Polymorphism

Specification: Vector Bounds

First, let’s see how to specify array bounds safety by refining the types for the key functions exported by Data.Vector, i.e. how to

1. define the size of a Vector
2. compute the size of a Vector
3. restrict the indices to those that are valid for a given size.

-- | Define the size
measure vlen :: Vector a -> Int

-- | Compute the size
assume length :: x:Vector a -> {v:Int | v = vlen x}

-- | Lookup at an index
assume (!) :: x:Vector a -> {v:Nat | v < vlen x} -> a

Using this new specification is now a simple matter of telling LiquidHaskell to include this file:

$ liquid -i include/ target.hs

LiquidHaskell ships with specifications for Prelude, Data.List, and Data.Vector which it includes by default.

Measures are used to define properties of Haskell data values that are useful for specification and verification. Think of vlen as the actual size of a Vector regardless of how the size was computed.

Assumes are used to specify types describing the semantics of functions that we cannot verify e.g. because we don’t have the code for them. Here, we are assuming that the library function Data.Vector.length indeed computes the size of the input vector. Furthermore, we are stipulating that the lookup function (!) requires an index that is between 0 and the real size of the input vector x.

Dependent Refinements are used to describe relationships between the elements of a specification. For example, notice how the signature for length names the input with the binder x that then appears in the output type to constrain the output Int. Similarly, the signature for (!) names the input vector x so that the index can be constrained to be valid for x. Thus, dependency lets us write properties that connect multiple program values.

Aliases are extremely useful for defining abbreviations for commonly occurring types. Just as we enjoy abstractions when programming, we will find it handy to have abstractions in the specification mechanism. To this end, LiquidHaskell supports type aliases. For example, we can define Vectors of a given size N as:

and now use this to type twoLangs above as:

Similarly, we can define an alias for Int values between Lo and Hi:

after which we can specify (!) as:

(!) :: x:Vector a -> Btwn 0 (vlen x) -> a

Verification: Vector Lookup

Let’s try to write some functions to sanity check the specifications. First, find the starting element – or head of a Vector

When we check the above, we get an error:

     src/03-poly.lhs:127:23: Error: Liquid Type Mismatch
       Inferred type
         VV : Int | VV == ?a && VV == 0

       not a subtype of Required type
         VV : Int | VV >= 0 && VV < vlen vec

       In Context
         VV  : Int | VV == ?a && VV == 0
         vec : Vector a | 0 <= vlen vec
         ?a  : Int | ?a == (0  :  int)

LiquidHaskell is saying that 0 is not a valid index as it is not between 0 and vlen vec. Say what? Well, what if vec had no elements! A formal verifier doesn’t make off by one errors.

To Fix the problem we can do one of two things.

1. Require that the input vec be non-empty, or
2. Return an output if vec is non-empty, or

Here’s an implementation of the first approach, where we define and use an alias NEVector for non-empty Vectors

Inference: Our First Recursive Function

Ok, let’s write some code! Let’s start with a recursive function that adds up the values of the elements of an Int vector.

LiquidHaskell verifies vectorSum – or, to be precise, the safety of the vector accesses vec ! i. Note you need to run liquid with the option --no-termination or make sure your source file has {-@ LIQUID "--no-termination" @-}. The verification works out because LiquidHaskell is able to *automatically infer* ^[In your editor, click ongo` to see the inferred type.]

go :: Int -> {v:Int | 0 <= v && v <= sz} -> Int

which states that the second parameter i is between 0 and the length of vec (inclusive). LiquidHaskell uses this and the test that i < sz to establish that i is between 0 and (vlen vec) to prove safety.

Higher-Order Functions: Bottling Recursion in a loop

Let’s refactor the above low-level recursive function into a generic higher-order loop.

We can now use loop to implement vectorSum:

Inference is a convenient option. LiquidHaskell finds:

loop :: lo:Nat -> hi:{Nat|lo <= hi} -> a -> (Btwn lo hi -> a -> a) -> a

In English, the above type states that

• lo the loop lower bound is a non-negative integer
• hi the loop upper bound is a greater then or equal to lo,
• f the loop body is only called with integers between lo and hi.

It can be tedious to have to keep typing things like the above. If we wanted to make loop a public or exported function, we could use the inferred type to generate an explicit signature.

At the call loop 0 n 0 body the parameters lo and hi are instantiated with 0 and n respectively, which, by the way is where the inference engine deduces non-negativity. Thus LiquidHaskell concludes that body is only called with values of i that are between 0 and (vlen vec), which verifies the safety of the call vec ! i.

Refinements and Polymorphism

While the standard Vector is great for dense arrays, often we have to manipulate sparse vectors where most elements are just 0. We might represent such vectors as a list of index-value tuples:

Implicitly, all indices other than those in the list have the value 0 (or the equivalent value for the type a).

The Alias SparseN is just a shorthand for the (longer) type on the right, it does not define a new type. If you are familiar with the index-style length encoding e.g. as found in DML or Agda, then note that despite appearances, our Sparse definition is not indexed.

Let’s write a function to compute a sparse product

LiquidHaskell verifies the above by using the specification to conclude that for each tuple (i, v) in the list y, the value of i is within the bounds of the vector x, thereby proving x ! i safe.

The sharp reader will have undoubtedly noticed that the sparse product can be more cleanly expressed as a fold:

foldl' :: (a -> b -> a) -> a -> [b] -> a

We can simply fold over the sparse vector, accumulating the sum as we go along

LiquidHaskell digests this without difficulty. The main trick is in how the polymorphism of foldl' is instantiated.

1. GHC infers that at this site, the type variable b from the signature of foldl' is instantiated to the Haskell type (Int, a).

2. Correspondingly, LiquidHaskell infers that in fact b can be instantiated to the refined (Btwn 0 (vlen x), a).

Thus, the inference mechanism saves us a fair bit of typing and allows us to reuse existing polymorphic functions over containers and such without ceremony.

Recap

This chapter gave you an idea of how one can use refinements to verify size related properties, and more generally, to specify and verify properties of recursive and polymorphic functions. Next, let’s see how we can use LiquidHaskell to prevent the creation of illegal values by refining data type definitions.