Dafny Tutorial
Dafny Tutorial
0. Introduction
Dafny is a language that is designed to make it easy to write correct code. This means
correct in the sense of not having any runtime errors, but also correct in actually doing
what the programmer intended it to do. To accomplish this, Dafny relies on high-level
annotations to reason about and prove correctness of code. The effect of a piece of code
can be given abstractly, using a natural, high-level expression of the desired behavior,
which is easier and less error prone to write. Dafny then generates a proof that the code
matches the annotations (assuming they are correct, of course!). Dafny lifts the burden
of writing bug-free code into that of writing bug-free annotations. This is often easier
than writing the code, because annotations are shorter and more direct. For example, the
following fragment of annotation in Dafny says that every element of the array is strictly
positive:
This says that for all integers k that are indices into the array, the value at that index is
greater than zero. By writing these annotations, one is confident that the code is correct.
Further, the very act of writing the annotations can help one understand what the code is
doing at a deeper level.
In addition to proving a correspondence to user supplied annotations, Dafny proves
that there are no run time errors, such as index out of bounds, null dereferences, division
by zero, etc. This guarantee is a powerful one, and is a strong case in and of itself for
the use of Dafny and tools like it. Dafny also proves the termination of code, except in
specially designated loops.
These lecture notes give a guide to Dafny, explaining the specification features that
get used routinely in all verification tasks. Although the lecture notes focus on Dafny, the
concepts explained apply to program specification and verificaton more generally.
2 J. Koenig and K.R.M. Leino / A Guide to Dafny
The appendix of these lecture notes contains a reference guide to a large subset of
the Dafny language. Turn to it if you need to know more about the syntax or constructs
available.
The best way to learn from these notes is to try things out yourself. There are three
ways to run the Dafny verifier. The slickest way is to run it in the Microsoft Visual
Studio integrated development environment, since this continuously runs the verifier in
the background and highlights errors in your program as you develop it. Another way is
to run Dafny from the command line, which requires a .NET virtual machine (built in
on Windows platforms, and provided by Mono on non-Windows platforms). The third
way to run Dafny uses any web browser and requires no installation.0 More information
about these options and installation are available from the Dafny homepage.1
Let’s get started writing some Dafny programs.
1. Methods
In many ways, Dafny is a typical imperative programming language. There are methods,
variables, types, loops, if statements, arrays, integers, and more. One of the basic units
of any Dafny program is the method. A method is a piece of imperative, executable
code. In other languages, they might be called procedures, or functions, but in Dafny the
term “function” is reserved for a different concept that we will cover later. A method is
declared in the following way:
This declares a method called “Abs” which takes a single integer parameter, called “x”,
and returns a single integer, called “r”. Note that the types are required for each param-
eter and return value, and follow each name after a colon (“:”). Also, the return values
(out parameters) are named, and there can be multiple return values, as in below:
The method body is the code contained within the braces, which until now has been
cleverly represented as “ . . . ” (which is not Dafny syntax). The body consists of a series
of statements, such as the familiar imperative assignments, if statements, loops, other
method calls, return statements, etc. For example, the MultipleReturns method may be
implemented as:
0 rise4fun.com
1 research.microsoft.com/dafny
J. Koenig and K.R.M. Leino / A Guide to Dafny 3
Assignments do not use “=”, but rather “:=”. (In fact, as Dafny uses “==” for equality,
there is no use of a single equals sign in Dafny statements and expressions.) Simple
statements must be followed by a semicolon, and whitespace and comments are ignored.
To return a value from a method, the value is assigned to one of the named return values
sometime before the method returns. In fact, the return values act very much like local
variables, and can be assigned to more than once. The input parameters, however, are
read only. Return statements are used when one wants to return before reaching the end of
the body block of the method. Return statements can be just the return keyword (where
the current value of the out parameters are used), or they can take a list of values to
return. There are also compound statements, such as if statements. if statements require
parentheses around the boolean condition, and act as one would expect:
One caveat is that they always need braces around the branches, even if the branch
only contains a single statement (compound or otherwise). Here the if statement checks
whether x is less than zero, using the familiar comparison operator syntax, and sets r to
the absolute value as appropriate. (Other comparison operators are <=, >, >=, !=, and ==,
with the expected meaning.) Here we use the implicit return at the end of the function.
Regardless of the branch taken, the correct value is put into r, so when we get to the end
of the function, the result is returned properly.
None of what we have seen so far has any annotations: the code could be written in
virtually any imperative language (with appropriate considerations for multiple return
values). The real power of Dafny comes from the ability to annotate these methods to
specify their behavior. For example, one property that we observe with the Abs method
is that the result is always greater than or equal to zero, regardless of the input. We could
put this observation in a comment, but then we would have no way to know whether
the method actually had this property. Further, if someone came along and changed the
method, we wouldn’t be guaranteed that the comment was changed to match. With anno-
4 J. Koenig and K.R.M. Leino / A Guide to Dafny
tations, we can have Dafny prove that the property we claim of the method is true. There
are several ways to give annotations, but some of the most common, and most basic, are
method pre- and postconditions (see also, e.g., [10,19,20,9,3,1]).
This property of the Abs method, that the result is always non-negative, is an exam-
ple of a postcondition: it is something that is true after the method returns. Postcondi-
tions, declared with the ensures keyword, are given as part of the method’s declaration,
after the return values (if present) and before the method body. The keyword is followed
by the boolean expression and a semicolon, which are both required. Like an if or while
condition and most annotations, a postcondition is always a boolean expression: some-
thing that can be false or true. In the case of the Abs method, a reasonable postcondition
is the following:
You can see here why return values are given names. This makes them easy to refer to in
the postcondition of a method. When the expression is true, we say that the postcondition
holds. The postcondition must hold for every possible invocation of the function, and for
every possible return point (including the implicit one at the end of the function body).
In this case, the only property we are expressing is that the return value is always at least
zero.
Sometimes there are multiple properties that we would like to establish about our
code. In this case, we have two options. We can either join the two conditions together
with the boolean and operator (&&), or we can write multiple ensures annotations. The
latter is the same as the former, but it helps make distinct properties clearer. For example,
the return value names from the MultipleReturns method might lead one to guess the
following postconditions:
or even:
because of the chaining comparison operator syntax in Dafny. (In general, most of the
comparison operators can be chained, but only “in one direction”, i.e., not mixing greater
than and less than; see Section A.8.)
The first way of expressing the postconditions separates the “less” part from the
“more” part, which may be desirable. Another thing to note is that we have included one
of the input parameters in the postcondition. This is useful because it allows us to relate
the input and output of the method to one another (this works because input parameters
are read only, and so are the same at the end as they were at the beginning).
Dafny actually rejects this program, claiming that the first postcondition does not
hold (i.e., is not true). This means that Dafny wasn’t able to prove that this annotation
holds every time the method returns. In general, there are two main causes for Dafny
verification errors: annotations that are inconsistent with the code, and situations where
it is not “clever” enough to prove the required properties. Differentiating between these
two possibilities can be a difficult task, but fortunately, Dafny and the Boogie/Z3 system
[0,18,5] on which it is based [14] are pretty smart, and will prove matching code and
annotations with a minimum of fuss.
In this situation, Dafny is correct in saying there is an error with the code. The
key to the problem is that y is an integer, so it can be negative. If y is negative (or
zero), then more can actually be smaller than or equal to x. Our method will not work as
intended unless y is strictly larger than zero. This is precisely the idea of a precondition.
A precondition is similar to a postcondition, except that it is something that must be true
before a method is called. When you call a method, it is your job to establish (make
true) the preconditions, something Dafny will enforce using a proof. Likewise, when
you write a method, you get to assume the preconditions, but you must establish the
postconditions. The caller of the method then gets to assume that the postconditions hold
after the method returns.
Preconditions have their own keyword, requires. We can give the necessary pre-
condition to MultipleReturns as below:
Like postconditions, multiple preconditions can be written either with the boolean and
operator (&&), or by multiple requires keywords. Traditionally, requires annotations
precede ensures annotations in the source code, though this is not strictly necessary.
(The order of the requires and ensures annotations with respect to others of the same
type can sometimes matter, as we will see later.) With the addition of this condition,
Dafny now verifies the code as correct, because this assumption is all that is needed to
guarantee the code in the method body is correct.
Exercise 0 Write a method Max that takes two integer parameters and returns their max-
imum. Add appropriate annotations and make sure your code verifies.
6 J. Koenig and K.R.M. Leino / A Guide to Dafny
Not all methods necessarily have preconditions. For example, the Abs method we
have already seen is defined for all integers, and so has no preconditions (other than the
trivial requirement that its argument is an integer, which is enforced by the type system).
Even though it has no need of preconditions, the Abs function as it stands now is not
very useful. To investigate why, we need to make use of another kind of annotation, the
assertion.
3. Assertions
method Testing()
{
assert 2 < 3;
}
Dafny proves this method correct, as 2 is always less than 3. Asserts have several uses,
but chief among them is checking whether your expectations of what is true at various
points is actually true. You can use this to check basic arithmetical facts, as above, but
they can also be used in more complex situations. Assertions are a powerful tool for
debugging annotations, by checking what Dafny is able to prove about your code. For
example, we can use it to investigate what Dafny knows about the Abs function.
To do this, we need one more concept: local variables. Local variables behave ex-
actly as you would expect, except maybe for a few issues with shadowing. Local vari-
ables are declared with the var keyword, and can optionally have type declarations. Un-
like method parameters, where types are required, Dafny can infer the types of local
variables in almost all situations. This is an example of an initialized, explicitly typed
variable declaration:
var x: int := 5;
var x := 5;
Explicit type declarations apply only to the immediately preceding variable, so here the
bool declaration applies only to z, and not x or y, which are both inferred to be of type
int. We needed variables because we want to talk about the return value of the Abs
method. We cannot put Abs inside a annotations directly, as the method could change
J. Koenig and K.R.M. Leino / A Guide to Dafny 7
memory state, among other problems. So we capture the return value of a call to Abs as
follows:
This is an example of a situation where we can ask Dafny what it knows about the values
in the code, in this case v. We do this by adding assertions, like the one above. Every time
Dafny encounters an assertion, it tries to prove that the condition holds for all executions
of the code. In this example, there is only one control path through the method, and
Dafny is able to prove the annotation easily because it is exactly the postcondition of the
Abs method. Abs guarantees that the return value is non-negative, so it trivially follows
that v, which is this value, is non-negative after the call to Abs.
Exercise 1 Write a test method that calls your Max method from Exercise 0 and then
asserts something about the result.
But we know something stronger about the Abs method. In particular, for non-
negative x, Abs(x) == x. Specifically, in the above program, the value of v is 3. If we try
adding an assertion (or changing the existing one) to say:
assert v == 3;
we find that Dafny cannot prove our assertion, and gives an error. The reason this happens
is that Dafny “forgets” about the body of every method except the one it is currently
working on. This simplifies Dafny’s job tremendously, and is one of the reasons it is able
to operate at reasonable speeds. It also helps us reason about our programs by breaking
them apart and so we can analyze each method in isolation (given the annotations for the
other methods). We don’t care at all what happens inside each method when we call it,
as long as it satisfies its annotations. This works because Dafny will prove that all the
methods satisfy their annotations, and refuse to compile our code until they do.
For the Abs method, this means that the only thing Dafny knows in the Testing
method about the value returned from Abs is what the postconditions say about it, and
nothing more. This means that Dafny won’t know the nice property about Abs and non-
negative integers unless we tell it by putting this in the postcondition of the Abs method.
Another way to look at it is to consider the method annotations (along with the type of
the parameters and return values) as fixing the behavior of the method. Everywhere the
method is used, we assume that it is any conceivable method that satisfies the pre- and
postconditions. In the Abs case, we might have written:
This method satisfies the postconditions, but clearly the program fragment:
var v := Abs(3);
assert v == 3;
would not be true in this case. Dafny is considering, in an abstract way, all methods with
those annotations. The mathematical absolute value certainly is such a method, but so are
all methods that return a positive constant, for example. We need stronger postconditions
to eliminate these other possibilities, and “fix” the method down to exactly the one we
want. We can partially do this with the following:
This expresses exactly the property we discussed before, that the absolute value is the
same for non-negative integers. The second ensures uses implication operator, ==>,
which basically says that the left-hand side implies the right in the mathematical sense (it
binds more weakly than boolean and and comparisons, so the above says 0 <= x implies
x == y). The left and right sides must both be boolean expressions.
The postcondition says that after Abs is called, if the value of x was non-negative,
then y is equal to x. One caveat of the implication is that it is still true if the left part
(the antecedent) is false. So the second postcondition is trivially true when x is negative.
In fact, the only thing that the annotations say when x is negative is that the result, y,
is non-negative. But this is still not enough to fix the function, so we must add another
postcondition, to make the following complete annotation covering all cases:
These annotations are enough to require that our method actually computes the absolute
value of x. But we still have an issue: there seems to be a lot of duplication. The body
of the method is reflected very closely in the annotations. While this is correct code, we
want to eliminate this redundancy. As you might guess, Dafny provides a means of doing
this: functions.
Exercise 2 Using a precondition, change Abs to say it can only be called on negative
values. Simplify the body of Abs into just one assignment statement and make sure the
method still verifies.
J. Koenig and K.R.M. Leino / A Guide to Dafny 9
Exercise 3 Keeping the postconditions of Abs the same as above, change the body of Abs
to y := x + 2;. What precondition do you need to annotate the method with in order
for the verification to go through? And what precondition do you need if the body is
y := x + 1;? What does that precondition say about calling the method?
4. Functions
This declares a function called abs which takes a single integer, and returns an integer
(the second int). Unlike a method, which can have all sorts of statements in its body,
a function body must consist of exactly one expression, with the correct type. Here our
body must be an integer expression. In order to implement the absolute value function,
we need to use an if expression. An if expression behaves like the ternary operator in
other languages.
Obviously, the condition must be a boolean expression, and the two branches must have
the same type. You might wonder why anyone would bother with functions, if they are
so limited compared to methods. The power of functions comes from the fact that they
can be used directly in annotations. So we can write:
assert abs(3) == 3;
In fact, not only can we write this statement directly without capturing to a local vari-
able, we didn’t even need to write all the postconditions that we did with the method
(though functions can and do have pre- and postconditions in general). The limitations
of functions are precisely what let Dafny do this. Unlike methods, Dafny does not forget
the body of a function when considering other functions. So it can expand the definition
of abs in the above assertion and determine that the result is actually 3.
Exercise 4 Write a function max that returns the larger of two given integer pa-
rameters. Write a test method that uses an assert statement to check some prop-
erty of the value of max on some arguments (for example, 34 < max(21,55) or
max(21,55) == max(55,21)).
10 J. Koenig and K.R.M. Leino / A Guide to Dafny
One caveat of functions is that not only can they appear in annotations, they can only
appear in annotations. One cannot write:
var v := abs(3);
as this is not an annotation. Functions are never part of the final compiled program,
they are just tools to help us verify our code. Nevertheless, sometimes it is convenient
to use a function in real code too, so one can define a function method, which can be
called from real code. Note that there are restrictions on what functions can be function
methods.
Exercise 5 Change the test method in Exercise 4 to put the result of max into a local
variable and then use an assert statement to check some property of that local vari-
able. Dafny will reject this program, because you’re calling max from real code. Fix this
problem using a function method.
Exercise 6 Now that we have an abs function, change the postcondition of method Abs
(at the end of Section 3) to make use of abs. After you make sure the method still verifies,
change the body of Abs to also use abs. (After doing this, you will also realize there’s not
much point in having a method that does exactly the same thing as a function method.)
Here we use nat, the type of natural numbers (non-negative integers), which is often
more convenient than annotating everything to be non-negative. It turns out that we could
make this function a function method if we wanted to. But executing that function method
would be extremely slow, as this version of calculating the Fibonacci numbers has ex-
ponential complexity. There are much better ways to calculate the Fibonacci function.
But this function is still useful, as we can have Dafny prove that our fast version really
matches the mathematical definition. We can get the best of both worlds: the guarantee
of correctness and the performance we want.
We can start by defining a method like the following:
We haven’t written the body yet, so Dafny will complain that our post condition doesn’t
hold. We need an algorithm to calculate the n th Fibonacci number. The basic idea is to
J. Koenig and K.R.M. Leino / A Guide to Dafny 11
keep a counter, and repeatedly calculate adjacent pairs of Fibonacci numbers until the
desired number is reached. To do this, we need a loop. In Dafny, this is done via a while
loop:
var i := 0;
while (i < n)
{
i := i + 1;
}
This is a trivial loop that just increments i until it reaches n. This will form the core of
our loop to calculate Fibonacci numbers.
5. Loop Invariants
Loops present a problem for Dafny. There is no way for Dafny to know in advance how
many times the code will go around the loop. But Dafny needs to consider all paths
through a program, which could include going around the loop any number of times.
To make it possible for Dafny to work with loops, you need to provide loop invariants,
another kind of annotation.
A loop invariant is an expression that holds upon entering a loop, and after every
execution of the loop body. It captures something that is invariant, i.e., does not change,
about every step of the loop. Now, obviously we are going to want to change variables,
etc. each time around the loop, or we wouldn’t need the loop. Like pre- and postcondi-
tions, an invariant is a property that is preserved for each execution of the loop, expressed
using the same boolean expressions we have seen. For example, we see in the above loop
that if i starts off positive, then it stays positive. So we can add the invariant, using its
own keyword, to the loop:
var i := 0;
while (i < n)
invariant 0 <= i;
{
i := i + 1;
}
When you specify an invariant, Dafny proves two things: the invariant holds upon enter-
ing the loop, and it is preserved by the loop. By preserved, we mean that assuming that
the invariant holds at the beginning of the loop, we must show that executing the loop
body once makes the invariant hold again. Dafny can only know upon analyzing the loop
body what the invariants say, in addition to the loop guard (the loop condition). Just as
Dafny will not discover properties of a method on its own, it will not know any but the
most basic properties of a loop are preserved unless it is told via an invariant.
In our example, the point of the loop is to build up the Fibonacci numbers one (well,
two) at a time until we reach the desired number. After we exit the loop, we will have
that i == n, because i will stop being incremented when it reaches n. We can use our
assertion trick to check to see if Dafny sees this fact as well:
12 J. Koenig and K.R.M. Leino / A Guide to Dafny
var i := 0;
while (i < n)
invariant 0 <= i;
{
i := i + 1;
}
assert i == n;
We find that this assertion fails. As far as Dafny knows, it is possible that i somehow
became much larger than n at some point during the loop. All it knows after the loop
exits (i.e., in the code after the loop) is that the loop guard failed, and the invariants hold.
In this case, this amounts to n <= i and 0 <= i. But this is not enough to guarantee that
i == n, just that n <= i. Somehow we need to eliminate the possibility of i exceeding
n. One first guess for solving this problem might be the following:
var i := 0;
while (i < n)
invariant 0 <= i < n;
{
i := i + 1;
}
This does not verify, as Dafny complains that the invariant is not preserved (also known
as maintained) by the loop. We want to be able to say that after the loop exits, then all the
invariants hold. Our invariant holds for every execution of the loop except for the very
last one. Because the loop body is executed only when the loop guard holds, in the last
iteration i goes from n - 1 to n, but does not increase further, at the loop exits. Thus, we
have only omitted exactly one case from our invariant, and repairing it relatively easy:
...
invariant 0 <= i <= n;
...
Now we can say both that n <= i from the loop guard and 0 <= i <= n from the invari-
ant, which allows Dafny to prove the assertion i == n. The challenge is finding a loop
invariant that is preserved by the loop, but also that lets you prove what you need after
the loop has executed.
Exercise 7 Change the loop invariant to 0 <= i <= n+2. Does the loop still verify?
Does the assertion i == n after the loop still verify?
Exercise 8 Once again using the invariant 0 <= i <= n for the loop above, change the
loop guard from i < n to i != n. Do the loop and the assertion after the loop still
verify?
In addition to the counter, our algorithm called for a pair of numbers which represent
adjacent Fibonacci numbers in the sequence. Unsurprisingly, we will have another invari-
J. Koenig and K.R.M. Leino / A Guide to Dafny 13
ant or two to relate these numbers to each other and the counter. To find these invariants,
we employ a common Dafny trick: working backwards from the postconditions.
Our postcondition for the Fibonacci method is that the return value b is equal to
fib(n). But after the loop, we have that i == n, so we need b == fib(i) at the end of
the loop. This might make a good invariant, as it relates something to the loop counter.
This observation is surprisingly common throughout Dafny programs. Many times a
method is just a loop that, when it ends, makes the postcondition true by having a counter
reach another number, often an argument or the length of an array or sequence. So we
have that the variable b, which is conveniently our out parameter, will be the current
Fibonacci number:
invariant b == fib(i);
We also note that in our algorithm, we can compute any Fibonacci number by keeping
track of a pair of numbers, and summing them to get the next number. So we want a way
of tracking the previous Fibonacci number, which we will call a. Another invariant will
express that number’s relation to the loop counter. The invariants are:
At each step of the loop, the two values are summed to get the next leading number,
while the trailing number is the old leading number. Using a parallel assignment, we can
write a loop that performs this operation:
var i := 1;
var a := 0;
b := 1;
while (i < n)
invariant 0 < i <= n;
invariant a == fib(i - 1);
invariant b == fib(i);
{
a, b := b, a + b;
i := i + 1;
}
Here a is the trailing number, and b is the leading number. The parallel assignment means
that the entire right-hand side is calculated before the assignments to the variables are
made. Thus a will get the old value of b, and b will get the sum of the two old values,
which is precisely the behavior we want.
We also have made a change to the loop counter i. Because we also want to track the
trailing number, we can’t start the counter at 0, as otherwise we would have to calculate a
negative Fibonacci number. The problem with doing this is that the loop counter invariant
may not hold when we enter the loop. The only problem is when n is 0. This can be
eliminated as a special case, by testing for this condition at the beginning of the loop.
The completed Fibonacci method becomes:
14 J. Koenig and K.R.M. Leino / A Guide to Dafny
Dafny no longer complains about the loop invariant not holding, because if n were 0, it
would return before reaching the loop. Dafny is also able to use the loop invariants to
prove that after the loop, i == n and b == fib(i), which together imply the postcondi-
tion, b == fib(n).
Exercise 9 The ComputeFib method above is more complicated than necessary. Write
a simpler program by not introducing a as the Fibonacci number that precedes b, but
instead introducing a variable c that that succeeds b. Verify that your program is correct.
Exercise 10 Starting with the final ComputeFib method above, delete the if statement
and change the initializations to start i at 0, a at 1, and b at 0. Verify this new program
by adjusting the loop invariants to match the new behavior.
One of the problems with using invariants is that it is easy to forget to have the loop
make progress, i.e., do work at each step. For example, we could have omitted the entire
body of the loop in the previous program. The invariants would be correct, because they
are still true upon entering the loop, and since the loop doesn’t change anything, they
would be preserved by the loop. But the crucial step from the loop to the postcondition
wouldn’t hold. We know that if we exit the loop, then we can assume the negation of
the guard and the invariants, but this says nothing about what happens if we never exit
the loop. Thus we would like to make sure the loop ends at some point, which gives us a
much stronger correctness guarantee.
6. Termination
Dafny proves code terminates, i.e., does not loop forever, by using decreases annota-
tions. For many things, Dafny is able to guess the right annotations, but sometimes it
needs to be made explicit. In fact, for all of the code we have seen so far, Dafny has been
able to do this proof on its own, which is why we haven’t needed the decreases annota-
J. Koenig and K.R.M. Leino / A Guide to Dafny 15
tion explicitly yet. There are two places Dafny proves termination: loops and recursion.
Both of these situations require either an explicit annotation or a correct guess by Dafny.
A decreases annotation, as its name suggests, gives Dafny an expression that de-
creases with every loop iteration or recursive call. There are two conditions that Dafny
needs to verify when using a decreases expression: that the expression actually gets
smaller, and that it is bounded. Many times, an integral value (natural or plain integer)
is the quantity that decreases, but other things can be used as well. In the case of inte-
gers, the bound is assumed to be zero. For example, the following is a proper use of a
decreases annotation on a loop:
while (0 < i)
invariant 0 <= i;
decreases i;
{
i := i - 1;
}
Here Dafny has all the ingredients it needs to prove termination. The variable i gets
smaller each loop iteration, and is bounded below by zero. This is fine, except the loop is
backwards from most loops, which tend to count up instead of down. In this case, what
decreases is not the counter itself, but rather the distance between the counter and the
upper bound. A simple trick for dealing with this situation is given below:
while (i < n)
invariant 0 <= i <= n;
decreases n - i;
{
i := i + 1;
}
This is actually Dafny’s guess for this situation, as it sees i < n and assumes that
n - i is the quantity that decreases. The upper bound of the loop invariant implies that
0 <= n - i, and gives Dafny a lower bound on the quantity. This also works when the
bound n is not constant, such as in the binary search algorithm, where two quantities
approach each other.
Exercise 11 In the loop above, the invariant i <= n and the negation of the loop guard
allow us to conclude i == n after the loop (as we checked in Section 5 by writing an
assert statement). Note that if the loop guard were instead written as i != n (as in
Exercise 8), then the negation of the guard immediately gives i == n after the loop, re-
gardless of the loop invariant. Change the loop guard to i != n and delete the invariant
annotation. Does the program verify?
The other situation that requires a termination proof is when methods or functions
are recursive. Similarly to looping forever, these methods could potentially call them-
selves forever, never returning to their original caller. When Dafny is not able to guess
the termination condition, an explicit decreases clause can be given along with pre- and
postconditions, as in the unnecessary annotation for the fib function:
16 J. Koenig and K.R.M. Leino / A Guide to Dafny
As before, Dafny can guess this condition on its own, but sometimes the decreasing
condition is hidden within a field of an object where Dafny cannot find it on its own, and
then it requires an explicit annotation.
7. Arrays
All that we have considered is fine for toy functions and little mathematical exercises,
but it really isn’t helpful for real programs. So far we have only considered a handful of
values at a time in local variables. Now we turn our attention to arrays of data. Arrays are
a built in part of the language, with their own type, array<T>, where T is another type.
For now we only consider arrays of integers, array<int>. Arrays can be null, and have
a built in length field called Length. Element access uses the standard bracket syntax
and are indexed from zero, so a[3] is preceded by the 3 elements a[0], a[1], and a[2],
in that order. All array accesses must be proven to be within bounds, which is part of
the no-runtime-errors safety guaranteed by Dafny. Because bounds checks are proven at
verification time, no runtime checks need to be made. To create a new array, it must be
allocated with the new keyword, but for now we will only work with methods that take a
previously allocated array as an argument.
One of the most basic things we might want to do with an array is search through it
for a particular key, and return the index of a place where we can find the key if it exists.
We have two outcomes for a search, with a different correctness condition for each. If
the algorithm returns an index (i.e., non-negative integer), then the key should be present
at that index. This might be expressed as follows:
The array index here is safe because the implication operator is short circuiting. Short cir-
cuiting means if the left part is false, then the implication is already true regardless of the
truth value of the second part, and thus it does not need to be evaluated. Using the short
circuiting property of the implication operator, along with the boolean and (&&), which is
also short circuiting, is a common Dafny practice. The condition index < a.Length is
necessary because otherwise the method could return a large integer which is not an in-
dex into the array. Together, the short circuiting behavior means that by the time control
reaches the array access, index must be a valid index.
If the key is not in the array, then we would like the method to return a negative
number. In this case, we want to say that the method did not miss an occurrence of the
J. Koenig and K.R.M. Leino / A Guide to Dafny 17
key; in other words, that the key is not in the array. To express this property, we turn to
another common Dafny tool: quantifiers.
8. Quantifiers
A quantifier in Dafny most often takes the form of a forall expression, also called a
universal quantifier. As its name suggests, this expression is true if some property holds
for all elements of some set. For now, we will consider the set of integers. An example
universal quantifier, wrapped in an assertion, is given below:
A quantifier introduces a temporary name for each element of the set it is considering.
This is called the bound variable, in this case k. The identifier has a type, which is almost
always inferred rather than given explicitly and is usually int anyway. (In general, one
can have any number of bound variables; we will encounter an example in Section 9.) A
pair of colons (::) separates the bound variable and its optional type from the quantified
property (which must be of type bool). In this case, the property is that adding 1 to
any integer makes a strictly larger integer. Dafny is able to prove this simple property
automatically. Generally it is not very useful to quantify over infinite sets, such as all the
integers. Instead, quantifiers are typically used to quantify over all elements in an array
or data structure. We do this for arrays by using the implication operator to make the
quantified property trivially true for values which are not indices:
This says that some property holds for each element of the array. The implication makes
sure that k is actually a valid index into the array before evaluating the second part of the
expression. Dafny can use this fact not only to prove that the array is accessed safely, but
also to reduce the set of integers it must consider to only those that are indices into the
array.
With a quantifier, saying the key is not in the array is straightforward:
Thus our method becomes (with the addition of the non-nullity requirement on a):
We can fill in the body of this method in a number of ways, but perhaps the easiest is a
linear search, implemented below:
18 J. Koenig and K.R.M. Leino / A Guide to Dafny
index := 0;
while (index < a.Length)
{
if (a[index] == key) { return; }
index := index + 1;
}
index := -1;
As you can see, we have omitted the loop invariants on the while loop, so Dafny gives us
a verification error on one of the postconditions. The reason we get an error is that Dafny
does not know that the loop actually covers all the elements. In order to convince Dafny
of this, we have to write an invariant that says that everything before the current index
has already been looked at (and is not the key). Just like the postcondition, we can use a
quantifier to express this property:
This says that everything before, but excluding, the current index is not the key. Notice
that upon entering the loop, i is 0, so the first part of the implication is always false, and
thus the quantified property is always true. This common situation is known as vacuous
truth: the quantifier holds because it is quantifying over an empty set of objects. This
means that it is true when entering the loop. We test the value of the array before we
extend the non-key part of the array, so Dafny can prove that this invariant is preserved.
One problem arises when we try to add this invariant: Dafny complains about the index
being out of range for the array access within the invariant.
This is erroneous because there is no invariant on index, so it could be greater than
the length of the array. Then the bound variable, k, could exceed the length of the ar-
ray. To fix this, we put the standard bounds on index, 0 <= index <= a.Length. Note
that because we say k < index, the array access is still protected from error even when
index == a.Length. The use of a variable that is one past the end of a growing range is
a common pattern when working with arrays, where it is often used to build a property
up one element at a time. The complete method is given below:
Exercise 12 Write a method that takes an integer array, which it requires to have at least
one element, and returns an index to the maximum of the array’s elements. Annotate the
method with pre- and postconditions that state the intent of the method, and annotate its
body with loop invariants to verify it.
A linear search is not very efficient, especially when many queries are made of the
same data. If the array is sorted, then we can use the efficient binary search procedure to
find the key. But in order for us to be able to prove our implementation correct, we need
some way to require that the input array actually is sorted. We could do this directly with
a quantifier inside a requires clause of our method, but a more modular way to express
this is through a predicate.
9. Predicates
A predicate is a function that returns a boolean. It is a simple but powerful idea that
occurs throughout Dafny programs. For example, preparing the way for the binary search
method in Section 11, we define the sorted predicate over arrays of integers as a function
that takes an array as an argument, and returns true if that array is sorted in increasing
order. The use of predicates makes our code shorter, as we do not need to write out a long
property over and over. It can also make our code easier to read by giving a common
property a name.
There are a number of ways we could write the sorted predicate, but the easiest is
to use a quantifier over the indices of the array. We can write a quantifier that expresses
the property, “if x is before y in the array, then x <= y”, as a quantifier over two bound
variables:
Here we have two bound variables, j and k, which are both integers. The comparisons
between the two guarantee that they are both valid indices into the array, and that j is
before k. Then the second part says that they are ordered properly with respect to one
another. The function, along with the standard non-null precondition required for using
the array, is thus:
Dafny rejects this code as given, claiming that the function cannot read the elements of
a. Dealing with this complaint requires another annotation, the reads annotation.
Exercise 13 To get practice with quantifiers, modify the definition of function sorted so
that it returns true only when the array is sorted and all its elements are distinct.
20 J. Koenig and K.R.M. Leino / A Guide to Dafny
10. Framing
The sorted predicate is not able to access the array because the array was not included in
the function’s reading frame. The reading frame of a function is all the memory locations
that the function is allowed to read. The reason we might limit what a function can read
is so that when we write to memory, we can be sure that functions that do not read that
part of memory yield the same value after the write as they did before the write. For
example, we might have two arrays, one of which we know is sorted. If we did not put
a reads annotation on the sorted predicate, then when we modify the unsorted array, we
cannot immediately determine whether the other array stopped being sorted. While we
might be able to give invariants to preserve it in this case, it gets even more complex
when manipulating data structures and dealing with recursive functions. In those cases,
framing is essential to making the verification process feasible.
To specify a reading frame, we use a reads annotation, again with its own keyword:
A reads annotation can appear anywhere along with the pre- and postconditions, but it
does not give a boolean expression like the other annotations we have seen; instead, it
specifies a set of memory locations that the function is allowed to access. The name of
an array, like a in the above example, stands for all the elements of that array. One can
also specify object fields and sets of objects, but we will not concern ourselves with those
topics here. Dafny will check that you do not read any memory location that is not stated
in the reading frame. This means that function calls within a function must have reading
frames that are a subset of the calling function’s reading frame. One thing to note is that
parameters to the function are not memory locations, so do not need to be declared.
Frames also affect methods. As you might have guessed, methods are not required
to list the things they read, as we have written a method which accesses an array with
no reads annotation. Methods are allowed to read whatever memory they like, but they
are required to list which parts of memory they modify, with a modifies annotation.
They are almost identical to their reads cousins, except they say what can be changed.
In combination with reads, modification restrictions allow Dafny to prove properties of
code that would otherwise be very difficult or impossible. Reads and modifies are two
of the tools that allow Dafny to work on one method at a time, because they restrict
what would otherwise be arbitrary modifications of memory to something that Dafny can
reason about.
ensures index < 0 ==> forall k :: 0 <= k < a.Length ==> a[k] != key;
{
...
}
Exercise 14 What happens if you remove the precondition a != null? Change the def-
inition of sorted so that it allows its argument to be null but returns false if it is.
We have the same postconditions that we did for the linear search, as the goal is the
same. The difference is that now we know the array is sorted. Because Dafny can unwrap
functions, inside the body of the method it knows this too. We can then use that property
to prove the correctness of the search. The method body is given below:
First we declare our range to search over. This can be thought of as the remaining space
where the key could possibly be. The first invariant expresses the fact that this range is
within the array. The second says that the key is not anywhere outside of this range. In
the first two branches of the if chain, we find the element in the middle of our range is
not the key, and so we move the range to exclude that element and all the other elements
on the appropriate side of it. We need the addition of 1 when moving the lower end of the
range because it is inclusive. If we do not add this, then the loop may continue forever
when mid == low, which happens when low + 1 == high. We could change this to say
that the loop exits when low and high are one apart, but this would mean we would need
an extra check after the loop to determine if the key was found at the one remaining
index. In the above formulation, this is unnecessary because when low == high, the loop
exits. But this means that no elements are left in the search range, so the key was not
present. This can be deduced from the loop invariant:
invariant forall i :: 0 <= i < a.Length && !(low <= i < high)
==> a[i] != key;
22 J. Koenig and K.R.M. Leino / A Guide to Dafny
When low == high, the negated condition in the first part of the implication is always
true (because no i can be both at least and strictly smaller than the same value). Thus the
invariant says that all elements in the array are not the key, and the second postcondition
holds. As you can see, it is easy to introduce subtle off-by-one errors in this code. With
the invariants, not only can Dafny prove the code correct, but we can understand the
operation of the code easier ourselves.
Exercise 15 Change the assignments in the body of BinarySearch to set low to mid or
to set high to mid - 1. In each case, what goes wrong?
For advanced uses of specification features in Dafny or other tools, we recommend study-
ing textbooks on program correctness (such as [6,8,7,11]), collections of verified pro-
grams (such as [23,4]), or, as have become more prevalent lately, the solutions submitted
to verification challenges ([22,16,17]) and verification competitions ([13]).
There are several techniques for specifying programs with dynamic data structures
(for a survey, see [10]). For this purpose, Dafny supports a specification idiom based on
dynamic frames [12,21]. That idiom goes beyond what we have covered in these lecture
notes, but see [15,14].
13. Conclusion
We’ve seen a whirlwind tour of the major features of Dafny, and used it for some inter-
esting, if a little on the small side, examples of what Dafny can do. Even if you do not
use Dafny regularly, writing down what it is that your code does, in a precise way, and
using this to prove code correct is a useful skill to have. Invariants, pre- and postcondi-
tions, and annotations are useful in debugging code, and also as documentation for future
developers. When modifying or adding to a codebase, they confirm that the guarantees
of existing code are not broken. If annotations are checked at run time, they act as test
oracles injected in the code. They also ensure that APIs are used correctly, by formaliz-
ing behavior and requirements and enforcing correct usage. Reasoning from invariants,
considering pre- and postconditions, and writing assertions to check assumptions are all
general computer science skills that will benefit you no matter what language you work
in.
References
[0] Mike Barnett, Bor-Yuh Evan Chang, Robert DeLine, Bart Jacobs, and K. Rustan M. Leino. Boogie: A
modular reusable verifier for object-oriented programs. In Frank S. de Boer, Marcello M. Bonsangue,
Susanne Graf, and Willem-Paul de Roever, editors, Formal Methods for Components and Objects: 4th
International Symposium, FMCO 2005, volume 4111 of Lecture Notes in Computer Science, pages
364–387. Springer, September 2006.
[1] Mike Barnett, Manuel Fähndrich, K. Rustan M. Leino, Peter Müller, Wolfram Schulte, and Herman
Venter. Specification and verification: the Spec# experience. Communications of the ACM, 54(6):81–91,
June 2011.
J. Koenig and K.R.M. Leino / A Guide to Dafny 23
[2] Raymond T. Boute. The Euclidian definition of the functions div and mod. ACM Transactions on
Programming Languages and Systems, 14(2):127–144, April 1992.
[3] Lilian Burdy, Yoonsik Cheon, David R. Cok, Michael D. Ernst, Joeseph R. Kiniry, Gary T. Leavens,
K. Rustan M. Leino, and Erik Poll. An overview of JML tools and applications. International Journal
on Software Tools for Technology Transfer, 7(3):212–232, June 2005.
[4] Dafny test suite. Files with extension .dfy in the subdirectories of Test at boogie.codeplex.com.
[5] Leonardo de Moura and Nikolaj Bjørner. Z3: An efficient SMT solver. In C. R. Ramakrishnan and Jakob
Rehof, editors, Tools and Algorithms for the Construction and Analysis of Systems, 14th International
Conference, TACAS 2008, volume 4963 of Lecture Notes in Computer Science, pages 337–340. Springer,
March–April 2008.
[6] Edsger W. Dijkstra. A Discipline of Programming. Prentice Hall, Englewood Cliffs, NJ, 1976.
[7] Edsger W. Dijkstra and W. H. J. Feijen. A Method of Programming. Addison-Wesley, July 1988.
[8] David Gries. The Science of Programming. Texts and Monographs in Computer Science. Springer-
Verlag, 1981.
[9] John V. Guttag and James J. Horning, editors. Larch: Languages and Tools for Formal Specification.
Texts and Monographs in Computer Science. Springer, 1993. With Stephen J. Garland, Kevin D. Jones,
Andrés Modet, and Jeannette M. Wing.
[10] John Hatcliff, Gary T. Leavens, K. Rustan M. Leino, Peter Müller, and Matthew Parkinson. Behavioral
interface specification languages. Technical Report CS-TR-09-01a, Department of Electrical Engineer-
ing and Computer Science, University of Central Florida, October 2010. To appear in ACM Computing
Surveys.
[11] Anne Kaldewaij. Programming: The Derivation of Algorithms. Series in Computer Science. Prentice-
Hall International, 1990.
[12] Ioannis T. Kassios. Dynamic frames: Support for framing, dependencies and sharing without restric-
tions. In Jayadev Misra, Tobias Nipkow, and Emil Sekerinski, editors, FM 2006: Formal Methods,
14th International Symposium on Formal Methods, volume 4085 of Lecture Notes in Computer Science,
pages 268–283. Springer, August 2006.
[13] Vladimir Klebanov, Peter Müller, Natarajan Shankar, Gary T. Leavens, Valentin Wüstholz, Eyad Alka-
ssar, Rob Arthan, Derek Bronish, Rod Chapman, Ernie Cohen, Mark Hillebrand, Bart Jacobs, K. Rus-
tan M. Leino, Rosemary Monahan, Frank Piessens, Nadia Polikarpova, Tom Ridge, Jan Smans, Stephan
Tobies, Thomas Tuerk, Mattias Ulbrich, and Benjamin Weiß. The 1st verified software competition: Ex-
perience report. In Michael Butler and Wolfram Schulte, editors, FM 2011: Formal Methods - 17th In-
ternational Symposium on Formal Methods, volume 6664 of Lecture Notes in Computer Science, pages
154–168. Springer, June 2011.
[14] K. Rustan M. Leino. Specification and verification of object-oriented software. In Manfred Broy, Was-
siou Sitou, and Tony Hoare, editors, Engineering Methods and Tools for Software Safety and Security,
volume 22 of NATO Science for Peace and Security Series D: Information and Communication Security,
pages 231–266. IOS Press, 2009. Summer School Marktoberdorf 2008 lecture notes.
[15] K. Rustan M. Leino. Dafny: An automatic program verifier for functional correctness. In Edmund M.
Clarke and Andrei Voronkov, editors, LPAR-16, volume 6355 of Lecture Notes in Computer Science,
pages 348–370. Springer, April 2010.
[16] K. Rustan M. Leino and Rosemary Monahan. Dafny meets the verification benchmarks challenge. In
Gary T. Leavens, Peter W. O’Hearn, and Sriram K. Rajamani, editors, Verified Software: Theories, Tools,
Experiments, Third International Conference, VSTTE 2010, volume 6217 of Lecture Notes in Computer
Science, pages 112–126. Springer, August 2010.
[17] K. Rustan M. Leino and Michał Moskal. VACID-0: Verification of ample correctness of invariants of
data-structures, edition 0. In Tools and Experiments workshop at VSTTE 2010, August 2010.
[18] K. Rustan M. Leino and Philipp Rümmer. A polymorphic intermediate verification language: Design
and logical encoding. In Javier Esparza and Rupak Majumdar, editors, Tools and Algorithms for the
Construction and Analysis of Systems, 16th International Conference, TACAS 2010, volume 6015 of
Lecture Notes in Computer Science, pages 312–327. Springer, March 2010.
[19] Barbara Liskov and John Guttag. Abstraction and Specification in Program Development. MIT Electri-
cal Engineering and Computer Science Series. MIT Press, 1986.
[20] Bertrand Meyer. Object-oriented Software Construction. Series in Computer Science. Prentice-Hall
International, 1988.
[21] Jan Smans, Bart Jacobs, Frank Piessens, and Wolfram Schulte. Automatic verifier for Java-like programs
24 J. Koenig and K.R.M. Leino / A Guide to Dafny
based on dynamic frames. In José Luiz Fiadeiro and Paola Inverardi, editors, Fundamental Approaches
to Software Engineering, 11th International Conference, FASE 2008, volume 4961 of Lecture Notes in
Computer Science, pages 261–275. Springer, March–April 2008.
[22] Bruce W. Weide, Murali Sitaraman, Heather K. Harton, Bruce Adcock, Paolo Bucci, Derek Bronish,
Wayne D. Heym, Jason Kirschenbaum, and David Frazier. Incremental benchmarks for software verifi-
cation tools and techniques. In Natarajan Shankar and Jim Woodcock, editors, Verified Software: Theo-
ries, Tools, Experiments, Second International Conference, VSTTE 2008, volume 5295 of Lecture Notes
in Computer Science, pages 84–98. Springer, October 2008.
[23] Why3 gallery of certified programs. proval.lri.fr/gallery/why3.en.html.
This appendix illustrates many of the most common language features in Dafny.
A.0. Programs
At the top level, a Dafny program (stored as a file with extension .dfy) is a set of dec-
larations. The declarations introduce fields, methods, and functions, as well as classes
and inductive datatypes, where the order of introduction is irrelevant. A class also con-
tains a set of declarations, introducing fields, methods, and functions. Fields, methods,
and functions declared outside a class go into an implicit class called _default, giving
the appearance of the program having global variables, procedures, and functions. If the
program contains a unique parameter-less method called Main, then program execution
starts there, but it is not necessary to have a main method to do verification.
Comments start with // and go to the end of the line, or start with /* and end with
* / and can be nested.
A.1. Fields
var x: T;
Unlike for local variables and bound variables, the type is required and will not be in-
ferred. The field can be declared to be a ghost field by preceding the declaration with
the keyword ghost. Ghost entities (fields, variables, functions, methods, code) are not
represented at run time; they are only used by the verifier, which treats them as it treats
non-ghost entities.
Dafny’s types include bool for booleans, int for mathematical (that is, unbounded)
integers, user-defined classes and inductive datatypes, set<T> for finite mathemati-
cal (immutable) sets of T values (where T is any type), and seq<T> for mathematical
(immutable) sequences of T values. In addition, there are array types (which are like
predefined “class” types) of one and more dimensions, written array<T>, array2<T>,
array3<T>, . . . . The type object is a supertype of all class types, that is, an object
denotes any reference, including null. Finally, the type nat denotes a subrange of int,
namely the non-negative integers.
J. Koenig and K.R.M. Leino / A Guide to Dafny 25
A.2. Methods
A.3. Functions
where a, b, c are the method’s parameters, T is the type of the function’s result, Pre is a
boolean expression denoting the function’s precondition, Frame denotes a set of objects
whose fields the function body may depend on, Post is a boolean expression denoting
the function’s postcondition, Rank is the function’s variant function, and Body is an ex-
pression that defines the function. The precondition allows a function to be partial, that
is, the precondition says when the function is defined (and Dafny will verify that every
use of the function meets the precondition). The postcondition is usually not needed,
since the body of the function gives the full definition. However, the postcondition can
be a convenient place to declare properties of the function that may require an inductive
proof to establish. For example:
says that the result of Factorial is always positive, which Dafny verifies inductively
from the function body. To refer to the function’s result in the postcondition, use the
function itself, as shown in the example.
By default, a function is ghost, and cannot be called from non-ghost code. To make
it non-ghost, replace the keyword function with the two keywords function method.
By default, a function has an implicit receiver parameter, this. This parameter can
be removed by preceding the function declaration with the keyword static. A static
function F in a class C can be invoked by C.F( . . . ). This can give a convenient way to
declare a number of helper functions in a separate class.
A.4. Classes
class C {
// member declarations go here
}
where the members of the class (fields, methods, and functions) are defined (as described
above) inside the curly braces.
A.5. Datatypes
An inductive datatype is a type whose values are created using a fixed set of constructors.
A datatype Tree with constructors Empty and Node is declared as follows:
The constructors are separated by vertical bars. Parameter-less constructors need not use
parentheses, as is shown here for Empty.
For each constructor Ct, the datatype implicitly declares a boolean member Ct?,
which returns true for those values that have been constructed using Ct. For example,
after the code snippet:
var t0 := Empty;
var t1 := Node(t0, 5, t0);
the expression t1.Node? evaluates to true and t0.Node? evaluates to false. Two
datatype values are equal if they have been created using the same constructor and equal
parameters to that constructor. Therefore, for parameter-less constructors like Empty,
t.Empty? gives the same result as t == Empty.
A constructor can optionally declare a destructor for any of its parameters, which is
done by introducing a name for the parameter. For example, if Tree were declared as:
then t1.data == 5 and t1.left == t0 hold after the code snippet above.
A.6. Generics
Dafny supports generic types. That is, any class, inductive datatype, method, and function
can have type parameters. These are declared in angle brackets after the name of what is
being declared. For example:
class Multiset<T> { /* . . . */ }
datatype Tree<T> = Empty | Node(Tree<T>, T, Tree<T>);
method Find<T>(key: T, collection: Tree<T>) { /* . . . */ }
function IfThenElse<T>(b: bool, x: T, y: T): T { /* . . . */ }
A.7. Statements
invariant Inv;
modifies Frame;
decreases Rank;
{
Stmts
}
match (Expr) {
case Empty => Stmts0
case Node(l, d, r) => Stmts1
}
break;
return;
The var statement introduces local variables (which are not allowed to shadow other
variables declared inside the same set of most tightly enclosing curly braces). Each vari-
able can optionally be followed by :T for any type T, which explicitly gives the preced-
ing variable the type T (rather than being inferred). The ExprList with initial values is
optional. To declare the variables as ghost variables, precede the declaration with the
keyword ghost.
The assignment statement assigns each right-hand side in ExprList to the corre-
sponding left-hand side in Lvalues. These assignments are performed in parallel, so the
left-hand sides must denote distinct L-values. Each right-hand side can be an expression
or an object creation of one of the following forms:
new T
new T.Init(ExprList)
new T[SizeExpr]
new T[SizeExpr0, SizeExpr1]
The first form allocates an object of type T. The second form additionally invokes an
initialization method or constructor Init on the newly allocated object. The other forms
show examples of array allocations, in particular a one- and a two-dimensional array of
T values, respectively.
The entire right-hand side of an assignment can also be a method call, in which
case the left-hand sides are the actual out-parameters (omitting the “:=” if there are no
out-parameters).
The assert statement claims that the given expression evaluates to true (which is
checked by the verifier).
The print statement outputs to standard output the values of the given print expres-
sions. A print expression is either an expression or a string literal (where \n is used to
denote a newline character).
The if statement is the usual one. The example shows stringing together alternatives
using else if. The else branch is optional, as usual.
The while statement is the usual loop, where the invariant declaration gives a
loop invariant, the modifies clause restricts the modification frame of the loop, and
the decreases clause introduces a variant function for the loop. By default, the loop
invariant is true, the modification frame is the same as in the enclosing context (usually
J. Koenig and K.R.M. Leino / A Guide to Dafny 29
the modifies clause of the enclosing method), and the variant function is guessed from
the loop guard.
The match statement evaluates the source Expr, an expression whose type is an
inductive datatype, and then executes the case corresponding to which constructor was
used to create the source datatype value, binding the constructor parameters to the given
names.
The break statement can be used to exit loops, and the return statement can be used
to exit a method.
A.8. Expressions
The expressions in Dafny are quite similar to those in Java-like languages. Here are some
noteworthy differences.
In addition to the short-circuiting boolean operators && (and) and || (or), Dafny
has a short-circuiting implication operator ==> and an if-and-only-if operator <==>. As
suggested by their widths, <==> has lower binding power than ==>, which in turn has
lower binding power than && and ||. Implication associates to the right.
Dafny comparison expressions can be chaining, which means that comparisons “in
the same direction” can be strung together. For example,
Note that boolean equality can be expressed using both == and <==>. There are
two differences between these. First, == has a stronger binding power than <==>. Sec-
ond, == is chaining while <==> is associative. That is, a == b == c is the same as
a == b && b == c, whereas a <==> b <==> c is the same as a <==> (b <==> c),
which is also the same as (a <==> b) <==> c.
Operations on integers are the usual ones, except that / (integer division) and %
(integer modulo) follow the Euclidean definition, which means that % always results in a
non-negative number [2]. (Hence, when the first argument to / or % is negative, the result
is different than what you get in C, Java, or C#.)
Dafny expressions include universal and existential quantifiers, which have the form:
and likewise for exists, where each bound variable can optionally be followed by an
explicit type, as in x: T, and Expr is a boolean expression.
Operations on sets include + (union), * (intersection), and - (set difference), as well
as the set comparison operators < (proper subset), <= (subset), their duals > and >=, and
!! (disjointness). The expression x in S says that x is a member of set S, and x !in S
is a convenient way of writing !(x in S). To make a set from some elements, enclose
them in curly braces. For example, {x,y} is the set consisting of x and y (which is a
singleton set if x == y), {x} is the singleton set containing x, and {} is the empty set.
30 J. Koenig and K.R.M. Leino / A Guide to Dafny
where Expr0 and Expr1 are any expressions of the same type. Unlike the if statement,
the if-then-else expression does not require parentheses around the guard expression,
uses the then keyword, and must include an explicit else branch.
The match statement also has an analogous match expression, which has a form like:
match Expr
case Empty => Expr0
case Node(l, d, r) => Expr1
As with the if statement versus the if-then-else expression, note that the match expres-
sion does not require parentheses around the source expression and does not surround
the cases with curly braces. A match expression can only be used in the body of func-
tion definitions, where it must either be the entire body or be the entire expression for
a case in an enclosing match expression; furthermore, the source expression must be a
parameter of the enclosing function.