary number, or
• one more than twice a binary number.

How does this work? Let's compare some binary and unary numbers. Zero is easy:

One isn't so bad either. To get 1 in unary, just add an S. To get it in binary, well, what could we do? If we simply enumerate possibilities, the next up after BZ itself are T₂ BZ and T2P₁. T₂ BZ is "twice BZ", i.e., "twice zero", i.e. "zero". Not what we want. T2P₁ BZ is "twice BZ plus one", i.e., "zero plus one", i.e., "one". Perfect!

For the number two, the nats do the same thing: just add an S. Noting that 2 = 2 * 1, we have our answer for binary, too.

Let's do three.

Let's skip a few. What about the number nine?

Eleven?

You may be wondering... what do T₂ and T2P₁ and BZ have to do with, you know, binary? 0s and 1s? To see, take a binary number and working from BZ out, write 1 for T2P₁ and then 0 for T₂.

• b_one: 1
• b_two: 10
• b_three: 11
• b_nine: 1001
• b_eleven: 1011

These are indeed the binary numbers! (a) First, write an increment function incr for binary numbers, and a function bin_to_nat to convert binary numbers to unary numbers. (b) Write five unit tests for your increment and binary-to-unary functions. (A "unit test" in Coq is just a Compute that you can quickly check.) Notice that incrementing a binary number and then converting it to unary should yield the same result as first converting it to unary and then incrementing. You only need to write five tests total, but please be certain to test both functions.

Exercise: 4 stars, standard (binary_inverse)

This exercise is a continuation of the previous exercise about binary numbers. It's a bit harder. (a) Next, write a function nat_to_bin that converts natural numbers to binary numbers. (b) You might think that starting with a binary number, converting to a natural, and then back to binary yields the same number we started with. However, this is not true! Explain what the problem is. (c) Define a "direct" normalization function -- i.e., a function normalize from binary numbers to binary numbers such that, for any binary number b, converting to a natural and then back to binary yields (normalize b). It's okay if you can't complete this problem, but the autograder will reject your submission if you don't at least define nat_to_bin and normalize at the right type. You can write Definition nat_to_bin : TYPE. Admitted., but you'll have to figure out what TYPE is. If you're stuck... ask for help!

Exercise: 3 stars, standard (trinary)

We saw above how to write a type definition for binary numbers. Now write one to represent trinary numbers, i.e., numbers made up of the digits 0, 1, and 2. Be careful to use fresh names (O and S and BZ and T₂ and T2P₁ are all taken!). Call your type trinary.

Define a function trin_to_nat to convert a trinary number to a unary nat.

For extra glory (but no credit), write nat_to_trin. You'll need some helper functions!

What's happening here?

You might be thinking, "What is this crap? I wanted to learn computer programming! Coq/Gallina is a stupid language with stupid chicken names." A few thoughts. 1. This is a computer science department, not a computer programming department. You'll learn to program because it's a critical skill for computer science, but we're not in it just for the programming. (Similarly, astronomers learn to use cool telescopes, and you could hardly have modern astronomy without telescopes... but astronomy isn't just about telescopes.) 2. Recursion is fundamental to computer science; many core concepts are expressed recursively. You'll have pleny of time to write iterative loops. This is practice! 3. We're trying to show you something essential about how computation and data interrelate, and recursion lets us highlight that. 4. Most people learn iterative languages early on. Studying recursion lets us level the playing field. 5. In the latter two thirds of the course, we'll be proving things about our programs. It's MUCH easier to prove things about recursive programs than iterative ones.