package tablecloth-native

  1. Overview
  2. Docs

A module for working with floating-point numbers. Valid syntax for floats includes:

0.
0.0
42.
42.0
3.14
0.1234
123_456.123_456
6.022e23   (* = (6.022 * 10^23) *)
6.022e+23  (* = (6.022 * 10^23) *)
1.602e−19  (* = (1.602 * 10^-19) *)
1e3        (* = (1 * 10 ** 3) = 1000. *)

Without opening this module you can use the . suffixed operators e.g

1. +. 2. /. 0.25 *. 2. = 17. 

But by opening this module locally you can use the un-suffixed operators

Float.((10.0 - 1.5 / 0.5) ** 3.0) = 2401.0

Historical Note: The particular details of floats (e.g. NaN) are specified by IEEE 754 which is literally hard-coded into almost all CPUs in the world.

type t = float

Constants

val zero : t

The literal 0.0 as a named value

val one : t

The literal 1.0 as a named value

val nan : t

NaN as a named value. NaN stands for not a number.

Note comparing values with Float.nan will always return false even if the value you are comparing against is also NaN.

e.g

let isNotANmber x = Float.(x = nan) in
isNotANumber nan = false

For detecting Nan you should use Float.isNaN

val infinity : t

Positive infinity

Float.log ~base:10.0 0.0 = Float.infinity
val negativeInfinity : t

Negative infinity, see Float.infinity

val negative_infinity : t
val e : t

An approximation of Euler's number.

val pi : t

An approximation of pi.

Basic arithmetic and operators

val add : t -> t -> t

Addition for floating point numbers.

Float.add 3.14 3.14 = 6.28
Float.(3.14 + 3.14 = 6.28)

Although ints and floats support many of the same basic operations such as addition and subtraction you cannot add an int and a float directly which means you need to use functions like Int.toFloat or Float.roundToInt to convert both values to the same type.

So if you needed to add a List.length to a float for some reason, you could:

Float.add 3.14 (Int.toFloat (List.length [1,2,3])) = 6.14

or

Float.roundToInt 3.14 + List.length [1,2,3] = 6

Languages like Java and JavaScript automatically convert int values to float values when you mix and match. This can make it difficult to be sure exactly what type of number you are dealing with and cause unexpected behavior.

OCaml has opted for a design that makes all conversions explicit.

val (+) : t -> t -> t
val subtract : t -> t -> t

Subtract numbers

Float.subtract 4.0 3.0 = 1.0

Alternatively the - operator can be used:

Float.(4.0 - 3.0) = 1.0
val (-) : t -> t -> t
val multiply : t -> t -> t

Multiply numbers like

Float.multiply 2.0 7.0 = 14.0

Alternatively the operator * can be used:

Float.(2.0 * 7.0) = 14.0
val (*) : t -> t -> t
val divide : t -> by:t -> t

Floating-point division:

Float.divide 3.14 ~by:2.0 = 1.57

Alternatively the / operator can be used:

Float.(3.14 / 2.0) = 1.57
val (/) : t -> t -> t
val power : base:t -> exponent:t -> t

Exponentiation, takes the base first, then the exponent.

Float.power ~base:7.0 ~exponent:3.0 = 343.0

Alternatively the ** operator can be used:

Float.(7.0 ** 3.0) = 343.0
val (**) : t -> t -> t
val negate : t -> t

Flips the 'sign' of a float so that positive floats become negative and negative integers become positive. Zero stays as it is.

Float.negate 8 = (-8)
Float.negate (-7) = 7
Float.negate 0 = 0

Alternatively an operator is available:

Float.(~- 4.0) = (-4.0)
val (~-) : t -> t
val absolute : t -> t

Get the absolute value of a number.

Float.absolute 8. = 8.
Float.absolute (-7) = 7
Float.absolute 0 = 0
val maximum : t -> t -> t

Returns the larger of two floats, if both arguments are equal, returns the first argument

Float.maximum 7. 9. = 9.
Float.maximum (-4.) (-1.) = (-1.)

If either (or both) of the arguments are NaN, returns NaN

Float.(isNaN (maximum 7. nan) = true
val minimum : t -> t -> t

Returns the smaller of two floats, if both arguments are equal, returns the first argument

Float.minimum 7.0 9.0 = 7.0
Float.minimum (-4.0) (-1.0) = (-4.0)

If either (or both) of the arguments are NaN, returns NaN

Float.(isNaN (minimum 7. nan) = true
val clamp : t -> lower:t -> upper:t -> t

Clamps n within the inclusive lower and upper bounds.

Float.clamp ~lower:0. ~upper:8. 5. = 5.
Float.clamp ~lower:0. ~upper:8. 9. = 8.
Float.clamp ~lower:(-10.) ~upper:(-5.) 5. = -5.

Throws an Invalid_argument exception if lower > upper

Fancier math

val squareRoot : t -> t

Take the square root of a number.

Float.squareRoot 4.0 = 2.0
Float.squareRoot 9.0 = 3.0

squareRoot returns NaN when its argument is negative. See Float.nan for more.

val square_root : t -> t
val log : t -> base:t -> t

Calculate the logarithm of a number with a given base.

Float.log ~base:10. 100. = 2.
Float.log ~base:2. 256. = 8.

Checks

val isNaN : t -> bool

Determine whether a float is an undefined or unrepresentable number.

Float.isNaN (0.0 / 0.0) = true
Float.(isNaN (squareRoot (-1.0)) = true
Float.isNaN (1.0 / 0.0) = false  (* Float.infinity {b is} a number *)
Float.isNaN 1. = false

Note this function is more useful than it might seem since NaN does not equal Nan:

Float.(nan = nan) = false
val is_nan : t -> bool
val isFinite : t -> bool

Determine whether a float is finite number. True for any float except Infinity, -Infinity or NaN

Float.isFinite (0. / 0.) = false
Float.(isFinite (squareRoot (-1.)) = false
Float.isFinite (1. / 0.) = false
Float.isFinite 1. = true
Float.(isFinite nan) = false

Notice that NaN is not finite!

For a float n to be finite implies that Float.(not (isInfinite n || isNaN n)) evaluates to true.

val is_finite : t -> bool
val isInfinite : t -> bool

Determine whether a float is positive or negative infinity.

Float.isInfinite (0. / 0.) = false
Float.(isInfinite (squareRoot (-1.)) = false
Float.isInfinite (1. / 0.) = true
Float.isInfinite 1. = false
Float.(isInfinite nan) = false

Notice that NaN is not infinite!

For a float n to be finite implies that Float.(not (isInfinite n || isNaN n)) evaluates to true.

val is_infinite : t -> bool
val inRange : t -> lower:t -> upper:t -> bool

Checks if n is between lower and up to, but not including, upper. If lower is not specified, it's set to to 0.0.

Float.inRange ~lower:2. ~upper:4. 3. = true
Float.inRange ~lower:1. ~upper:2. 2. = false
Float.inRange ~lower:5.2 ~upper:7.9 9.6 = false

Throws an Invalid_argument exception if lower > upper

val in_range : t -> lower:t -> upper:t -> bool

Angles

val hypotenuse : t -> t -> t

hypotenuse x y returns the length of the hypotenuse of a right-angled triangle with sides of length x and y, or, equivalently, the distance of the point (x, y) to (0, 0).

Float.hypotenuse 3. 4. = 5.
val degrees : t -> t

Converts an angle in degrees to Float.radians.

Float.degrees 180. = v
val radians : t -> t

Convert a Float.t to radians.

Float.(radians pi) = 3.141592653589793

Note This function doesn't actually do anything to its argument, but can be useful to indicate intent when inter-mixing angles of different units within the same function.

val turns : t -> t

Convert an angle in turns into Float.radians.

One turn is equal to 360°.

Float.(turns (1. / 2.)) = pi
Float.(turns 1. = degrees 360.)

Polar coordinates

val fromPolar : (float * float) -> float * float

Convert polar coordinates (r, θ) to Cartesian coordinates (x,y).

Float.(fromPolar (squareRoot 2., degrees 45.)) = (1., 1.)
val from_polar : (float * float) -> float * float
val toPolar : (float * float) -> float * float

Convert Cartesian coordinates (x,y) to polar coordinates (r, θ).

Float.toPolar (3.0, 4.0) = (5.0, 0.9272952180016122)
Float.toPolar (5.0, 12.0) = (13.0, 1.1760052070951352)
val to_polar : (float * float) -> float * float
val cos : t -> t

Figure out the cosine given an angle in radians.

Float.(cos (degrees 60.)) = 0.5000000000000001
Float.(cos (radians (pi / 3.))) = 0.5000000000000001
val acos : t -> t

Figure out the arccosine for adjacent / hypotenuse in radians:

Float.(acos (radians 1.0 / 2.0)) = Float.radians 1.0471975511965979 (* 60° or pi/3 radians *)
val sin : t -> t

Figure out the sine given an angle in radians.

Float.(sin (degrees 30.)) = 0.49999999999999994
Float.(sin (radians (pi / 6.)) = 0.49999999999999994
val asin : t -> t

Figure out the arcsine for opposite / hypotenuse in radians:

Float.(asin (1.0 / 2.0)) = 0.5235987755982989 (* 30° or pi / 6 radians *)
val tan : t -> t

Figure out the tangent given an angle in radians.

Float.(tan (degrees 45.)) = 0.9999999999999999
Float.(tan (radians (pi / 4.)) = 0.9999999999999999
Float.(tan (pi / 4.)) = 0.9999999999999999
val atan : t -> t

This helps you find the angle (in radians) to an (x, y) coordinate, but in a way that is rarely useful in programming.

You probably want atan2 instead!

This version takes y / x as its argument, so there is no way to know whether the negative signs comes from the y or x value. So as we go counter-clockwise around the origin from point (1, 1) to (1, -1) to (-1,-1) to (-1,1) we do not get angles that go in the full circle:

Float.atan (1. /. 1.) = 0.7853981633974483  (* 45° or pi/4 radians *)
Float.atan (1. /. -1.) = -0.7853981633974483  (* 315° or 7 * pi / 4 radians *)
Float.atan (-1. /. -1.) = 0.7853981633974483 (* 45° or pi/4 radians *)
Float.atan (-1. /.  1.) = -0.7853981633974483 (* 315° or 7 * pi/4 radians *)

Notice that everything is between pi / 2 and -pi/2. That is pretty useless for figuring out angles in any sort of visualization, so again, check out Float.atan2 instead!

val atan2 : y:t -> x:t -> t

This helps you find the angle (in radians) to an (x, y) coordinate. So rather than saying Float.(atan (y / x)) you can Float.atan2 ~y ~x and you can get a full range of angles:

Float.atan2 ~y:1. ~x:1. = 0.7853981633974483  (* 45° or pi/4 radians *)
Float.atan2 ~y:1. ~x:(-1.) = 2.3561944901923449  (* 135° or 3 * pi/4 radians *)
Float.atan2 ~y:(-1.) ~x:(-1.) = -(2.3561944901923449) (* 225° or 5 * pi/4 radians *)
Float.atan2 ~y:(-1.) ~x:1.) = -(0.7853981633974483) (* 315° or 7 * pi/4 radians *)

Conversion

type direction = [
  1. | `Zero
  2. | `AwayFromZero
  3. | `Up
  4. | `Down
  5. | `Closest of [ `Zero | `AwayFromZero | `Up | `Down | `ToEven ]
]
val round : ?direction:direction -> t -> t

Round a number, by default to the to the closest int with halves rounded `Up (towards positive infinity)

Float.round 1.2 = 1.0
Float.round 1.5 = 2.0
Float.round 1.8 = 2.0
Float.round -1.2 = -1.0
Float.round -1.5 = -1.0
Float.round -1.8 = -2.0

Other rounding strategies are available by using the optional ~direction label.

Towards zero

Float.round ~direction:`Zero 1.2 = 1.0
Float.round ~direction:`Zero 1.5 = 1.0
Float.round ~direction:`Zero 1.8 = 1.0
Float.round ~direction:`Zero (-1.2) = -1.0
Float.round ~direction:`Zero (-1.5) = -1.0
Float.round ~direction:`Zero (-1.8) = -1.0

Away from zero

Float.round ~direction:`AwayFromZero 1.2 = 1.0
Float.round ~direction:`AwayFromZero 1.5 = 1.0
Float.round ~direction:`AwayFromZero 1.8 = 1.0
Float.round ~direction:`AwayFromZero (-1.2) = -1.0
Float.round ~direction:`AwayFromZero (-1.5) = -1.0
Float.round ~direction:`AwayFromZero (-1.8) = -1.0

Towards infinity

This is also known as Float.ceiling

Float.round ~direction:`Up 1.2 = 1.0
Float.round ~direction:`Up 1.5 = 1.0
Float.round ~direction:`Up 1.8 = 1.0
Float.round ~direction:`Up (-1.2) = -1.0
Float.round ~direction:`Up (-1.5) = -1.0
Float.round ~direction:`Up (-1.8) = -1.0

Towards negative infinity

This is also known as Float.floor

List.map  ~f:(Float.round ~direction:`Down) [-1.8; -1.5; -1.2; 1.2; 1.5; 1.8] = [-2.0; -2.0; -2.0; 1.0 1.0 1.0]

To the closest integer

Rounding a number x to the closest integer requires some tie-breaking for when the fraction part of x is exactly 0.5.

Halves rounded towards zero

List.map  ~f:(Float.round ~direction:(`Closest `AwayFromZero)) [-1.8; -1.5; -1.2; 1.2; 1.5; 1.8] = [-2.0; -1.0; -1.0; 1.0 1.0 2.0]

Halves rounded away from zero

This method is often known as commercial rounding

List.map  ~f:(Float.round ~direction:(`Closest `AwayFromZero)) [-1.8; -1.5; -1.2; 1.2; 1.5; 1.8] = [-2.0; -2.0; -1.0; 1.0 2.0 2.0]

Halves rounded down

List.map  ~f:(Float.round ~direction:(`Closest `Down)) [-1.8; -1.5; -1.2; 1.2; 1.5; 1.8] = [-2.0; -2.0; -1.0; 1.0 1.0 2.0]

Halves rounded up

This is the default.

Float.round 1.5 is the same as Float.round ~direction:(`Closest `Up) 1.5

Halves rounded towards the closest even number

This tie-breaking rule is the default rounding mode using in

Float.round ~direction:(`Closest `ToEven) -1.5 = -2.0
Float.round ~direction:(`Closest `ToEven) -2.5 = -2.0
val floor : t -> t

Floor function, equivalent to Float.round ~direction:`Down.

Float.floor 1.2 = 1.0
Float.floor 1.5 = 1.0
Float.floor 1.8 = 1.0
Float.floor -1.2 = -2.0
Float.floor -1.5 = -2.0
Float.floor -1.8 = -2.0
val ceiling : t -> t

Ceiling function, equivalent to Float.round ~direction:`Up.

Float.ceiling 1.2 = 2.0
Float.ceiling 1.5 = 2.0
Float.ceiling 1.8 = 2.0
Float.ceiling -1.2 = (-1.0)
Float.ceiling -1.5 = (-1.0)
Float.ceiling -1.8 = (-1.0)
val truncate : t -> t

Ceiling function, equivalent to Float.round ~direction:`Zero.

Float.truncate 1.0 = 1
Float.truncate 1.2 = 1
Float.truncate 1.5 = 1
Float.truncate 1.8 = 1
Float.truncate (-1.2) = -1
Float.truncate (-1.5) = -1
Float.truncate (-1.8) = -1
val fromInt : int -> float

Convert an int to a float

Float.fromInt 5 = 5.0
Float.fromInt 0 = 0.0
Float.fromInt -7 = -7.0
val from_int : int -> float
val toInt : t -> int option

Converts a float to an Int by ignoring the decimal portion. See Float.truncate for examples.

Returns None when trying to round a float which can't be represented as an int such as Float.nan or Float.infinity or numbers which are too large or small.

Float.(toInt nan) = None
Float.(toInt infinity) = None

You probably want to use some form of Float.round prior to using this function.

Float.(round 1.6 |> toInt) = Some 2
val to_int : t -> int option
OCaml

Innovation. Community. Security.