Namespace collision

[joshday]

I apparently can't use a variable with the name d

julia> using RecipesBase, Plots

julia> type A end

julia> @recipe function f(::A)
           d = 100
           randn(d)
       end
apply_recipe (generic function with 66 methods)

julia> plot(A())
ERROR: MethodError: `convert` has no method matching convert(::Type{Dict{Symbol,Any}}, ::Array{Float64,1})
This may have arisen from a call to the constructor Dict{Symbol,Any}(...),
since type constructors fall back to convert methods.
Closest candidates are:
  Dict{K,V}(::Any)
  call{T}(::Type{T}, ::Any)
  convert{K,V}(::Type{Dict{K,V}}, ::Dict{K,V})
  ...
 in apply_recipe at /Users/joshday/.julia/v0.4/RecipesBase/src/RecipesBase.jl:237
 in _plot! at /Users/joshday/.julia/v0.4/Plots/src/plot.jl:292

[tbreloff]

Yup. I don't think that's going to change.

On Saturday, June 18, 2016, Josh Day [email protected] wrote:

I apparently can't use a variable with the name d

julia> using RecipesBase, Plots

julia> type A end

julia> @recipe function f(::A)
d = 100
randn(d)
end
apply_recipe (generic function with 66 methods)

julia> plot(A())
ERROR: MethodError: convert has no method matching convert(::Type{Dict{Symbol,Any}}, ::Array{Float64,1})
This may have arisen from a call to the constructor Dict{Symbol,Any}(...),
since type constructors fall back to convert methods.
Closest candidates are:
Dict{K,V}(::Any)
call{T}(::Type{T}, ::Any)
convert{K,V}(::Type{Dict{K,V}}, ::Dict{K,V})
...
in apply_recipe at /Users/joshday/.julia/v0.4/RecipesBase/src/RecipesBase.jl:237
in _plot! at /Users/joshday/.julia/v0.4/Plots/src/plot.jl:292

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
#10, or mute the
thread
https://github.com/notifications/unsubscribe/AA492k0w_byMBxyB05ICVmyuNEmeCOL3ks5qM_e0gaJpZM4I4839
.

[joshday]

I commonly see/use n, m, p, and d used to denote dimensions of matrices. Could it be changed to something more obscure? It's a pretty confusing error as it's not clear what I'm doing "wrong".

[joshday]

For clarity, I ran into this doing something like

type A
    mat::Matrix{Float64}
end

@recipe function f(a::A)
    n, d = size(a.mat)
    ...
end

[tbreloff]

I could see changing it to 'kw', which shouldn't clash as much. The problem
is that it's a far-reaching breaking change and would require a coordinated
effort for every package using recipes.

On Saturday, June 18, 2016, Josh Day [email protected] wrote:

For clarity, I ran into this doing something like

type A
mat::Matrix{Float64}end
@recipe function f(a::A)
n, d = size(a.mat)
...end

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
#10 (comment),
or mute the thread
https://github.com/notifications/unsubscribe/AA492s32MW3UYI5wS-RUJ1OdHQ1JHG6Aks5qNAHGgaJpZM4I4839
.

[joshday]

That's a fair reason for wanting to keep it as is, but it's a strange gotcha. Is there a way to get a more informative error?

[tbreloff]

I suppose i could search for 'd' on the lhs of an equals and warn.

And I do see that 'kw' might be better than 'd'. It's just gonna cause a
dependency mess to change.

On Saturday, June 18, 2016, Josh Day [email protected] wrote:

That's a fair reason for wanting to keep it as is, but it's a strange
gotcha. Is there a way to get a more informative error?

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
#10 (comment),
or mute the thread
https://github.com/notifications/unsubscribe/AA492ojNj6Q7XpJXPALU2BYKFK2aeWvOks5qNAbxgaJpZM4I4839
.

[diegozea]

The whole point of having arg --> value and similar notations was not to expose d at the user. Maybe the remaining uses of d should be manage in a similar way, letting to the macro hygiene the manage of this problem (I guest... macros aren't my strong point).

[tbreloff]

'-->' and ':=' notation is certainly to reduce the need to refer to 'd'
explicitly, but there's enough cases where it's still necessary. It's a
tricky problem to avoid clashes completely through macro-fu. A warning on
re-binding should be plenty.

On Saturday, June 18, 2016, Diego Javier Zea [email protected]
wrote:

The whole point of having arg --> value and similar notations was not to
expose d at the user. Maybe the remaining uses of d should be manage in a
similar way, letting to the macro hygiene the manage of this problem (I
guest... macros aren't my strong point).

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
#10 (comment),
or mute the thread
https://github.com/notifications/unsubscribe/AA492gG-Dq-TsXY4EF7zCKMsncPCRIwyks5qNCKngaJpZM4I4839
.

[jw3126]

If you want to change from d to kw (I would actually prefer something more verbose, maybekeywords) I don't think it would be more hurtful then any of the other gazillion breaking changes julia programmers face day by day ;) You could proceed as follows:

Step 1) emit deprecations warnings if kw is used inside recipes.
Step 2) allow both d and kw, but emit warnings if d is used.
Step 3) disallow d

[mkborregaard]

@jw3126 I'd support this if you'd make a coordinated set of PR's implementing this to 1) all repos in JuliaPlots, and 2) all packages depending on RecipesBase that have used d (or the term keywords), and 3) documented the change in the PlotDocs. There is a map of the dependencies here:

[jw3126]

@mkborregaard Wow. How does one find the packages depending on a given package? If you have a list of packages depending on Plots/RecipiesBase I can at least do a bit of grepping and see if these PRs are possible with sane amount of work.

[mkborregaard]

👍 Pkg.dependents

[jw3126]

ApproxFun
    Plot.jl
AtariAlgos
    AtariAlgos.jl
AverageShiftedHistograms
    bivariate.jl
    univariate.jl
BenchmarkProfiles
ConcaveHull
    ConcaveHull.jl
ConstructiveSolidGeometry
ControlSystems
DiffEqBase
    integrator_interface.jl
    monte_solutions.jl
    solution_interface.jl
DiffEqDevTools
    plotrecipes.jl
DiffEqNoiseProcess
    recipes.jl
DiffEqPDEBase
    plotrecipes.jl
DimensionalPlotRecipes
    DimensionalPlotRecipes.jl
DynamicMovementPrimitives
    utilities.jl
        92     delete!(d,:y0)
EEG
ExperimentalAnalysis
ImageQuilting
    plot_recipes.jl
ImplicitEquations
    plot_recipe.jl
        143     xlims = get(d,:xlims, (-5,5))
        144     ylims = get(d, :ylims, (-5,5))
IntervalArithmetic
    plot_recipes.jl
IterativeSolvers
    history.jl
JWAS
LPVSpectral
    plotting.jl
        21     delete!(d,:Fs)
        89     delete!(d, :normalization)
        90     delete!(d, :normdim)
        93         delete!(d, :dims)
        137     delete!(d, :phase)
        138     delete!(d, :bounds)
        139     delete!(d, :nMC)
        140     delete!(d, :mcmean)
LossFunctions
    io.jl
MIToS
    Plots.jl
    Plots.jl
Mimi
PenaltyFunctions
    elementpenalty.jl
PhyloTrees
    plot.jl
PlotRecipes
    finance.jl
        20         @series Plots.isvertical(d) ? (sx, sy) : (sy, sx)
    graphs.jl
        300     if get(d, :linewidth, 1) > 0
        316                                     xview=d[:xlims], yview=d[:ylims], root=root)
        335             grad = get(d, :linecolor, nothing)
        362         scalefactor = pop!(d, :markersize, nodesize)
        364         nodeshape = get(d, :markershape, nodeshape)
        416         colorgradient = ColorGradient(get(d,:linecolor,cgrad()))
        498     grad = cgrad(get(d,:linecolor,:inferno), get(d, :linealpha, nothing))
    misc.jl
    pkg_deps.jl
Plots
    recipes.jl
        82     xmin, xmax = hvline_limits(d[:subplot][:xaxis])
        94     ymin, ymax = hvline_limits(d[:subplot][:yaxis])
        126     d[:x], d[:y] = make_steps(x, y, :steppre)
        130     if d[:markershape] != :none
        147     d[:x], d[:y] = make_steps(x, y, :steppost)
        151     if d[:markershape] != :none
        173     fr = d[:fillrange]
        175         yaxis = d[:subplot][:yaxis]
        194     if d[:markershape] != :none
        227     fr = d[:fillrange]
        277     procx, procy, xscale, yscale, baseline = _preprocess_barlike(d, x, y)
        279     axis = d[:subplot][isvertical(d) ? :xaxis : :yaxis]
        290     bw = d[:bar_width]
        298     fillto = d[:fillrange]
        323     if !isvertical(d)
        392     edge, weights, xscale, yscale, baseline = _preprocess_binlike(d, x, y)
        393     if (d[:bar_width] == nothing)
        405     edge, weights, xscale, yscale, baseline = _preprocess_binlike(d, x, y)
        468     axis = d[:subplot][Plots.isvertical(d) ? :xaxis : :yaxis]
        470     edge, weights, xscale, yscale, baseline = _preprocess_binlike(d, x, y)
        473     if !isvertical(d)
        478     if d[:markershape] != :none
        559     h = _make_hist((y,), d[:bins], normed = d[:normalize], weights = d[:weights])
        568     h = _make_hist((y,), d[:bins], normed = d[:normalize], weights = d[:weights])
        577     h = _make_hist((y,), d[:bins], normed = d[:normalize], weights = d[:weights])
        593     seriestype := get(st_map, d[:seriestype], d[:seriestype])
        595     if d[:seriestype] == :scatterbins
        597         edge, weights, xscale, yscale, baseline = _preprocess_binlike(d, h.edges[1], h.weights)
        644     h = _make_hist((x, y), d[:bins], normed = d[:normalize], weights = d[:weights])
        665     if d[:markershape] == :none
        728     error_style!(d)
        730     d[:x], d[:y] = error_coords(d[:x], d[:y], error_zipit(d[:yerror]))
        736     error_style!(d)
        738     d[:y], d[:x] = error_coords(d[:y], d[:x], error_zipit(d[:xerror]))
        837         quiver_using_arrows(d)
        839         quiver_using_hack(d)
        943     if d[:markershape] == :none
        946     if d[:markersize] == default(:markersize)
    series.jl
        128     xs, _ = convertToAnyVector(x, d)
        129     ys, _ = convertToAnyVector(y, d)
        130     zs, _ = convertToAnyVector(z, d)
        132     fr = pop!(d, :fillrange, nothing)
        136         convertToAnyVector(fr, d)
        148             di = copy(d)
        205     newx = _apply_type_recipe(d, x)
        207     newy = _apply_type_recipe(d, y)
        209     newz = _apply_type_recipe(d, z)
        219     newx = _apply_type_recipe(d, x)
        221     newy = _apply_type_recipe(d, y)
        230     newy = _apply_type_recipe(d, y)
        243         newv = _apply_type_recipe(d, v)
        270 @recipe f(n::Integer) = is3d(get(d,:seriestype,:path)) ? (SliceIt, n, n, n) : (SliceIt, n, n, nothing)
        274     if all3D(d)
        276         wrap_surfaces(d)
        285     if all3D(d)
        288         wrap_surfaces(d)
        328         z, d[:fillcolor] = replace_image_with_heatmap(mat)
        358     plt = d[:plot_object]
        407     wrap_surfaces(d)
        417     wrap_surfaces(d)
        425     if !like_surface(get(d, :seriestype, :none))
        426         d[:seriestype] = :contour
        428     wrap_surfaces(d)
        480 @recipe f{R1<:Number,R2<:Number,R3<:Number,R4<:Number}(xyuv::AVec{Tuple{R1,R2,R3,R4}}) = get(d,:seriestype,:path)==:ohlc ? OHLC[OHLC(t...) for t in xyuv] : unzip(xyuv)
    runtests.jl
PyDSTool
    bifurcation.jl
        9   x = bif.d[coords[1]]
        10   y = bif.d[coords[2]]
RationalFunctions
    plotting.jl
RecursiveArrayTools
    vector_of_array.jl
Reinforce
    cartpole.jl
    mountain_car.jl
    pendulum.jl
Robotlib
    Frames.jl

WARNING: deprecated syntax "typealias Vect{Ty} Union{AbstractVector{Ty},Vec{3,Ty}}".
Use "Vect{Ty} = Union{AbstractVector{Ty},Vec{3,Ty}}" instead.

WARNING: deprecated syntax "typealias Matr{Ty} Union{AbstractMatrix{Ty},Mat{3,3,Ty}}".
Use "Matr{Ty} = Union{AbstractMatrix{Ty},Mat{3,3,Ty}}" instead.

WARNING: deprecated syntax "abstract GeometricObject".
Use "abstract type GeometricObject end" instead.

WARNING: deprecated syntax "typealias Vect{Ty} Union{AbstractVector{Ty},Vec{3,Ty}}".
Use "Vect{Ty} = Union{AbstractVector{Ty},Vec{3,Ty}}" instead.

WARNING: deprecated syntax "typealias Matr{Ty} Union{AbstractMatrix{Ty},Mat{3,3,Ty}}".
Use "Matr{Ty} = Union{AbstractMatrix{Ty},Mat{3,3,Ty}}" instead.

WARNING: deprecated syntax "abstract GeometricObject".
Use "abstract type GeometricObject end" instead.
    robotplot.jl
Sims
    utils.jl
SingularIntegralEquations
    plot.jl
SpatialEcology
    PlotRecipes.jl
StatPlots
    bar.jl
        10     isstack = pop!(d, :bar_position, :dodge) == :stack
        30         bws = d[:bar_width] / nc
        56         get(d, :fillrange, nothing)
    boxplot.jl
        12     bw = d[:bar_width]
        33         center = Plots.discrete_value!(d[:subplot][:xaxis], glabel)[1]
        86     if d[:linecolor] == d[:fillcolor]
        87 	d[:linecolor] = d[:markerstrokecolor]
        94 	    if get!(d, :markershape, :circle) == :none
        95             	d[:markershape] = :circle
        97             markercolor := d[:fillcolor]
        98             markerstrokecolor := d[:linecolor]
    cornerplot.jl
        18     labs = pop!(d, :label, ["x$i" for i=1:N])
        43     fillcolor --> Plots.fg_color(d)
        44     linecolor --> Plots.fg_color(d)
        49     grad = cgrad(get(d, :markercolor, cgrad()))
    corrplot.jl
        15     labs = pop!(d, :label, [""])
        23     fillcolor --> Plots.fg_color(d)
        24     linecolor --> Plots.fg_color(d)
        26     grad = cgrad(get(d, :markercolor, cgrad()))
        28     title = get(d,:title,"")
        37             update_ticks_guides(d, labs, i, i, n)
        50             update_ticks_guides(d, labs, i, j, n)
        63                     seriestype := get(d, :seriestype, :histogram2d)
    dataframes.jl
        68         if haskey(d, k)
        69             d[k] = processDFarg(df, d[k])
        74     tuple(Any[handle_dfs(df, d, (i==1 ? "x" : i==2 ? "y" : "z"), arg) for (i,arg) in enumerate(args)]...)
    distributions.jl
    groupederror.jl
        289     if !(:seriestype in keys(d)) || (d[:seriestype] == :path)
        300     elseif d[:seriestype] == :scatter
        313     elseif d[:seriestype] == :bar
    hist.jl
        7     if Plots.isvertical(d)
        25     if Plots.isvertical(d)
    marginalhist.jl
        4     x, y = d[:x], d[:y]
        27     fillcolor --> Plots.fg_color(d)
        28     linecolor --> Plots.fg_color(d)
        38     bns = get(d, :bins, nothing)
    shadederror.jl
        15     x,y = d[:x],d[:y]
    StatPlots.jl
    violin.jl
        21     bw = d[:bar_width]
        32         xcenter = Plots.discrete_value!(d[:subplot][:xaxis], glabel)[1]
SymEngine
    recipes.jl
SymPy
    plot_recipes.jl
        184     xlims = get(d,:xlims, (-5,5))
        185     ylims = get(d, :ylims, (-5,5))
TimeSeries
    plotrecipes.jl
        3     st = get(d, :seriestype, :path)
        47     st = get(d, :seriestype, :candlestick)
        55     bw = get(d, :bar_width, nothing)
        60     cols = get(d, :seriescolor, nothing)
ValueHistories
    recipes.jl
        21         if get(d, :layout, nothing) != nothing
        22             title  --> d[:label]

[jw3126]

This is all files that contain @recipe and are part of packages depending on Plots or RecipesBase. Lines inside @recipes that contain d are printed. There are no such lines containing keywords.
I think the @mkborregaard s PR idea looks feasible.

[jw3126]

Since keywords is used in none of these packages, I think we could skip Step 1 in

Step 1) emit deprecations warnings if keywords is used inside recipes.
Step 2) allow both d and keywords , but emit warnings if d is used.
Step 3) disallow d

Any thoughts?

[mkborregaard]

Hey, thanks this is great. One thought - with regards the potential for conflicts, I believe they may in fact be even bigger with keywords (which is often used to pass kwargs, I think) than d - maybe a better name would be the descriptive plotattributes?
I should say I'm headed offline so I cannot give any support to this over the next weeks - you can choose to wait if you'd need me to help or go ahead with the assistance of other JuliaPlots members.

[daschw]

fixed by #29