Support plotting surface data

[jebej]

This would require a method for Coordinates that takes x,y, and z where z is a matrix such that size(z,1)==length(x) and size(z,2)==length(y)

[KristofferC]

Good suggestion. Could you for reference link to an example or a stackoverflow post with some code so I can see how the tex should look.

[jebej]

The answers to this question show the format, and a use both with "matrix plot" and "surf". Basically you need to give x y z points.

I am personally trying to do a matrix plot (page 171 of the manual), but I find that doing

Plot(data,
    {
        "matrix plot"
    })

does not work.

The code for the data coordinates looks like this, which is what the Coordinates method could do automatically given x,y,z:

xx = repmat(x,length(y))
yy = vec(repmat(y.',length(x))
zz = vec(z)
data = Coordinates(xx,yy,zz)

[KristofferC]

Are the empty lines needed for the output to be correct or are they just for clarity in the examples?

[jebej]

I think there needs to be one, although the number of rows or columns can be specified in which case it's not necessary:

Limitations: Due to current implementational restrictions, matrix plot can only update axis limits
successfully if the length of the matrix scanlines is known in advance. The example above contains
mesh/cols=3, and a “scanline” is a row (with 3 columns each). Consequently, the example above works
fine. However, If mesh/cols=3 would be unknown in advance, the current implementation assumes that
end-of-scanline markers are given in the input file according to empty line=scanline. As a rule of
thumb, you have to follow the following guidelines when using matrix plot:

1. The input matrix should have an empty line whenever one line of input is finished.
2. If there is no empty line, the length of each scanline must be given. For the standard con-
   figuration mesh/ordering=x varies, pgfplots expects mesh/cols. For the alternative choice
   mesh/ordering=y varies, pgfplots expects mesh/rows to be set.
3. The input matrix must have at least 2 rows and at least 2 columns. This allows pgfplots to
   interpolate/extrapolate the vertices in a well-defined way

[jebej]

I think my case will be easier with a table. Will make an example as soon as I figure it out.

[pevnak]

I have encountered the same problem.
at the moment the package produces output like this

\RequirePackage{luatex85}
\documentclass[tikz]{standalone}
% Default preamble
\usepackage{pgfplots}
\pgfplotsset{compat=newest}
\usepgfplotslibrary{groupplots}
\usepgfplotslibrary{polar}
\usepgfplotslibrary{statistics}
\begin{document}
\begin{tikzpicture}[]
\begin{axis}[xlabel={trnuses}, ylabel={tstuses}]
\addplot[matrix plot*, point meta=explicit]
table [x={trnuses}, y={tstuses}, meta={tsterr}]
{ trnuses tstuses tsterr
0.01 0.01 0.0241
0.01 0.02 0.0237
0.01 0.05 0.023600000000000003
0.01 0.1 0.040600000000000004
0.01 0.2 0.2051
0.01 0.5 0.45289999999999997
0.01 1.0 0.49420000000000003
0.02 0.01 0.024399999999999998
0.02 0.02 0.023799999999999998
0.02 0.05 0.023899999999999998
0.02 0.1 0.0403
0.02 0.2 0.1996
0.02 0.5 0.45099999999999996
0.02 1.0 0.49450000000000005
0.05 0.01 0.0242
0.05 0.02 0.0242
0.05 0.05 0.0237
0.05 0.1 0.042
0.05 0.2 0.2113
0.05 0.5 0.4574
0.05 1.0 0.49450000000000005
0.1 0.01 0.0256
0.1 0.02 0.0257
0.1 0.05 0.0253
0.1 0.1 0.0361
0.1 0.2 0.1131
0.1 0.5 0.378
0.1 1.0 0.4755000000000001
0.2 0.01 0.036500000000000005
0.2 0.02 0.037000000000000005
0.2 0.05 0.036500000000000005
0.2 0.1 0.042600000000000006
0.2 0.2 0.05720000000000001
0.2 0.5 0.2472
0.2 1.0 0.3833
0.5 0.01 0.0867
0.5 0.02 0.0869
0.5 0.05 0.0865
0.5 0.1 0.0891
0.5 0.2 0.09729999999999998
0.5 0.5 0.1554
0.5 1.0 0.27549999999999997
1.0 0.01 0.146
1.0 0.02 0.14609999999999998
1.0 0.05 0.1458
1.0 0.1 0.1473
1.0 0.2 0.15430000000000002
1.0 0.5 0.1793
1.0 1.0 0.2505
}
;
\end{axis}
\end{tikzpicture}
\end{document}

which does not compile. After you add the blank lines

\RequirePackage{luatex85}
\documentclass[tikz]{standalone}
% Default preamble
\usepackage{pgfplots}
\pgfplotsset{compat=newest}
\usepgfplotslibrary{groupplots}
\usepgfplotslibrary{polar}
\usepgfplotslibrary{statistics}
\begin{document}
\begin{tikzpicture}[]
\begin{axis}[xlabel={trnuses}, ylabel={tstuses}]
\addplot[matrix plot*, point meta=explicit]
table [x={trnuses}, y={tstuses}, meta={tsterr}]
{ trnuses tstuses tsterr
0.01 0.01 0.0241
0.01 0.02 0.0237
0.01 0.05 0.023600000000000003
0.01 0.1 0.040600000000000004
0.01 0.2 0.2051
0.01 0.5 0.45289999999999997
0.01 1.0 0.

       0.02	0.01	0.024399999999999998
       0.02	0.02	0.023799999999999998
       0.02	0.05	0.023899999999999998
       0.02	0.1	0.0403
       0.02	0.2	0.1996
       0.02	0.5	0.45099999999999996
       0.02	1.0	0.49450000000000005

       0.05	0.01	0.0242
       0.05	0.02	0.0242
       0.05	0.05	0.0237
       0.05	0.1	0.042
       0.05	0.2	0.2113
       0.05	0.5	0.4574
       0.05	1.0	0.49450000000000005

       0.1	0.01	0.0256
       0.1	0.02	0.0257
       0.1	0.05	0.0253
       0.1	0.1	0.0361
       0.1	0.2	0.1131
       0.1	0.5	0.378
       0.1	1.0	0.4755000000000001

       0.2	0.01	0.036500000000000005
       0.2	0.02	0.037000000000000005
       0.2	0.05	0.036500000000000005
       0.2	0.1	0.042600000000000006
       0.2	0.2	0.05720000000000001
       0.2	0.5	0.2472
       0.2	1.0	0.3833

       0.5	0.01	0.0867
       0.5	0.02	0.0869
       0.5	0.05	0.0865
       0.5	0.1	0.0891
       0.5	0.2	0.09729999999999998
       0.5	0.5	0.1554
       0.5	1.0	0.27549999999999997
       
       1.0	0.01	0.146
       1.0	0.02	0.14609999999999998
       1.0	0.05	0.1458
       1.0	0.1	0.1473
       1.0	0.2	0.15430000000000002
       1.0	0.5	0.1793
       1.0	1.0	0.2505
       }
   ;

\end{axis}
\end{tikzpicture}

\end{document}

it works.

I apologize for broken formatting, just do not know, how to do it properly.

[tpapp]

Supporting the table format would also allow contour plots. See the countour prepared keywords in Section 4.6.8 of the pgfplots manual. A tool like Contour.jl would allow the computation of the curves without external tools.

[KristofferC]

Using Contour.jl already works (see last example in https://github.com/KristofferC/PGFPlotsXExamples/blob/master/examples/custom_types.ipynb).

[tpapp]

@jebej: this reproduces the first example on p 171:

x = repeat(0:2, outer = 3)
y = repeat(0:2, inner = 3)
meta = ["color=$c" for c in ["red", "blue", "yellow", "black", "brown", "magenta", "green", "red", "white"]]
c = Coordinates(Float64.(hcat(x, y))'; metadata = meta)
p = @pgf Axis(Plot(c,
          {
            matrix_plot,
            mark = "*",
            nodes_near_coords = raw"\coordindex",
            "mesh/color input" = "explicit",
            "mesh/cols" = 3,
        }), { enlargelimits = false })

Will make a PR to the examples repo soon.

[CarolinGao12]

I have the same problem, but in my case, my x,y, are not easily to be formed in a table. Is there any solution for x, y, z , which are all matrix , so that it is accepted by surface plot ?

[tpapp]

@CarolinGao12 : please provide a minimal working example. There is a

Table([options], ::AbstractVector, ::AbstractVector, ::AbstractMatrix; ...)

constructor (implemented by TableData) but I don't know how you specify coordinates from matrices.

[CarolinGao12]

it seems the x, y input should be a vector , and z can be a matrix
but in my case, the xp_k and yp_k are matrixs

for k = 1: numAnt
    xp_k=[xp[k,:,:];reshape(xp[k,1,:],(1,91))]
    yp_k=[yp[k,:,:];reshape(yp[k,1,:],(1,91))]
    zp_k=[zp[k,:,:];reshape(zp[k,1,:],(1,91))]
    @pgf Plot3(
    {
        surf,
    },
    Coordinates(xp_k,yp_k,zp_k)
    )
end     

[KristofferC]

So perhaps we should add:

function Coordinates(x::AbstractMatrix, y::AbstractMatrix, z::AbstractMatrix;
                     meta::Union{Void, AbstractMatrix} = nothing)
    @argcheck size(x) == size(y) == size(z)
    meta ≠ nothing && @argcheck size(meta) == size(z)
    insert_scanlines(Coordinates(vec(x), vec(y), vec(z);
                     meta = meta ≠ nothing ? vec(meta) : meta),
                     size(z, 2))
end

giving

julia> print_tex(Coordinates([1 1; 2 2], [1 2; 1 2], [1 2; 3 4]))
coordinates {
    (1, 1, 1)
    (2, 1, 3)

    (1, 2, 2)
    (2, 2, 4)
}

But what if both x and y is changing in one "scan line", not sure what happens then.

[CarolinGao12]

thanks, I change the matrix to the vector and this problem solved

[tpapp]

1. Following the pgfplots manual, I would suggest the Table constructors.

2. @CarolinGao12, it would be great if you could provide a self-contained minimal working example. I am assuming you are generating the x and y somehow by repeating elements, that step could be omitted.

3. @KristofferC: interesting question. Apparently pgfplots will just omit the node that does not belong there. Example:

using PGFPlotsX
@pgf Plot3({ surf, shader = "interp" },
           Table([:x => vec([1 1; 2 2]),
                  :y => vec([1 2; 1 2]),
                  :z => vec([1 2; 3 4])],
                 scanlines = 2))
pgfsave("/tmp/regular.png", ans)
@pgf Plot3({ surf, shader = "interp" },
           Table([:x => vec([1 1; 1 2]),
                  :y => vec([1 2; 2 2]),
                  :z => vec([1 2; 3 4])],
                 scanlines = 2))
pgfsave("/tmp/irregular.png", ans)

I don't think we should add a constructor for this, as it would just be confusing.

[CarolinGao12]

    @pgf begin 
     
    p1=Plot3(
    {
        surf,
        opacity=0.6,
        "colormap/jet",
           shader=>"faceted interp",
            "faceted color= black",
            
   }, 
   Table([:x => vec(xp_1),
                  :y => vec(yp_1),
                  :z => vec(zp_1)],
                 scanlines = 2)
   )

    

it works fine