Permalink
Find file
494 lines (381 sloc) 24.9 KB

Julia v0.6.0 Release Notes

New language features

  • New type system capabilities (#8974, #18457)

    • Type parameter constraints can refer to previous parameters, e.g. type Foo{R<:Real, A<:AbstractArray{R}}. Can also be used in method definitions.

    • New syntax Array{T} where T<:Integer, indicating a union of types over all specified values of T (represented by a UnionAll type). This provides behavior similar to parametric methods or typealias, but can be used anywhere a type is accepted. This syntax can also be used in method definitions, e.g. function inv(M::Matrix{T}) where T<:AbstractFloat. Anonymous functions can have type parameters via the syntax ((x::Array{T}) where T<:Real) -> 2x.

    • Implicit type parameters, e.g. Vector{<:Real} is equivalent to Vector{T} where T<:Real, and similarly for Vector{>:Int} (#20414).

    • Much more accurate subtype and type intersection algorithms. Method sorting and identification of equivalent and ambiguous methods are improved as a result.

Language changes

  • "Inner constructor" syntax for parametric types is deprecated. For example, in this definition:

    type Foo{T,S<:Real}
        x
        Foo(x) = new(x)
    end
    

    the syntax Foo(x) = new(x) actually defined a constructor for Foo{T,S}, i.e. the case where the type parameters are specified. For clarity, this definition now must be written as Foo{T,S}(x) where {T,S<:Real} = new(x) (#11310).

  • The keywords used to define types have changed (#19157, #20418).

    • immutable changes to struct

    • type changes to mutable struct

    • abstract changes to abstract type ... end

    • bitstype 32 Char changes to primitive type Char 32 end

    In 0.6, immutable and type are still allowed as synonyms without a deprecation warning.

  • Multi-line and single-line nonstandard command literals have been added. A nonstandard command literal is like a nonstandard string literal, but the syntax uses backquotes (`) instead of double quotes, and the resulting macro called is suffixed with _cmd. For instance, the syntax q`xyz` is equivalent to @q_cmd "xyz" (#18644).

  • Nonstandard string and command literals can now be qualified with their module. For instance, Base.r"x" is now parsed as Base.@r_str "x". Previously, this syntax parsed as an implicit multiplication (#18690).

  • For every binary operator , a .⨳ b is now automatically equivalent to the broadcast call (⨳).(a, b). Hence, one no longer defines methods for .* etcetera. This also means that "dot operations" automatically fuse into a single loop, along with other dot calls f.(x) (#17623). Similarly for unary operators (#20249).

  • Newly defined methods are no longer callable from the same dynamic runtime scope they were defined in (#17057).

  • isa is now parsed as an infix operator with the same precedence as in (#19677).

  • @. is now parsed as @__dot__, and can be used to add dots to every function call, operator, and assignment in an expression (#20321).

  • The identifier _ can be assigned, but accessing its value is deprecated, allowing this syntax to be used in the future for discarding values (#9343, #18251).

  • The typealias keyword is deprecated, and should be replaced with Vector{T} = Array{T,1} or a const assignment (#20500).

  • Experimental feature: x^n for integer literals n (e.g. x^3 or x^-3) is now lowered to Base.literal_pow(^, x, Val{n}), to enable compile-time specialization for literal integer exponents (#20530, #20889).

Breaking changes

This section lists changes that do not have deprecation warnings.

  • readline, readlines and eachline return lines without line endings by default. You must use readline(s, chomp=false), etc. to get the old behavior where returned lines include trailing end-of-line character(s) (#19944).

  • Strings no longer have a .data field (as part of a significant performance improvement). Use Vector{UInt8}(str) to access a string as a byte array. However, allocating the Vector object has overhead. You can also use codeunit(str, i) to access the ith byte of a String. Use sizeof(str) instead of length(str.data), and pointer(str) instead of pointer(str.data) (#19449).

  • Operations between Float16 and Integers now return Float16 instead of Float32 (#17261).

  • Keyword arguments are processed left-to-right: if the same keyword is specified more than once, the rightmost occurrence takes precedence (#17785).

  • The lgamma(z) function now uses a different (more standard) branch cut for real(z) < 0, which differs from log(gamma(z)) by multiples of 2π in the imaginary part (#18330).

  • broadcast now handles tuples, and treats any argument that is not a tuple or an array as a "scalar" (#16986).

  • broadcast now produces a BitArray instead of Array{Bool} for functions yielding a boolean result. If you want Array{Bool}, use broadcast! or .= (#17623).

  • Operations like .+ and .* on Range objects are now generic broadcast calls (see above) and produce an Array. If you want a Range result, use + and *, etcetera (#17623).

  • broadcast now treats Ref (except for Ptr) arguments as 0-dimensional arrays (#18965).

  • broadcast now handles missing data (Nullables) allowing operations to be lifted over mixtures of Nullables and scalars, as if the Nullable were like an array with zero or one element (#16961, #19787).

  • The runtime now enforces when new method definitions can take effect (#17057). The flip-side of this is that new method definitions should now reliably actually take effect, and be called when evaluating new code (#265).

  • The array-scalar methods of /, \, *, +, and - now follow broadcast promotion rules. (Likewise for the now-deprecated array-scalar methods of div, mod, rem, &, |, and xor; see "Deprecated or removed" below.) (#19692).

  • broadcast!(f, A) now calls f() for each element of A, rather than doing fill!(A, f()) (#19722).

  • rmprocs now throws an exception if requested workers have not been completely removed before waitfor seconds. With a waitfor=0, rmprocs returns immediately without waiting for worker exits.

  • quadgk has been moved from Base into a separate package (#19741).

  • The Collections module has been removed, and all functions defined therein have been moved to the DataStructures package (#19800).

  • The RepString type has been moved to the LegacyStrings.jl package.

  • In macro calls with parentheses, e.g. @m(a=1), assignments are now parsed as = expressions, instead of as kw expressions (#7669).

  • When used as an infix operator, ~ is now parsed as a call to an ordinary operator with assignment precedence, instead of as a macro call (#20406).

  • (µ "micro" and ɛ "latin epsilon") are considered equivalent to the corresponding Greek characters in identifiers. \varepsilon now tab-completes to U+03B5 (greek small letter epsilon) (#19464).

  • retry now inputs the keyword arguments delays and check instead of n and max_delay. The previous functionality can be achieved setting delays to ExponentialBackOff (#19331).

  • transpose(::AbstractVector) now always returns a RowVector view of the input (which is a special 1×n-sized AbstractMatrix), not a Matrix, etc. In particular, for v::AbstractVector we now have (v.').' === v and v.' * v is a scalar (#19670).

  • Parametric types with "unspecified" parameters, such as Array, are now represented as UnionAll types instead of DataTypes (#18457).

  • Union types have two fields, a and b, instead of a single types field. The empty type Union{} is represented by a singleton of type TypeofBottom (#18457).

  • The type NTuple{N} now refers to tuples where every element has the same type (since it is shorthand for NTuple{N,T} where T). To get the old behavior of matching any tuple, use NTuple{N,Any} (#18457).

  • FloatRange has been replaced by StepRangeLen, and the internal representation of LinSpace has changed. Aside from changes in the internal field names, this leads to several differences in behavior (#18777):

    • Both StepRangeLen and LinSpace can represent ranges of arbitrary object types---they are no longer limited to floating-point numbers.

    • For ranges that produce Float64, Float32, or Float16 numbers, StepRangeLen can be used to produce values with little or no roundoff error due to internal arithmetic that is typically twice the precision of the output result.

    • To take advantage of this precision, linspace(start, stop, len) now returns a range of type StepRangeLen rather than LinSpace when start and stop are FloatNN. LinSpace(start, stop, len) always returns a LinSpace.

    • StepRangeLen(a, step, len) constructs an ordinary-precision range using the values and types of a and step as given, whereas range(a, step, len) will attempt to match inputs a::FloatNN and step::FloatNN to rationals and construct a StepRangeLen that internally uses twice-precision arithmetic. These two outcomes exhibit differences in both precision and speed.

  • A=>B expressions are now parsed as calls instead of using => as the expression head (#20327).

  • The count function no longer sums non-boolean values (#20404)

  • Bessel, Hankel, Airy, error, Dawson, eta, zeta, digamma, inverse digamma, trigamma, and polygamma special functions have been moved from Base to the SpecialFunctions.jl package (#20427). Note that airy, airyx and airyprime have been deprecated in favor of more specific functions (airyai, airybi, airyaiprime, airybiprimex, airyaix, airybix, airyaiprimex, airybiprimex) (#18050).

  • write on an IOBuffer now returns a signed integer in order to be consistent with other buffers (#20609).

  • @async, @spawn, @spawnat, @fetch and @fetchfrom no longer implicitly localize variables. Previously, the expression would be wrapped in an implicit let block ([#19594]).

  • Closures shipped for remote execution via @spawn or remotecall now automatically serialize globals defined under Main. For details, please refer to the paragraph on "Global variables" under the "Parallel computing" chapter in the manual ([#19594]).

Library improvements

  • A new @views macro was added to convert a whole expression or block of code to use views for all slices (#20164).

  • max, min, and related functions (minmax, maximum, minimum, extrema) now return NaN for NaN arguments (#12563).

  • oneunit(x) function to return a dimensionful version of one(x) (which is clarified to mean a dimensionless quantity if x is dimensionful) (#20268).

  • The chop and chomp functions now return a SubString (#18339).

  • Numbered stackframes printed in stacktraces can now be opened in an editor by entering the corresponding number in the REPL and pressing ^Q (#19680).

  • The REPL now supports something called prompt pasting (#17599). This activates when pasting text that starts with julia>  into the REPL. In that case, only expressions starting with julia>  are parsed, the rest are removed. This makes it possible to paste a chunk of code that has been copied from a REPL session without having to scrub away prompts and outputs. This can be disabled or enabled at will with Base.REPL.enable_promptpaste(::Bool).

  • The function print_with_color can now take a color represented by an integer between 0 and 255 inclusive as its first argument (#18473). For a number-to-color mapping, please refer to this chart. It is also possible to use numbers as colors in environment variables that customizes colors in the REPL. For example, to get orange warning messages, simply set ENV["JULIA_WARN_COLOR"] = 208. Please note that not all terminals support 256 colors.

  • The function print_with_color no longer prints text in bold by default (#18628). Instead, the function now take a keyword argument bold::Bool which determines whether to print in bold or not. On some terminals, printing a color in non bold results in slightly darker colors being printed than when printing in bold. Therefore, light versions of the colors are now supported. For the available colors see the help entry on print_with_color.

  • The default text style for REPL input and answers has been changed from bold to normal (#11250). They can be changed back to bold by setting the environment variables JULIA_INPUT_COLOR and JULIA_ANSWER_COLOR to "bold". For example, one way of doing this is adding ENV["JULIA_INPUT_COLOR"] = :bold and ENV["JULIA_ANSWER_COLOR"] = :bold to the .juliarc.jl file. See the manual section on customizing colors for more information.

  • The default color for info messages has been changed from blue to cyan (#18442), and for warning messages from red to yellow (#18453). This can be changed back to the original colors by setting the environment variables JULIA_INFO_COLOR to "blue" and JULIA_WARN_COLOR to "red".

  • Iteration utilities that wrap iterators and return other iterators (enumerate, zip, rest, countfrom, take, drop, cycle, repeated, product, flatten, partition) have been moved to the module Base.Iterators (#18839).

  • BitArrays can now be constructed from arbitrary iterables, in particular from generator expressions, e.g. BitArray(isodd(x) for x = 1:100) (#19018).

  • hcat, vcat, and hvcat now work with UniformScaling objects, so you can now do e.g. [A I] and it will concatenate an appropriately sized identity matrix (#19305).

  • New accumulate and accumulate! functions were added, which generalize cumsum and cumprod. Also known as a scan operation (#18931).

  • reshape now allows specifying one dimension with a Colon() (:) for the new shape, in which case that dimension's length will be computed such that its product with all the other dimensions is equal to the length of the original array (#19919).

  • A new titlecase function was added, to capitalize the first character of each word within a string (#19469).

  • any and all now always short-circuit, and mapreduce never short-circuits (#19543). That is, not every member of the input iterable will be visited if a true (in the case of any) or false (in the case of all) value is found, and mapreduce will visit all members of the iterable.

  • Additional methods for ones and zeros functions were added to support the same signature as the similar function (#19635).

  • count now has a count(itr) method equivalent to count(identity, itr) (#20403).

  • Methods for map and filter with Nullable arguments have been implemented; the semantics are as if the Nullable were a container with zero or one elements (#16961).

  • New @test_warn and @test_nowarn macros were added in the Base.Test module to test for the presence or absence of warning messages (#19903).

  • logging can now be used to redirect info, warn, and error messages either universally or on a per-module/function basis (#16213).

  • A new iszero(x) function was added, to quickly check whether x is zero (or is all zeros, for an array) (#19950).

  • notify now returns a count of tasks woken up (#19841).

  • A new nonstandard string literal raw"..." was added, for creating strings with no interpolation or unescaping (#19900).

  • A new Dates.Time type was added that supports representing the time of day with up to nanosecond resolution (#12274).

  • A new @macroexpand macro was added as a convenient alternative to the macroexpand function (#18660).

  • A new ConjArray type was added, as a wrapper type for lazy complex conjugation of arrays. Currently, it is used by default for the new RowVector type only, and enforces that both transpose(vec) and ctranspose(vec) are views not copies (#20047).

Compiler/Runtime improvements

  • ccall is now implemented as a macro, removing the need for special code-generator support for Intrinsics.

  • ccall gained limited support for a llvmcall calling-convention. This can replace many uses of llvmcall with a simpler, shorter declaration.

  • All Intrinsics are now Builtin functions instead and have proper error checking and fall-back static compilation support.

Deprecated or removed

  • Linear indexing is now only supported when there is exactly one non-cartesian index provided. Allowing a trailing index at dimension d to linearly access the higher dimensions from array A (beyond size(A, d)) has been deprecated as a stricter constraint during bounds checking. Instead, reshape the array such that its dimensionality matches the number of indices (#20079).

  • isdefined(a::Array, i::Int) has been deprecated in favor of isassigned (#18346).

  • is has been deprecated in favor of === (which used to be an alias for is) (#17758).

  • num and den have been deprecated in favor of numerator and denominator respectively (#19233).

  • infix operator $ has been deprecated in favor of infix or function xor() (#18977).

  • Dates.recur has been deprecated in favor of filter (#19288)

  • cummin and cummax have been deprecated in favor of accumulate.

  • sumabs and sumabs2 have been deprecated in favor of sum(abs, x) and sum(abs2, x), respectively. maxabs and minabs have similarly been deprecated in favor of maximum(abs, x) and minimum(abs, x). Likewise for the in-place counterparts of these functions (#19598).

  • produce, consume and iteration over a Task object have been deprecated in favor of using Channels for inter-task communication (#19841).

  • @test_approx_eq x y has been deprecated in favor of @test isapprox(x,y) or @test x ≈ y (#4615).

  • Vectorized functions have been deprecated in favor of dot syntax (#17302, #17265, #18558, #19711, #19712, #19791, #19802, #19931, #20543, #20228).

  • The two-argument forms of map (map!(f, A)) and asyncmap! (asyncmap!(f, A)) have been deprecated in anticipation of future semantic changes (#19721).

  • isimag has been deprecated (#19949).

  • broadcast_zpreserving has been deprecated (#19533, #19720).

  • convert methods from Diagonal and Bidiagonal to subtypes of AbstractTriangular have been deprecated (#17723).