AdvancedUnicode
plotting library designed for use inJulia
'sREPL
.
Here is a list of the main high-level functions for common scenarios:
lineplot
(Line Plot)scatterplot
(Scatter Plot)stairs
(Staircase Plot)barplot
(Bar Plot - horizontal)histogram
(Histogram - horizontal / vertical)boxplot
(Box Plot - horizontal)spy
(Sparsity Pattern)densityplot
(Density Plot)contourplot
(Contour Plot)polarplot
(Polar Plot)heatmap
(Heatmap Plot)imageplot
(Image Plot)surfaceplot
(Surface Plot - 3D)isosurface
(Isosurface Plot - 3D)
Introduction
Here is a quick hello world example of a typical use-case:
usingUnicodePlots
lineplot([-1,2,3,7], [-1,2,9,4], title="Example",name="my line",xlabel="x",ylabel="y")
There are other types ofCanvas
available (see sectionLow-level Interface).
In some situations, such as printing to a file, usingAsciiCanvas
,DotCanvas
orBlockCanvas
might lead to better results:
plt=lineplot([-1,2,3,7], [-1,2,9,4], title="Example",name="my line",
xlabel="x",ylabel="y",canvas=DotCanvas, border=:ascii)
Some plot methods have a mutating variant that ends with an exclamation mark:
lineplot!(plt, [0,4,8], [10,1,10], color=:cyan,name="other line")
These mutating methods cannot update the limits of the axes as plots are drawn onto a fixed canvas. The limits must be set beforehand by the plotting function that creates the figure or by creating an emptyPlot
:
p=Plot(; xlim=(-1,3), ylim=(-1,3))
lineplot!(p,1:2)
One can adjust the plotheight
andwidth
to the current terminal size by usingheight=:auto
and/orwidth=:auto
.
You can reverse/flip thePlot
axes by settingxflip=true
and/oryflip=true
on plot creation.
Lineplot
lineplot([1,2,7], [9,-6,8], title="My Lineplot")
It's also possible to specify a function and a range:
plt=lineplot(-π/2,2π, [cos, sin])
You can also plot lines by specifying an intercept and slope:
lineplot!(plt,-.5,.2,name="line")
Plotting multiple series is supported by providing aMatrix
(<: AbstractMatrix
) for they
argument, with the individual series corresponding to its columns. Auto-labeling is by default, but you can also label each series by providing aVector
or a1xn
Matrix
such as[ "series 1" "series2"...]
:
lineplot(1:10,[0:93:12reverse(5:14)fill(4,10)], color=[:green:red:yellow:cyan])
Physical quantities ofUnitful.jl
are supported throughpackage extensions - weak dependencies:
usingUnitful
a, t=1u"m/s^2",(0:100)*u"s"
lineplot(a/2*t.^2,a*t, xlabel="position",ylabel="speed",height=10)
Intervals fromIntervalSets.jl
are supported:
usingIntervalSets
lineplot(-1..3,x->x^5-5x^4+5x^3+5x^2-6x-1;name="quintic")
Usehead_tail
to mimic plotting arrows (:head
,:tail
or:both
) where the length of the "arrow" head or tail is controlled usinghead_tail_frac
where e.g. giving a value of0.1
means10%
of the segment length:
lineplot(1:10,1:10,head_tail=:head,head_tail_frac=.1,height=4)
UnicodePlots
exportshline!
andvline!
for drawing vertical and horizontal lines on a plot:
p=Plot([NaN], [NaN]; xlim=(0,8), ylim=(0,8))
vline!(p, [2,6], [2,6], color=:red)
hline!(p, [2,6], [2,6], color=:white)
hline!(p,7,color=:cyan)
vline!(p,1,color=:yellow)
Scatterplot
scatterplot(randn(50),randn(50), title="My Scatterplot")
Axis scaling (xscale
and/oryscale
) is supported: choose from (:identity
,:ln
,:log2
,:log10
) or use an arbitrary scale function:
scatterplot(1:10,1:10,xscale=:log10,yscale=:log10)
For the axis scale exponent, one can revert to usingASCII
characters instead ofUnicode
ones using the keywordunicode_exponent=false
:
scatterplot(1:4,1:4,xscale=:log10,yscale=:ln,unicode_exponent=false,height=6)
Using amarker
is supported, choose aChar
,a unit lengthString
or a symbol name such as:circle
(more fromkeys(UnicodePlots.MARKERS)
).
One can also provide a vector ofmarker
s and/orcolor
s as in the following example:
scatterplot([1,2,3], [3,4,1], marker=[:circle,'',"∫"],
color=[:cyan,nothing,:yellow], height=2)
As withlineplot
,scatterplot
supports plotting physicalUnitful
quantities, or plotting multiple series (Matrix
argument).
Staircase plot
stairs([1,2,4,7,8], [1,3,4,2,7],
color=:yellow,style=:post,height=6,title="Staircase")
Barplot
barplot(["Paris","New York","Madrid"], [2.244,8.406,3.165], title="Population")
Note:You can use the keyword argumentsymbols
to specify the characters that should be used to plot the bars (e.g.symbols=['#']
).
Histogram
histogram(randn(1_000).*.1,nbins=15,closed=:left)
Thehistogram
function also supports axis scaling using the parameterxscale
:
histogram(randn(1_000).*.1,nbins=15,closed=:right,xscale=:log10)
Vertical histograms are supported:
histogram(randn(100_000).*.1,nbins=60,vertical=true,height=10)
Boxplot
boxplot([1,3,3,4,6,10])
boxplot(["one","two"],
[[1,2,3,4,5], [2,3,4,5,6,7,8,9]],
title="Grouped Boxplot",xlabel="x")
Sparsity Pattern
usingSparseArrays
spy(sprandn(50,120,.05))
Plotting the zeros pattern is also possible usingshow_zeros=true
:
usingSparseArrays
spy(sprandn(50,120,.9), show_zeros=true)
Density Plot
plt=densityplot(randn(10_000),randn(10_000))
densityplot!(plt,randn(10_000).+2,randn(10_000).+2)
Using a scale function (e.g. damping peaks) is supported using thedscale
keyword:
x=randn(10_000); x[1_000:6_000].=2
densityplot(x,randn(10_000); dscale=x->log(1+x))
Contour Plot
contourplot(-3:.01:3,-7:.01:3,(x, y)->exp(-(x/2)^2-((y+2)/4)^2))
The keywordlevels
controls the number of contour levels. One can also choose acolormap
as withheatmap
,and disable the colorbar usingcolorbar=false
.
Polar Plot
Plots data in polar coordinates withθ
the angles in radians.
polarplot(range(0,2π, length=20),range(0,2,length=20))
Heatmap Plot
heatmap(repeat(collect(0:10)',outer=(11,1)), zlabel="z")
Theheatmap
function also supports axis scaling using the parametersxfact
,yfact
and axis offsets after scaling usingxoffset
andyoffset
.
Thecolormap
parameter may be used to specify a named or custom colormap. See theheatmap
function documentation for more details.
In addition, thecolorbar
andcolorbar_border
options may be used to toggle the colorbar and configure its border.
Thezlabel
option andzlabel!
method may be used to set thez
axis (colorbar) label.
Use thearray
keyword in order to display the matrix in the array convention (as in the repl).
heatmap(collect(0:30)*collect(0:30)',xfact=.1,yfact=.1,xoffset=-1.5,colormap=:inferno)
Image Plot
Draws an image, surround it with decorations.Sixel
are supported (experimental) under a compatible terminal throughImageInTerminal
(which must be imported beforeUnicodePlots
).
importImageInTerminal#mandatory (triggers glue code loading)
usingTestImages
imageplot(testimage("monarch_color_256"), title="monarch")
Surface Plot
Plots a colored surface using height valuesz
above ax-y
plane, in three dimensions (masking values usingNaN
s is supported).
sombrero(x, y)=15sinc(√(x^2+y^2)/π)
surfaceplot(-8:.5:8,-8:.5:8,sombrero, colormap=:jet)
Uselines=true
to increase the density (underlying call tolineplot
instead ofscatterplot
,with color interpolation).
By default,surfaceplot
scales heights to adjust aspect wrt the remaining axes withzscale=:aspect
.
To plot a slice in 3D, use an anonymous function which maps to a constant value:zscale=z -> a_constant
:
surfaceplot(
-2:2,-2:2,(x, y)->15sinc(√(x^2+y^2)/π),
zscale=z->0,lines=true,colormap=:jet
)
Isosurface Plot
UsesMarchingCubes.jl
to extract an isosurface, whereisovalue
controls the surface isovalue.
Usingcentroid
enables plotting the triangulation centroids instead of the triangle vertices (better for small plots).
Back face culling (hide not visible facets) can be activated usingcull=true
.
One can use the legacy 'Marching Cubes' algorithm usinglegacy=true
.
torus(x, y, z, r=0.2,R=0.5)=(√(x^2+y^2)-R)^2+z^2-r^2
isosurface(-1:.1:1,-1:.1:1,-1:.1:1,torus, cull=true,zoom=2,elevation=50)
...
To install UnicodePlots, start upJulia
and type the following code snippet into theREPL
(makes use of the nativeJulia
package managerPkg
):
julia>usingPkg
julia>Pkg.add("UnicodePlots")
...
Saving plots aspng
ortxt
files using thesavefig
command is supported (saving aspng
is experimental and requiresimport FreeType, FileIO
before loadingUnicodePlots
).
To recover the plot as a string with ansi color codes usestring(p; color=true)
.
...
When theCOLORTERM
environment variable is set to either24bit
ortruecolor
,UnicodePlots
will use24bit colorsas opposed to8bit colorsor even4bit colorsfor named colors.
One can force a specific colormode using eitherUnicodePlots.truecolors!()
orUnicodePlots.colors256!()
.
Named colors such as:red
or:light_red
will use256
color values (rendering will be terminal dependent). In order to force named colors to use true colors instead, useUnicodePlots.USE_LUT[]=true
.
The default color cycle can be changed to bright (high intensity) colors usingUnicodePlots.brightcolors!()
instead of the defaultUnicodePlots.faintcolors!()
.
...
3d plots use a so-called "Model-View-Projection" transformation matrixMVP
on input data to project 3D plots to a 2D screen.
Use keywordselevation
,azimuth
,up
orzoom
to control the view matrix, a.k.a. camera.
Theprojection
type forMVP
can be set to either:persp(ective)
or:ortho(graphic)
.
Displaying thex
,y
,andz
axes can be controlled using theaxes3d
keyword.
For enhanced resolution, use a wider and/or tallerPlot
(this can be achieved usingdefault_size!(width=60)
for all future plots).
...
UnicodePlots
is integrated inPlots
as a backend, with support forbasic layout.
For a more complex layout, use thegridplot
function (requires loadingTerm
as extension).
usingUnicodePlots, Term
(
UnicodePlots.panel(lineplot(1:2))*
UnicodePlots.panel(scatterplot(rand(100)))
)/(
UnicodePlots.panel(lineplot(2:-1:1))*
UnicodePlots.panel(densityplot(randn(1_000),randn(1_000)))
)|>display
gridplot(map(i->lineplot(-i:i),1:5); show_placeholder=true)|>display
gridplot(map(i->lineplot(-i:i),1:3); layout=(2,nothing))|>display
gridplot(map(i->lineplot(-i:i),1:3); layout=(nothing,1))|>display
...
Using a nontrue monospace font
can lead to visual problems on aBrailleCanvas
(border versus canvas).
Either change the font to e.g.JuliaMonoor useborder=:dotted
keyword argument in the plots.
For aJupyter
notebook with theIJulia
kernel seehere.
(Experimental) Terminals seem to respect a standard aspect ratio of4:3
,hence a square matrix does not often look square in the terminal.
You can pass the experimental keywordfix_ar=true
tospy
orheatmap
in order to recover a unit aspect ratio.
...
Non-exhaustive methods description:
-
title!(plot::Plot, title::String)
title
the string to write in the top center of the plot window. If the title is empty the whole line of the title will not be drawn
-
xlabel!(plot::Plot, xlabel::String)
xlabel
the string to display on the bottom of the plot window. If the title is empty the whole line of the label will not be drawn
-
ylabel!(plot::Plot, xlabel::String)
ylabel
the string to display on the far left of the plot window.
The methodlabel!
is responsible for the setting all the textual decorations of a plot. It has two functions:
-
label!(plot::Plot, where::Symbol, value::String)
where
can be any of::tl
(top-left),:t
(top-center),:tr
(top-right),:bl
(bottom-left),:b
(bottom-center),:br
(bottom-right),:l
(left),:r
(right)
-
label!(plot::Plot, where::Symbol, row::Int, value::String)
-
where
can be any of::l
(left),:r
(right) -
row
can be between 1 and the number of character rows of the canvas
-
x=y=collect(1:10)
plt=lineplot(x, y, canvas=DotCanvas, height=10,width=30)
lineplot!(plt, x,reverse(y))
title!(plt,"Plot Title")
forlocin(:tl,:t,:tr,:bl,:b,:br)
label!(plt, loc,string(':',loc))
end
label!(plt,:l,":l")
label!(plt,:r,":r")
foriin1:10
label!(plt,:l,i,string(i))
label!(plt,:r,i,string(i))
end
plt
annotate!(plot::Plot, x::Number, y::Number, text::AbstractString; kw...)
text
arbitrary annotation at position (x, y)
...
All plots support the set (or a subset) of the following named parameters:
-
symbols::Array = ['■']
:collection of characters used to render the bars. -
title::String = ""
:text displayed on top of the plot. -
name::String = ""
:current drawing annotation displayed on the right. -
xlabel::String = ""
:text displayed on thex
axis of the plot. -
ylabel::String = ""
:text displayed on they
axis of the plot. -
zlabel::String = ""
:text displayed on thez
axis (colorbar) of the plot. -
xscale::Symbol =:identity
:x
-axis scale (:identity
,:ln
,:log2
,:log10
), or scale function e.g.x -> log10(x)
. -
yscale::Symbol =:identity
:y
-axis scale. -
labels::Bool = true
:show plot labels.lineplot(1:.5:20,sin, labels=false)
-
border::Symbol =:solid
:plot bounding box style (:corners
,:solid
,:bold
,:dashed
,:dotted
,:ascii
,:none
).lineplot([-1.,2,3,7], [1.,2,9,4], canvas=DotCanvas, border=:dashed)
lineplot([-1.,2,3,7], [1.,2,9,4], canvas=DotCanvas, border=:ascii)
lineplot([-1.,2,3,7], [1.,2,9,4], canvas=DotCanvas, border=:bold)
lineplot([-1.,2,3,7], [1.,2,9,4], border=:dotted)
lineplot([-1.,2,3,7], [1.,2,9,4], border=:none)
-
margin::Int = 3
:number of empty characters to the left of the whole plot. -
padding::Int = 1
:left and right space between the labels and the canvas. -
color::Symbol =:auto
:choose from (:green
,:blue
,:red
,:yellow
,:cyan
,:magenta
,:white
,:normal
,:auto
), use an integer in[0-255]
,or provide3
integers asRGB
components. -
height::Int = 15
:number of canvas rows, or:auto
.lineplot(1:.5:20,sin, height=18)
-
width::Int = 40
:number of characters per canvas row, or:auto
.lineplot(1:.5:20,sin, width=60)
-
xlim::Tuple = (0, 0)
:plotting range for thex
axis ((0, 0)
stands for automatic). -
ylim::Tuple = (0, 0)
:plotting range for they
axis. -
zlim::Tuple = (0, 0)
:colormap scaled data range. -
xticks::Bool = true
:setfalse
to disable ticks (labels) onx
-axis. -
yticks::Bool = true
:setfalse
to disable ticks (labels) ony
-axis. -
xflip::Bool = false
:settrue
to flip thex
axis. -
yflip::Bool = false
:settrue
to flip they
axis. -
colorbar::Bool = false
:toggle the colorbar. -
colormap::Symbol =:viridis
:choose a symbol fromColorSchemes.jl
e.g.:viridis
,or supply a functionf: (z, zmin, zmax) -> Int(0-255)
,or a vector of RGB tuples. -
colorbar_lim::Tuple = (0, 1)
:colorbar limit. -
colorbar_border::Symbol =:solid
:color bar bounding box style (:solid
,:bold
,:dashed
,:dotted
,:ascii
,:none
). -
canvas::UnionAll = BrailleCanvas
:type of canvas used for drawing. -
grid::Bool = true
:draws grid-lines at the origin. -
compact::Bool = false
:compact plot labels. -
unicode_exponent::Bool = true
:useUnicode
symbols for exponents: e.g.10²⸱¹
instead of10^2.1
. -
thousands_separator::Char = ' '
:thousands separator character (useChar(0)
to disable grouping digits). -
projection::Symbol =:orthographic
:projection for 3D plots (:ortho(graphic)
,:persp(ective)
,orModel-View-Projection
(MVP) matrix). -
axes3d::Bool = true
:draw 3d axes (x ->:red
,y ->:green
,z ->:blue
). -
elevation::Float = 35.264389682754654
:elevation angle above or below thefloor
plane (-90 ≤ θ ≤ 90
). -
azimuth::Float = 45.0
:azimutal angle around theup
vector (-180° ≤ φ ≤ 180°
). -
zoom::Float = 1.0
:zooming factor in 3D. -
up::Symbol =:z
:up vector (:x
,:y
or:z
), prefix withm -> -
orp -> +
to change the sign e.g.:mz
for-z
axis pointing upwards. -
near::Float = 1.0
:distance to the near clipping plane (:perspective
projection only). -
far::Float = 100.0
:distance to the far clipping plane (:perspective
projection only). -
canvas_kw::NamedTuple = NamedTuple()
:extra canvas keywords. -
blend::Bool = true
:blend colors on the underlying canvas. -
fix_ar::Bool = false
:fix terminal aspect ratio (experimental). -
visible::Bool = true
:visible canvas.
Note:If you want to print the plot into a file but have monospace issues with your font, you should probably try settingborder=:ascii
andcanvas=AsciiCanvas
(orcanvas=DotCanvas
for scatterplots).
...
The primary structures that do all the heavy lifting behind the curtain are subtypes ofCanvas
.A canvas is a graphics object for rasterized plotting. Basically, it uses Unicode characters to represent pixel.
Here is a simple example:
importUnicodePlots:lines!, points!, pixel!
canvas=BrailleCanvas(15,40,#number of rows and columns (characters)
origin_y=0.,origin_x=0.,#position in virtual space
height=1.,width=1.)#size of the virtual space
lines!(canvas,0.,0.,1.,1.;color=:cyan)#virtual space
points!(canvas,rand(50),rand(50); color=:red)#virtual space
lines!(canvas,0.,1.,.5,0.;color=:yellow)#virtual space
pixel!(canvas,5,8;color=:red)#pixel space
Plot(canvas)
You can access the height and width of the canvas (in characters) withnrows(canvas)
andncols(canvas)
respectively. You can use those functions in combination withprint_row
to embed the canvas anywhere you wish. For example,print_row(STDOUT, canvas, 3)
writes the third character row of the canvas to the standard output.
As you can see, one issue that arises when multiple pixel are represented by one character is that it is hard to assign color. That is because each of the "pixel" of a character could belong to a different color group (each character can only have a single color). This package deals with this using a color-blend for the whole group. You can disable canvas color blending / mi xing by passingblend=false
to any function.
importUnicodePlots:lines!
canvas=BrailleCanvas(15,40;origin_y=0.,origin_x=0.,height=1.,width=1.)
lines!(canvas,0.,0.,1.,1.;color=:cyan)
lines!(canvas,.25,1.,.5,0.;color=:yellow)
lines!(canvas,.2,.8,1.,0.;color=:red)
Plot(canvas)
The following types ofCanvas
are implemented:
-
BrailleCanvas: This type of canvas is probably the one with the highest resolution for
Unicode
plotting. It essentially uses the Unicode characters of theBraillesymbols as pixels. This effectively turns every character into eight pixels that can individually be manipulated using binary operations. -
BlockCanvas: This canvas is also
Unicode
based. It has half the resolution of the BrailleCanvas. In contrast toBrailleCanvas
,the pixels don't have visible spacing between them. This canvas effectively turns every character into four pixels that can individually be manipulated using binary operations. -
HeatmapCanvas: This canvas is also
Unicode
based. It has half the resolution of theBlockCanvas
.This canvas effectively turns every character into two color pixels, using the foreground and background terminal colors. As such, the number of rows of the canvas is half the number ofy
coordinates being displayed. -
AsciiCanvasandDotCanvas: These two canvas utilizes only standard
ASCII
character for drawing. Naturally, it doesn't look quite as nice as the Unicode-based ones. However, in some situations it might yield better results. Printing plots to a file is one of those situations. -
DensityCanvas: Unlike the
BrailleCanvas
,the density canvas does not simply mark a "pixel" as set. Instead it increments a counter per character that keeps track of the frequency of pixels drawn in that character. Together with a variable that keeps track of the maximum frequency, the canvas can thus draw the density of data-points. -
BarplotGraphics: This graphics area is special in that it does not support any pixel manipulation. It is essentially the barplot without decorations but the numbers. It does only support one method
addrow!
which allows the user to add additional bars to the graphics object.
...
Because Julia uses column-major inde xing order for an array type, and because displaying data on a terminal is row based, we need an internal buffer compatible with efficient columns based iteration. We solve this by using the transpose of a (width
,height
) array for inde xing into an internal buffer likebuf[row, col]
orbuf[y, x]
.
Common users of UnicodePlots don't need to be aware of this axis difference if sticking to public interface.
p=Plot([NaN], [NaN]; xlim=(1,10), ylim=(1,10), title="internal buffer conventions")
#plot axes
vline!(p,1,head_tail=:head,color=:green,name="y-axis (rows)")
hline!(p,1,head_tail=:head,color=:red,name="x-axis (cols)")
#square
vline!(p,2,[2,9], color=:cyan,name="buf[y, x] - buf[row, col]")
vline!(p, [2,9], [2,9], color=:cyan)
hline!(p, [2,9], [2,9], color=:cyan)
#internal axes
vline!(p,3,range(3,8;length=20), head_tail=:tail,color=:light_green,name="y-buffer (rows)")
hline!(p,8,range(3,8;length=20), head_tail=:head,color=:light_red,name="x-buffer (cols)")
#mem layout
vline!(p,4,[4,7]; color=:yellow,name="memory layout")
vline!(p,7,[4,7]; color=:yellow)
hline!(p, [4,7], [4,7]; color=:yellow)
hline!(p, [4.5,5,5.5,6], [4.5,6.5]; color=:yellow)
...
Run the following snippet to analyze invalidations:
usingSnoopCompileCore
invalidations=@snooprusingUnicodePlots
tinf=@snoopi_deepUnicodePlots.precompile_workload()
usingSnoopCompile, AbstractTrees, PrettyTables#must occur after `invalidations`
print_tree(tinf; maxdepth=typemax(Int))
trees=invalidation_trees(invalidations)
trees=filtermod(UnicodePlots, trees; recursive=true)
@showlength(uinvalidated(invalidations))#all invalidations
#only from `UnicodePlots`
@showlength(staleinstances(tinf))
@showlength(trees)
SnoopCompile.report_invalidations(;
invalidations,
process_filename=x->last(split(x,".julia/packages/")),
n_rows=0,
)
...
The following snippet:
$cddocs
$ julia gen_docs.jl
$ (cd imgs;julia gen_imgs.jl)
will regenerateREADME.md
and the example images with root (prefix) urlhttps://github /JuliaPlots/UnicodePlots.jl/raw/unicodeplots-docs.
This code is free to use under the terms of the MIT license.
Inspired byTextPlots.jl,which in turn was inspired byDrawille.