[Tex/LaTex] Waterfall chart

pgfplots

I am trying to generate the following figure more automatically based on this example. Currently, the nodes above and below the bars are generated manually, which I like to avoid. I would prefer to add the text for the upper labels in the datatable and calculate their positions. In the same way, I would like to change the lower values that state the current total.

Also I would like to be able to add another bar in between 4 and 7 that contains the current total (to add emphasis to the current sum like best case, expected case, worst case)

Bonus Question

I tried to use the pgfmathabs function to have only non-negative values for node near coord. This is does not work, therefore I am first multiplying and then taking the square root, which seems rather inappropriate.

In order to better explain the desired outcome, I added a picture. I am not sure if the number belows the "step" bars (e.g. "oh no") are needed, if a "full" bar (e.g. "expected case") follows. If another "step" bars follows, they should be there. So if it would be possible to omit the lower numbers, in front of full bars that would be great. (In this minimal example, this does not really make sense, since all number would be ommited but in my full example with 10 "step" bars and 3 "full" bars its a different story.)

MWE

\documentclass[border=5mm, 10pt]{standalone}
\usepackage{pgfplots, pgfplotstable}
\usepackage{filecontents}

\pgfplotsset{compat=1.8}

\begin{filecontents}{datatable.csv}
16 
 -4  
 -7  
5
\end{filecontents}

\newcommand\barwidthval{22.5pt}%bar width value

\pgfplotstableset{
    create on use/accumyprev/.style={
        create col/expr={\prevrow{0}+\pgfmathaccuma}
    }
}

\makeatletter
\pgfplotsset{
    centered nodes near coords/.style={
    calculate offset/.code={
        \pgfkeys{/pgf/fpu=true,/pgf/fpu/output format=fixed}
        \pgfmathsetmacro\testmacro{(\pgfplotspointmeta*10^\pgfplots@data@scale@trafo@EXPONENT@y)/2*\pgfplots@y@veclength)}
        \pgfkeys{/pgf/fpu=false}
    },
    every node near coord/.style={
        /pgfplots/calculate offset,
        yshift=-\testmacro,
        black,
    },
    nodes near coords align=center
    },  
}

\begin{document}

\begin{tikzpicture}
\begin{axis}[
    no markers,
    axis x line=center,
    ybar stacked,
    ymin=-20,
    point meta=explicit,
    centered nodes near coords, 
    axis lines*=left,
    xtick=data,
    enlarge x limits=0.05,
    enlarge y limits=0.20,
    major tick length=0pt,
    bar width=\barwidthval,
    xticklabels={},
    xticklabel style={text width=2cm, align=center},
    ytick=\empty,
    x axis line style={opacity=0},
    y axis line style={opacity=0},
    ylabel={Values},
    axis on top
]

% 0-x-Axis
\draw ({rel axis cs:0,0}|-{axis cs:0,0}) -- ({rel axis cs:1,0}|-{axis cs:0,0});

% The first plot sets the "baseline": Uses the sum of all previous y values, except for the last bar, where it becomes 0
\addplot +[
    y filter/.code={\ifnum\coordindex>2 \def\pgfmathresult{0}\fi},
    draw=none,
    fill=none
] table [x expr=\coordindex, y=accumyprev] {datatable.csv};

% The values
\addplot +[
    fill=orange,
    draw=orange,
    ybar stacked,
    nodes near coords,
    nodes near coords={
    \pgfkeys{
      /pgf/fpu=true,  % PGFplots uses floating points internally
     }
    % pgfmathabs does not work here   
    \pgfmathparse{sqrt(\pgfplotspointmeta*\pgfplotspointmeta)}\pgfmathprintnumber{\pgfmathresult}
    },
] table [x expr=\coordindex, y index=0, meta index=0] {datatable.csv};

% The connecting line. Uses a bit of magic to typeset the ranges
\addplot [const plot, black] table [x expr=\coordindex, y expr=0] {datatable.csv};

\tikzstyle{upper node}=[above, font=\tiny, text width=\barwidthval, align=center, inner xsep=0, inner ysep=1pt,]
\tikzstyle{lower node}=[below,]

\node [upper node] at (axis cs: 0,16) {Best\\ case};
\node [upper node] at (axis cs: 1,16) {Oh no};
\node [lower node] at (axis cs: 1,12) {12};
\node [upper node] at (axis cs: 2,12) {Not again};
\node [lower node] at (axis cs: 2,5) {5};
\node [upper node] at (axis cs: 3,5) {Worst\\ case};

\end{axis}
\end{tikzpicture}
\end{document}

Best Answer

This is a version without pgfplots, but rather with TikZ.

Code

\documentclass[tikz, border=2mm]{standalone}
\usetikzlibrary{calc}
\usepackage{xifthen}

\begin{document}


\newcommand{\Waterfall}[7]%
% bar width,
% bar sep,
% bar color,
%
% max height,
% total
% description,
%
% parts/descriptions
%
{   \colorlet{BarColor}{#3}
    \pgfmathsetmacro{\SF}{#4/#5}% scale factor
    \fill[BarColor] (0,0) rectangle node[black] {#5} ++(#1,#4) coordinate (temp);
    \node[above] at ($(temp)+(-#1/2,0)$) {\tiny#6};
    \xdef\Rest{#5}
    \foreach \drop/\description in {#7}
    {   \pgfmathtruncatemacro{\NewRest}{\Rest-\drop}
        \draw (temp) -- ++ (#2,0) coordinate (temp);
        \node[above] at ($(temp)+(#1/2,0)$) {\tiny\description};
        \fill[BarColor!90] (temp) rectangle node[black] {\drop} ++(#1,-\SF*\drop) coordinate (temp);
        \ifthenelse{\NewRest = 0}
            {}
            {   %\node[below] at ($(temp)+(-#1/2,0)$) {\NewRest};
                \fill[BarColor!90!black] (temp) rectangle node[black] {\NewRest} ++ (-#1,-\SF*\NewRest);
            }
        \xdef\Rest{\NewRest}
    }
    \draw (temp) -- (0,0);
}

\begin{tikzpicture}
    \Waterfall{0.7}{1.5}{orange!50}%
        {5}{16}{Best case}%
        {4/Oh no,7/Not again,5/Worst case}%
\end{tikzpicture}

\end{document}

Output

Edit 1: Now with itermittent rest bars and conditionaly skipped labels.

Code

\documentclass[tikz, border=2mm]{standalone}
\usetikzlibrary{calc}
\usepackage{xifthen}

\begin{document}

\newcommand{\Waterfall}[7]%
% bar width,
% bar sep,
% bar color,
% max height,
%
% total
% description,
%
% parts/descriptions/rest label
%
{   \colorlet{BarColor}{#3}
    \pgfmathsetmacro{\SF}{#4/#5}% scale factor
    \fill[BarColor] (0,0) rectangle node[black] {#5} ++(#1,#4) coordinate (temp);
    \node[above] at ($(temp)+(-#1/2,0)$) {\tiny#6};
    \xdef\Rest{#5}
    \foreach \drop/\description/\restlabel in {#7}
    {   \pgfmathtruncatemacro{\NewRest}{\Rest-\drop}
        \draw (temp) -- ++ (#2,0) coordinate (temp);
        \node[above] at ($(temp)+(#1/2,0)$) {\tiny\description};
        \ifthenelse{\NewRest = 0}
        {   \fill[BarColor] (temp) rectangle node[black] {\drop} ++(#1,-\SF*\drop) coordinate (temp);
        }
        {   \fill[BarColor!50] (temp) rectangle node[black] {\drop} ++(#1,-\SF*\drop) coordinate (temp);
        }
        \ifthenelse{\equal{\restlabel}{}}
        {   \ifthenelse{\NewRest = 0}
            {}
            {   \node[below] at ($(temp)+(-#1/2,0)$) {\NewRest};
                %\fill[BarColor!90!black] (temp) rectangle node[black] {\NewRest} ++ (-#1,-\SF*\NewRest);
            }
        }
        {   \draw (temp) -- ++ (#2,0) ++ (#1,0) coordinate (temp);
            \fill[BarColor] (temp) rectangle node[black] {\NewRest} ++ (-#1,-\SF*\NewRest);
            \node[above] at ($(temp)+(-#1/2,0)$) {\tiny\restlabel};
        }       
        \xdef\Rest{\NewRest}
    }
    \draw (temp) -- (0,0);
}

\begin{tikzpicture}
    \Waterfall{0.7}{1.5}{orange!50}{5}%
        {16}{Best case}%
        {4/Oh no/Expected case,7/Not again/,5/Worst case/}
\end{tikzpicture}

\begin{tikzpicture}
    \Waterfall{0.5}{0.5}{cyan}{8}%
        {21}{good}%
        {6/Oh no/,5/Oh no/worse,4/Oh no/worse,3/Oh no/,2/Oh no/,1/worst/}
\end{tikzpicture}

\end{document}

Output

Edit 2: A workaround to remove the label before the last fall is to use IGNORE as the second to last rest label:

Code

\documentclass[tikz, border=2mm]{standalone}
\usetikzlibrary{calc}
\usepackage{xifthen}

\begin{document}

\newcommand{\Waterfall}[7]%
% bar width,
% bar sep,
% bar color,
% max height,
%
% total
% description,
%
% parts/descriptions/rest label
%
{   \colorlet{BarColor}{#3}
    \pgfmathsetmacro{\SF}{#4/#5}% scale factor
    \fill[BarColor] (0,0) rectangle node[black] {#5} ++(#1,#4) coordinate (temp);
    \node[above] at ($(temp)+(-#1/2,0)$) {\tiny#6};
    \xdef\Rest{#5}
    \foreach \drop/\description/\restlabel in {#7}
    {   \pgfmathtruncatemacro{\NewRest}{\Rest-\drop}
        \draw (temp) -- ++ (#2,0) coordinate (temp);
        \node[above] at ($(temp)+(#1/2,0)$) {\tiny\description};
        \ifthenelse{\NewRest = 0}
        {   \fill[BarColor] (temp) rectangle node[black] {\drop} ++(#1,-\SF*\drop) coordinate (temp);
        }
        {   \fill[BarColor!50] (temp) rectangle node[black] {\drop} ++(#1,-\SF*\drop) coordinate (temp);
        }
        \ifthenelse{\equal{\restlabel}{}}
        {   \ifthenelse{\NewRest = 0}
            {}
            {   \node[below] at ($(temp)+(-#1/2,0)$) {\NewRest};
                %\fill[BarColor!90!black] (temp) rectangle node[black] {\NewRest} ++ (-#1,-\SF*\NewRest);
            }
        }
        {   \ifthenelse{\equal{\restlabel}{IGNORE}}
            {}
            {   \draw (temp) -- ++ (#2,0) ++ (#1,0) coordinate (temp);
                \fill[BarColor] (temp) rectangle node[black] {\NewRest} ++ (-#1,-\SF*\NewRest);
                \node[above] at ($(temp)+(-#1/2,0)$) {\tiny\restlabel};
            }
        }       
        \xdef\Rest{\NewRest}
    }
    \draw (temp) -- (0,0);
}

\begin{tikzpicture}
    \Waterfall{0.7}{1.5}{orange!50}{5}%
        {16}{Best case}%
        {4/Oh no/Expected case,7/Not again/IGNORE,5/Worst case/}
\end{tikzpicture}

\begin{tikzpicture}
    \Waterfall{0.5}{0.5}{cyan}{8}%
        {45}{good}%
        {10/Oh no/,9/Oh no/worse,8/Oh no/worse,7/Oh no/,6/Oh no/IGNORE,5/worst/}
\end{tikzpicture}

\end{document}

Output

Edit 2: Now with pgfkeys to supply key-value options, which makes customization easier:

Code

\documentclass[tikz, border=2mm]{standalone}
\usetikzlibrary{calc}
\usepackage{xifthen}

\tikzset{
    waterfall/.is family,
    waterfall,
    bar width/.initial=0.7,
    bar sep/.initial=0.7,
    total height/.initial=5,
    bar color/.initial=blue!50!cyan,
    drop color/.initial=blue!50!cyan!50,
    draw color/.initial=transparent,
    label options/.style={font=\tiny},
    bar label options/.style={text=black},
    rest label options/.style={text=black},
    total/.initial=20,
    total label/.initial=Total,
}

\newcommand{\WFKey}[1] % access a specific key by name
{\pgfkeysvalueof{/tikz/waterfall/#1}}

\newcommand{\Waterfall}[2][]%
% [options], parts/descriptions/rest label
{   \tikzset{waterfall,#1} % Process Keys passed to command
    \pgfmathsetmacro{\SF}{\WFKey{total height}/\WFKey{total}}% scale factor
    \fill[\WFKey{bar color}] (0,0) rectangle node[waterfall/bar label options] {\WFKey{total}} ++(\WFKey{bar width},\WFKey{total height}) coordinate (temp);
    \node[above, waterfall/label options] at ($(temp)+(-\WFKey{bar width}/2,0)$) {\WFKey{total label}};
    \xdef\Rest{\WFKey{total}}
    \foreach \drop/\description/\restlabel in {#2}
    {   \pgfmathtruncatemacro{\NewRest}{\Rest-\drop}
        \draw (temp) -- ++ (\WFKey{bar sep},0) coordinate (temp);
        \node[above, waterfall/label options] at ($(temp)+(\WFKey{bar width}/2,0)$) {\description};
        \ifthenelse{\NewRest = 0}
        {   \fill[\WFKey{bar color}] (temp) rectangle node[waterfall/bar label options] {\drop} ++(\WFKey{bar width},-\SF*\drop) coordinate (temp);
        }
        {   \fill[\WFKey{drop color}] (temp) rectangle node[waterfall/bar label options] {\drop} ++(\WFKey{bar width},-\SF*\drop) coordinate (temp);
        }
        \ifthenelse{\equal{\restlabel}{}}
        {   \ifthenelse{\NewRest = 0}
            {}
            {   \node[below, waterfall/rest label options] at ($(temp)+(-\WFKey{bar width}/2,0)$) {\NewRest};
                %\fill[BarColor!90!black] (temp) rectangle node[black] {\NewRest} ++ (-#1,-\SF*\NewRest);
            }
        }
        {   \ifthenelse{\equal{\restlabel}{IGNORE}}
            {}
            {   \draw (temp) -- ++ (\WFKey{bar sep},0) ++ (\WFKey{bar width},0) coordinate (temp);
                \fill[\WFKey{bar color}] (temp) rectangle node[waterfall/bar label options] {\NewRest} ++ (-\WFKey{bar width},-\SF*\NewRest);
                \node[above, waterfall/label options] at ($(temp)+(-\WFKey{bar width}/2,0)$) {\restlabel};
            }
        }       
        \xdef\Rest{\NewRest}
    }
    \draw (temp) -- (0,0);
}

\begin{document}

\begin{tikzpicture}
    \Waterfall[total=16, total label=Best case, bar sep=1.4]%
        {4/Oh no/Expected case,7/Not again/IGNORE,5/Worst case/}
\end{tikzpicture}

\begin{tikzpicture}
    \Waterfall%
    [   total=45,
        total label=Good,
        bar color=green!50!gray,
        drop color=gray,
        label options/.style={font=\tiny\sffamily, draw=red, rounded corners=2pt, text=black, inner sep=2pt, above=1mm, minimum height=4mm},
        bar label options/.style={font=\tiny\sffamily, circle, fill= white, text=black, inner sep=2pt},
        rest label options/.style={font=\tiny\sffamily, circle, draw=gray, fill=gray!50, text=black, inner sep=2pt, below=-2mm},
    ]%
        {10/Oh no/,9/Oh no/bad,8/Oh no/worse,7/Oh no/,6/Oh no/IGNORE,5/worst/}
\end{tikzpicture}

\end{document}

Output

Related Solutions

[Tex/LaTex] plotting two time series with bounds

This happens because PGFPlots only uses one "stack" per axis: You're stacking the second confidence interval on top of the first. The easiest way to fix this is probably to use the approach described in "Is there an easy way of using line thickness as error indicator in a plot?": After plotting the first confidence interval, stack the upper bound on top again, using stack dir=minus. That way, the stack will be reset to zero, and you can draw the second confidence interval in the same fashion as the first:

\documentclass{standalone}
\usepackage{pgfplots, tikz}

\usepackage{pgfplotstable}

\pgfplotstableread{
temps   y_h y_h__inf    y_h__sup    y_f y_f__inf    y_f__sup    

1   0.237340    0.135170    0.339511    0.237653    0.135482    0.339823    
2   0.561320    0.422007    0.700633    0.165871    0.026558    0.305184    
3   0.694760    0.534205    0.855314    0.074856    -0.085698   0.235411    
4   0.728306    0.560179    0.896432    0.003361    -0.164765   0.171487    
5   0.711710    0.544944    0.878477    -0.044582   -0.211349   0.122184    
6   0.671241    0.511191    0.831291    -0.073347   -0.233397   0.086703    
7   0.621177    0.471219    0.771135    -0.088418   -0.238376   0.061540    
8   0.569354    0.431826    0.706882    -0.094382   -0.231910   0.043146    
9   0.519973    0.396571    0.643376    -0.094619   -0.218022   0.028783    
10  0.475121    0.366990    0.583251    -0.091467   -0.199598   0.016664    
}{\table}

\begin{document}
\begin{tikzpicture}

    \begin{axis}
    % y_h confidence interval
    \addplot [stack plots=y, fill=none, draw=none, forget plot]   table [x=temps, y=y_h__inf]   {\table} \closedcycle;
    \addplot [stack plots=y, fill=gray!50, opacity=0.4, draw opacity=0, area legend]   table [x=temps, y expr=\thisrow{y_h__sup}-\thisrow{y_h__inf}]   {\table} \closedcycle;
    % subtract the upper bound so our stack is back at zero
    \addplot [stack plots=y, stack dir=minus, forget plot, draw=none] table [x=temps, y=y_h__sup] {\table};

    % y_f confidence interval
    \addplot [stack plots=y, fill=none, draw=none, forget plot]   table [x=temps, y=y_f__inf]   {\table} \closedcycle;
    \addplot [stack plots=y, fill=gray!50, opacity=0.4, draw opacity=0, area legend]   table [x=temps, y expr=\thisrow{y_f__sup}-\thisrow{y_f__inf}]   {\table} \closedcycle;

    % the line plots (y_h and y_f)    
    \addplot [stack plots=false, very thick,smooth,blue]  table [x=temps, y=y_h]   {\table};
    \addplot [stack plots=false, very thick,smooth,blue]  table [x=temps, y=y_f]   {\table};
    \end{axis}

\end{tikzpicture}
\end{document}

[Tex/LaTex] Horizontal waterfall chart with labels for each bar

Well, just for fun (and to better learn new tools) I did an implementation in lua, which can be run through LuaLatex.

The lua code reads from an external .cvs, in which data is expected to be in different lines (a number per line), and generates a tikz graphic which self-adapts its axis to the data read. Also, the generated tikz defines a set of coordinates named row1, row2 and so on located a the center of each bar, and a coordinate named min located at the x position of the leftmost bar. These coordinates can be used to put labels in the diagram, either centered on the bars or at the left of the graphic.

The colors used in the graphic are user-definable. If there are less colors than bars, they are cycled.

This is the LaTeX code:

\documentclass{book}
\usepackage{xcolor}
\usepackage{tikz}
\usepackage{filecontents}
\directlua{dofile("luaFunctions.lua")}

%create a pair of datafiles
\begin{filecontents*}{datafile1.csv}
  2
  3
  2
  -6
  -3
\end{filecontents*}
\begin{filecontents*}{datafile2.csv}
  3 
  4 
  2 
  3 
  -4   
  -6
  2 
  -5
\end{filecontents*}

% latex commands to execute the lua functions
\def\waterfallChart#1{\directlua{waterfallChart("#1")}}
\def\setColors#1{%
   \directlua{emptyColors()}%
   \foreach \c in {#1} {\directlua{addColor("\c")}}
}

% set some styles
\tikzset{bar connection/.style = {black!50, thick}}
\setColors{cyan!80!black!50, orange, blue!80!black, red, yellow}

\begin{document}
\begin{tikzpicture} % First graph
  \waterfallChart{datafile1.csv}  % This draws the chart
  % Now, adding labels, centered at each bar
  \foreach \label [count=\n from 1] in {foo, bar, bad, foobar, spam}
    \node at (row\n) {\label};
\end{tikzpicture}

\vskip 2cm

\begin{tikzpicture} % Second graph
  \setColors{brown,red,orange}   % Different colors for this one
  \waterfallChart{datafile2.csv} % Draw the chart
  % Put labels (at the left of the figure in this case)
  \foreach \label [count=\n from 1] in {foo, bar, bad, foobar, spam, eggs, lorem, ipsum}
    \node[left] at (row\n-|min) {\label};
\end{tikzpicture}
\end{document}

This is the result:

Result

And this is the content of the file luaFunctions.lua:

colors = {"blue","green"}  -- Default colors

function readDataFile(filename)
    local input = io.open(filename, 'r')
    local dataTable = {}
    local n
    for line in input:lines() do
       table.insert(dataTable, line)
    end
    input:close()
    return dataTable
end

function emptyColors()
    colors = {}
end

function addColor(c)
    table.insert(colors,c)
end

function computeExtremes(dataTable)
    local max, min, x
    x = 0.0
    min = 0.0
    max = 0.0
    for i,p in ipairs(dataTable) do
        x = x + p
        if (x<min) then min = x end
        if (x>max) then max = x end
    end
    return min, max
end

function waterfallChart(filename)
    local data = readDataFile(filename)
    local min, max, n_steps, step, color, xpos, ypos, barwidth, aux, spread

    -- Configure here as required
    barwidth = 0.5  -- Height of each bar in the chart
    spread = 1.4    -- Distance among baselines of the bars (in barwidth units)
    n_ticks = 6     -- Number of ticks in the x-axis

    min, max = computeExtremes(data)
    step = (max-min)/n_ticks
    max = min + n_ticks*step
    xpos = 0.0
    ypos = 0.0
    aux = 0
    -- Draw axes
    -- Vertical axis
    tex.print(string.format("\\draw (0,%f) -- (0, %f);",
               1.1*barwidth, -#data*spread*barwidth))
    -- Horizontal axis
    tex.print(string.format("\\foreach \\tick in {%d, %d, ..., %d}",
              min, min+step, max))
    tex.print(string.format("  \\draw (\\tick, %f) -- +(0, -2mm) node[below] {\\tick};",
              -#data*spread*barwidth))
    tex.print(string.format("\\coordinate (min) at (%f,%f);",
                  min+0.0, -#data*spread*barwidth))

    -- Draw the bars
    color = 1
    for i,p in ipairs(data) do
        tex.print(string.format("\\fill[%s] (%f,%f) rectangle +(%f, %f) coordinate[midway] (row%d);",
                  colors[color], xpos, ypos, p+0.0, barwidth, i))
        tex.print(string.format("\\draw[bar connection] (%f, %f) -- +(0,%f);",
                  xpos, ypos, spread*aux*barwidth))
        aux =   1
        ypos = ypos - spread*barwidth
        xpos = xpos + 1.0 * p
        color = color + 1
        if (color > #colors) then color = 1; end
    end
end

MWE

Best Answer

Code

Output

Code

Output

Code

Output

Code

Output

Related Solutions

[Tex/LaTex] plotting two time series with bounds

[Tex/LaTex] Horizontal waterfall chart with labels for each bar

Related Question