Add support for engineering notation with Ryu (#31)

Co-authored-by: Tim Holy <tim.holy@gmail.com>

Project.toml

@@ -14,6 +14,7 @@ julia = "1"
 [extras]
 Dates = "ade2ca70-3891-5945-98fb-dc099432e06a"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
+Printf = "de0858da-6303-5e67-8744-51eddeeeb8d7"
 
 [targets]
-test = ["Dates", "Test"]
+test = ["Dates", "Test", "Printf"]

src/Showoff.jl

@@ -2,10 +2,10 @@ module Showoff
 
 using Dates
 
-if isdefined(Base, :Grisu)
-    include("grisu.jl")
-else
+if isdefined(Base, :Ryu)
     include("ryu.jl")
+else
+    include("grisu.jl")
 end
 
 export showoff

src/ryu.jl

@@ -4,12 +4,16 @@ using Base.Ryu
 
 function plain_precision_heuristic(xs::AbstractArray{<:AbstractFloat})
     ys = filter(isfinite, xs)
-    precision = 0
+    e10max = -(e10min = typemax(Int))
     for y in ys
-        b, e10 = Ryu.reduce_shortest(convert(Float32, y))
-        precision = max(precision, -e10)
+        if isapprox(y, 0, atol=1e-16)
+            continue
+        end
+        _, e10 = Ryu.reduce_shortest(y)
+        e10min = min(e10min, e10)
+        e10max = max(e10max, e10)
     end
-    return max(precision, 0)
+    return precision = min(-e10min, -e10max+16)
 end
 
 # Print a floating point number at fixed precision. Pretty much equivalent to
@@ -41,39 +45,72 @@ function format_fixed_scientific(x::AbstractFloat, precision::Integer,
     end
 
     if engineering
-        b, e10 = Ryu.reduce_shortest(convert(Float32, x))
-        d, r = divrem(e10, 3)
-        if d < 0 &&
-            d += sign(r)
-
+        # precision applies before convertic to engineering format
+        # that is, number of digits would be the same as the with non-engineering format
+        base_digits, power = get_engineering_string(x, precision)
+    else
+        e_format_number = Ryu.writeexp(x, precision)
+        base_digits, power = split(e_format_number, 'e')
     end
-    ryustr = Ryu.writeexp(x, precision)
-    @show x ryustr
 
-    # Rewrite the exponent
+
     buf = IOBuffer()
-    ret = iterate(ryustr)
-    while ret !== nothing
-        c, state = ret
-        c === 'e' && break
-        print(buf, c)
-        ret = iterate(ryustr, state)
+
+    print(buf, base_digits)
+    print(buf, "×10")
+
+    if power[1] == '-'
+        print(buf, '⁻')
     end
-    if ret !== nothing
-        print(buf, "×10")
-        _, state = ret
-        ret = iterate(ryustr, state)
-        while ret !== nothing
-            c, state = ret
-            if '0' <= c <= '9'
-                print(buf, superscript_numerals[c - '0' + 1])
-            elseif c == '-'
-                print(buf, '⁻')
-            end
-            ret = iterate(ryustr, state)
+    leading_index = findfirst(c -> '1' <= c <= '9', power)
+
+    if leading_index === nothing
+        print(buf, superscript_numerals[1])
+        return String(take!(buf))
+    end
+
+    for digit in power[leading_index:end]
+        if digit == '-'
+            print(buf, '⁻')
+        elseif '0' <= digit <= '9'
+            print(buf, superscript_numerals[digit - '0' + 1])
         end
+
     end
 
     return String(take!(buf))
 end
 
+
+function get_engineering_string(x::AbstractFloat, precision::Integer)
+    e_format_number = Ryu.writeexp(x, precision)
+    base_digits, power = split(e_format_number, 'e')
+
+    int_power = parse(Int, power)
+    positive = int_power >= 0
+
+    # round the power to the nearest multiple of 3
+    # positive power -> move the "." to the right by mode, round the power to the higher power
+    # negative power -> move the "." to the right by mode, round the power to the lower power
+    # ex:
+    # 1.2334e5 = 123.334e3
+    # 1.2334-5 = 12.3334e-6
+
+    if positive
+        indices_to_move = int_power % 3
+    else
+        indices_to_move = 3 - abs(int_power) % 3
+    end
+
+    buf = IOBuffer()
+    for i in eachindex(base_digits)
+        if base_digits[i] != '.'
+            print(buf, base_digits[i])
+        end
+        if i == 2 + indices_to_move
+            print(buf, '.')
+        end
+    end
+
+    return String(take!(buf)), string(int_power - indices_to_move)
+end

test/runtests.jl

@@ -1,6 +1,7 @@
 using Showoff
 using Test
 using Dates
+using Printf
 
 @testset "Internals" begin
     @test Showoff.@grisu_ccall(1, 2, 3) === nothing
@@ -32,21 +33,25 @@ end
     @test Showoff.format_fixed_scientific(Inf, 1, false) == "∞"
     @test Showoff.format_fixed_scientific(-Inf, 1, false) == "-∞"
     @test Showoff.format_fixed_scientific(NaN, 1, false) == "NaN"
-    @test Showoff.format_fixed_scientific(0.012345678, 4, true) == "12.34568×10⁻³"
-    @test Showoff.format_fixed_scientific(0.012345678, 4, false) == "1.234568×10⁻²"
-    @test Showoff.format_fixed_scientific(-10.0, 4, false) == "-1.000×10¹"
+    @test Showoff.format_fixed_scientific(0.012345678, 4, true) == "12.346×10⁻³"
+    @test Showoff.format_fixed_scientific(0.012345678, 4, false) == "1.2346×10⁻²"
+    @test Showoff.format_fixed_scientific(-10.0, 4, false) == "-1.0000×10¹"
+    @test Showoff.format_fixed_scientific(-10.0, 4, false) == "-1.0000×10¹"
+    @test Showoff.format_fixed_scientific(-10.0, 4, false)[1:end-5] == @sprintf("%0.4e", -10.0)[1:end-4]
+    @test Showoff.format_fixed_scientific(1.23456e7, 3, false)[1:end-5] == @sprintf("%0.3e", 1.23456e7)[1:end-4]
+    @test Showoff.format_fixed_scientific(2.99999999999999956E-16, 2, false) == "3.00×10⁻¹⁶"
 end
 
 @testset "Showoff" begin
     x = [1.12345, 4.5678]
     @test showoff(x) == ["1.12345", "4.56780"]
-    @test showoff([0.0, 50000.0]) == ["0", "5×10⁴"]
+    @test showoff([0.0, 50000.0]) == ["0", "5.0×10⁴"]
     @test showoff(x, :plain) == ["1.12345", "4.56780"]
-    @test showoff(x, :scientific) == ["1.12345×10⁰", "4.56780×10⁰"]
-    @test showoff(x, :engineering) == ["1.12345×10⁰", "4.56780×10⁰"]
+    @test showoff(x, :scientific) == ["1.123450×10⁰", "4.567800×10⁰"]
+    @test showoff(x, :engineering) == ["1.123450×10⁰", "4.567800×10⁰"]
     @test showoff([DateTime("2017-04-11", "yyyy-mm-dd")]) == ["Apr 11, 2017"]
     @test showoff(["a", "b"]) == ["\"a\"", "\"b\""]
-    @test showoff([1, 1e39]) == ["1×10⁰", "1×10³⁹"]
+    @test showoff([1, 1e39]) == ["1.0×10⁰", "1.0×10³⁹"]
     @test_throws ArgumentError showoff(x, :nevergonnagiveyouup)
     @test_throws ArgumentError showoff([Inf, Inf, NaN])
 end