type T, then the three argument method should throw an InexactError. For example, if we have an Interval type which represents a range of possible values similar to https://github.com/Ju- liaPhysics/Measurements following julia> struct Interval{T} min::T max::T end julia> Base.round(x::Interval, r::RoundingMode) = Interval(round(x.min, r), round(x.max, r)) julia> x = Interval(1.7, 2.2) Interval{Float64}(1.7, 2.2) 2) julia> round(x) Interval{Float64}(2.0, 2.0) julia> floor(x) Interval{Float64}(1.0, 2.0) julia> ceil(x) Interval{Float64}(2.0, 3.0) julia> trunc(x) Interval{Float64}(1.0, 2.0)CHAPTER 16. INTERFACES

type T, then the three argument method should throw an InexactError. For example, if we have an Interval type which represents a range of possible values similar to https://github.com/Ju- liaPhysics/Measurements following julia> struct Interval{T} min::T max::T end julia> Base.round(x::Interval, r::RoundingMode) = Interval(round(x.min, r), round(x.max, r)) julia> x = Interval(1.7, 2.2) Interval{Float64}(1.7, 2.2) 2) julia> round(x) Interval{Float64}(2.0, 2.0) julia> floor(x) Interval{Float64}(1.0, 2.0) julia> ceil(x) Interval{Float64}(2.0, 3.0) julia> trunc(x) Interval{Float64}(1.0, 2.0)CHAPTER 16. INTERFACES

type T, then the three argument method should throw an InexactError. For example, if we have an Interval type which represents a range of possible values similar to https://github.com/Ju- liaPhysics/Measurements following julia> struct Interval{T} min::T max::T end julia> Base.round(x::Interval, r::RoundingMode) = Interval(round(x.min, r), round(x.max, r)) julia> x = Interval(1.7, 2.2) Interval{Float64}(1.7, 2.2) 2) julia> round(x) Interval{Float64}(2.0, 2.0) julia> floor(x) Interval{Float64}(1.0, 2.0) julia> ceil(x) Interval{Float64}(2.0, 3.0) julia> trunc(x) Interval{Float64}(1.0, 2.0)CHAPTER 16. INTERFACES

type T, then the three argument method should throw an InexactError. For example, if we have an Interval type which represents a range of possible values similar to https://github.com/Ju- liaPhysics/Measurements following julia> struct Interval{T} min::T max::T end julia> Base.round(x::Interval, r::RoundingMode) = Interval(round(x.min, r), round(x.max, r)) julia> x = Interval(1.7, 2.2) Interval{Float64}(1.7, 2.2) 2) julia> round(x) Interval{Float64}(2.0, 2.0) julia> floor(x) Interval{Float64}(1.0, 2.0) julia> ceil(x) Interval{Float64}(2.0, 3.0) julia> trunc(x) Interval{Float64}(1.0, 2.0)CHAPTER 16. INTERFACES

type T, then the three argument method should throw an InexactError. For example, if we have an Interval type which represents a range of possible values similar to https://github.com/Ju- liaPhysics/Measurements following julia> struct Interval{T} min::T max::T end julia> Base.round(x::Interval, r::RoundingMode) = Interval(round(x.min, r), round(x.max, r)) julia> x = Interval(1.7, 2.2) Interval{Float64}(1.7, 2.2) 2) julia> round(x) Interval{Float64}(2.0, 2.0) julia> floor(x) Interval{Float64}(1.0, 2.0) julia> ceil(x) Interval{Float64}(2.0, 3.0) julia> trunc(x) Interval{Float64}(1.0, 2.0)CHAPTER 16. INTERFACES

type T, then the three argument method should throw an InexactError. For example, if we have an Interval type which represents a range of possible values similar to https://github.com/Ju- liaPhysics/Measurements following julia> struct Interval{T} min::T max::T end julia> Base.round(x::Interval, r::RoundingMode) = Interval(round(x.min, r), round(x.max, r)) julia> x = Interval(1.7, 2.2) Interval{Float64}(1.7, 2.2) 2) julia> round(x) Interval{Float64}(2.0, 2.0) julia> floor(x) Interval{Float64}(1.0, 2.0) julia> ceil(x) Interval{Float64}(2.0, 3.0) julia> trunc(x) Interval{Float64}(1.0, 2.0)CHAPTER 16. INTERFACES

type T, then the three argument method should throw an InexactError. For example, if we have an Interval type which represents a range of possible values similar to https://github.com/Ju- liaPhysics/Measurements following julia> struct Interval{T} min::T max::T end julia> Base.round(x::Interval, r::RoundingMode) = Interval(round(x.min, r), round(x.max, r)) julia> x = Interval(1.7, 2.2) Interval{Float64}(1.7, 2.2) 2) julia> round(x) Interval{Float64}(2.0, 2.0) julia> floor(x) Interval{Float64}(1.0, 2.0) julia> ceil(x) Interval{Float64}(2.0, 3.0) julia> trunc(x) Interval{Float64}(1.0, 2.0)CHAPTER 16. INTERFACES

type T, then the three argument method should throw an InexactError. For example, if we have an Interval type which represents a range of possible values similar to https://github.com/Ju- liaPhysics/Measurements following julia> struct Interval{T} min::T max::T end julia> Base.round(x::Interval, r::RoundingMode) = Interval(round(x.min, r), round(x.max, r)) julia> x = Interval(1.7, 2.2) Interval{Float64}(1.7, 2.2) 2) julia> round(x) Interval{Float64}(2.0, 2.0) julia> floor(x) Interval{Float64}(1.0, 2.0) julia> ceil(x) Interval{Float64}(2.0, 3.0) julia> trunc(x) Interval{Float64}(1.0, 2.0)CHAPTER 16. INTERFACES

1 and uniformly distributed values. For floating point types T, the values lie in the half-open interval [0, 1). randn(T, dims...) an Array with random, iid and standard normally distributed values maybe_collect() are printed. If the value of the environment variable begins with the character 'r', then the interval between garbage collection events is randomized. Note These environment variables only have an jitter=0.1) A Float64 iterator of length n whose elements exponentially increase at a rate in the interval factor * (1 ± jitter). The first element is first_delay and all elements are clamped to max_delay

1 and uniformly distributed values. For floating point types T, the values lie in the half-open interval [0, 1). randn(T, dims...) an Array with random, iid and standard normally distributed values maybe_collect() are printed. If the value of the environment variable begins with the character 'r', then the interval between garbage collection events is randomized. Note These environment variables only have an jitter=0.1) A Float64 iterator of length n whose elements exponentially increase at a rate in the interval factor * (1 ± jitter). The first element is first_delay and all elements are clamped to max_delay