# Practical Persimmon.Dried #FsAdvent

This article is the English translation of the 3rd day post from F# Advent Calendar Japan 2015. The article of the 2nd day is “使いたくない型を見えなくする #FsAdvent - 2つのアンコール (EN: Shadowing unfavorable types)” from @htid46.

I will introduce “Persimmon.Dried”, another random testing library in F#, and explain the practical way to use it.

## What is Persimmon?

Persimmon is one of a testing framework written in F#; it named because the project began in Autumn and the name reminds many Japanese people the season. Persimmon has some related products to make it more useful.

Persimmon has some features which enables you to:

- Write tests with computation expressions
- Execute each tests to the end even if the test fails in any assertion
- Write parameterized tests
- etc

```
open Persimmon
let test1 = test "Sample test 1" { // a test by the name of "Sample test 1"
let x = 5
do! assertPred (x * 2 = x + 2) // fails but continue the test
do! assertEquals (x * 3) (x + x + x) // and assert this line
}
```

You can also make naming tests more simplified with opening UseTestNameByReflection module.

```
open Persimmon
open UseTestNameByReflection
let ``Sample test 2`` = test { // a test by the name of "Sample test 2"
do! assertEquals 33 4
}
```

Because of its usefulness, we use Persimmon with opening this module.

If you want to know more of the product, please read the official documents.

## What is Persimmon.Dried?

Persimmon.Dried is a random testing library in the Persimmon series.

This tool is a kind of implementations of QuickCheck. When we wrote properties, we can test them with random input. You can also write tests with computation expressions like Persimmon.

```
open Persimmon
open Persimmon.Dried
open UseTestNameByReflection
let ``Property sample`` = Prop.forAll(Arb.int, Arb.int)(fun i j ->
i + j = j + i // write an expression to return bool
)
let ``Check property`` = property {
apply ``Property sample`` // check your property
}
// write properties and assertions together
let ``another way to write property`` =
let ``another way to write property`` = Prop.forAll(Arb.list Arb.string)(fun xs ->
List.length xs >= 0
)
property {
apply ``another way to write property``
}
```

Although Persimmon.Dried has the official documents too, I will show you various topics (mainly out of the document).

## How to write

When we write properties in Persimmon.Dried, it usually forms like following:

```
Prop-Function (Tuple-Of-Arbitrary) (Predicate-Function-To-Express-Property)
```

Arbitrary is candidates of arguments. For example, you write Arb.int, Arb.int then the tool passes two random integer parameters to a predicate function. The arguments of the predicate function is inferred by the tuple of Arbitrary. In the next snipet:

```
let ``yourself go`` = Prop.forAll(Arb.list Arb.int)(fun ls ->
List.length ls >= 0
)
```

the tuple of Arbitrary is Arb.list Arb.int and it stands for the candidates of argument which are int lists. The predicate function takes a parameter ls, which is passed as int list generated randomly.

## Arbitrary

You can use arbitraries flexibly: there are various arbitraries for basic types like Arb.int, Arb.byte and Arb.string, and for collection types like Arb.list Arb.int, Arb.array Arb.float and Arb.map Arb.int Arb.string.

You can also use arbitraries for functions or System.Func:

```
// generate F# function (int -> int)
let ``function as parameter`` = Prop.forAll(Arb.func CoArb.int Arb.int)(fun f ->
f (f 0) = (f >> f) 0
)
// generate System.Func<int, string>
let ``System.Func as parameter`` = Prop.forAll(Arb.systemFunc(CoArb.int, Arb.string))(fun f ->
f.Invoke(0) = "0"
)
```

A module CoArb represents the module of “arbitraries for functions (which are arbitraries)”.

If you want to treat arbitraries of collection types as non-empty or non-null, you can:

```
// non empty int list
let ``non empty list`` = Prop.forAll(Arb.nonEmpty(Arb.list Arb.int), Arb.int)(fun xs i ->
List.head xs = i
)
// non null string []
let ``non null array`` = Prop.forAll(Arb.nonNull(Arb.array Arb.string))(fun xs ->
Array.length xs >= 0
)
```

## Arbitrary and Gen

When you use Persimmon.Dried, you may want to generate arguments as well as you think. You can make another arbitrary from the present one.

```
// generate int list arguments whose length are no more than 3
let arbLs = {
Gen = Gen.listOfMaxLength 3 <| Arb.int.Gen
Shrinker = Shrink.shrinkList <| Arb.int.Shrinker
PrettyPrinter = Pretty.prettyList
}
let ``list of max length 3`` = Prop.forAll(arbLs)(fun xs ->
List.length xs <= 3
)
```

The basic arbitraries are F# records which have 3 values (labels), Gen, Shrinker and PrettyPrinter. In these labels, Gen : Gen<'T> takes charge of generating values, so controling it makes the result better.

Gen module has a lot of useful functions fo Gen<T> values. There are examples in the following subsections.

### oneOf

oneOf function takes a sequence of Gen<’T> then returns one of elements in it.

```
// use one of Gen element in the sequence
let arbOneOf = {
Arb.int with Gen = Gen.oneOf <| List.map Gen.constant [2; 3; 5; 7]
}
let ``get one of prime numbers`` = Prop.forAll(arbOneOf)(fun i ->
let contains v = List.exists (fun x -> x = v)
contains i [1 .. 10] // [2; 3; 5; 7] ⊂ [1 .. 10]
)
```

### suchThat

suchThat function takes Gen<’T> value and a condition function, then returns Gen<’T> to generate values which satisfy the condition.

```
// generate values satisfying the condition
let arbEven = {
Arb.int with Gen = Gen.suchThat (fun i -> i % 2 = 0) Arb.int.Gen // filter only even number
}
let ``only even number`` = Prop.forAll(arbEven)(fun i ->
(i * i) % 4 = 0
)
```

### listOfLength

listOfLength function takes a length number N and Gen<’T>, then returns Gen<’T list> which generates N-length ‘T list.

```
// generate int lists whose length is 3
let arbList3 = {
Gen = Gen.listOfLength 3 <| Arb.int.Gen
Shrinker = Shrink.shrinkList <| Arb.int.Shrinker
PrettyPrinter = Pretty.prettyList
}
// generate int lists whose length is 5
let arbList5 = {
arbList3 with Gen = Gen.listOfLength 5 <| Arb.int.Gen
}
let ``list of length`` = Prop.forAll(arbList3, arbList5)(fun xs3 xs5 ->
List.length xs3 < List.length xs5
)
```

There are similar functions such as listOfMaxLength which generates lists less than or equal to length N, and listOfMinLength which generates lists greater than or equal to length N.

If you want to know more about Gen, let’s read the source code.

## Properties

In Prop module, there are useful functions and operators to write properties.

### And / Or operator

The And operator .&. succeeds if both of properties succeed, otherwise fails.

```
// lhs .&. lazy rhs succeeds both of lhs and rhs
let ``and operator`` = Prop.forAll(Arb.int, Arb.int)(fun x y ->
x + 1 > x |@ "x + 1 is greater than x" .&.
lazy (y - 1 < y |@ "y - 1 is less than y") .&.
lazy (x * y > x + y |@ "x * y is greather than x + y") // This property may fail
)
```

On the other hand, the Or operator .|. succeeds if either property succeeds, fails if both fail.

```
// lhs .|. lazy rhs succeeds either of lhs and rhs
let ``or operator`` = Prop.forAll(Arb.int, Arb.int)(fun x y ->
"x + 1 is greater than x" @| (x + 1 > x) .|.
lazy ("y - 1 is less than y" @| (y - 1 < y)) .|.
lazy ("x * y is greather than x + y" @| (x * y > x + y)) // This property may fail
)
```

If you execute each 2 properties, and operator will fail and or operator will succeed.

|@ and @| you see in the code snipets above are the labeling operators. You can use them as p |@ s or s @| p, then label p s.

### Conditional property

==> operator checks a right hand side property only if a left hand side property succeeds.

```
// lhs ==> lazy rhs checks rhs if lhs succeeds.
let ``conditional property`` = Prop.forAll(Arb.int, Arb.list Arb.int)(fun i ls ->
let ils = List.Cons(i, ls)
List.length ls > 5 ==> lazy (List.length ils > 6)
)
```

This snipet checks whether the length of i :: ls is greater than 6 or not if the input list ls has greater length than 5.

Note that the binary operators which take both properties, same as .&. or .|., must take its right hand side property as lazy. This restriction is for the reason that F# mainly uses eager evaluation. In the case of rasing an exception when the right property is evaluated, the exception is occurred at the time of passing both properties to the operator, then the check fails.

```
// What if we could...
let ``right hand side executes... what?`` = Prop.forAll(Arb.int)(fun i ->
(i + 1 = i) ==> (i / 0 = 0) // write like this?
)
```

### Classify properties

classify function is similar to the labeling operators in the way of naming properties. Moreover, it can classify them by any conditions.

```
let ``classifying test case`` = Prop.forAll(Arb.int, Arb.int)(fun x y ->
x + y > x
|> Prop.classify(x < 0, "x is negative")
|> Prop.classify((x = 0), "x is zero")
|> Prop.classify(y < 0, "y is negative")
|> Prop.classify((y = 0), "y is zero")
)
```

When you execute the snipet above, it may show you as follows:

```
Assertion Violated: classifying test case
1. Falsified after 1 passed tests.
> ARG_0: 1513071627
> ARG_1: 0
> ARG_1_ORIGINAL: -1890354640
> current random state: (abbrev.)
> Collected test data:
50% x is negative
50% y is zero
```

You can get the ratio of failing arguments from here. However, it’s difficult to use this function because it won’t show the ratio without failing the classified properties.

## Conclusion

We saw some practical ways to use Persimmon.Dried briefly.

This is the random testing library available on the Persimmon testing framework; it’s as easy as the preceding product FsCheck, and improving now.

Try it out and enjoy!