Create publication-quality plots with a simple interface over matplotlib. Are you bored of copying and pasting the code to make a plot every time? Try this!

This module provides only one (highly customizable) function to plot some data. It uses matplotlib in its internal, but helps in setting all graphical and plot paramenters (tick rotate, font, size, labels, etc.).

For information about this Readme file and this tool please write to [email protected]

• Installation
• Usage
• Modes
• Arguments
• Cyclers
• Examples

Simply clone this repo or use:

pip3 install fastplot

Dependencies are: matplotlib numpy pandas statsmodels. FastPlot requires updated versions of such libraries, so, in case of error try first to upgrade them. For serif fonts you need Times New Roman, that, on Ubuntu, can be installed with:

sudo apt-get install msttcorefonts

This module has only one function, called plot. You must call it to make a plot, providing the data, the output file name, and other parameters.

• If you provide a path, the figure will be saved on disk, and no value returned by the function.
• If path is None, the function returns the current plt object for further processing or interactive show().
• Note: if you want to use the show() method of matplotlib for interactive view, you must import pyplot before importing fastplot, like:

import matplotlib.pyplot as plt
import fastplot

In this way, you can run fastplot also inside a Jupyter Notebook:

Note: If you want to both show() and savefig(), please do savefig() before, to prevent matplotlib from clearing the figure. See the following example.

Note: You may need to add the IPython magic %matplotlib inline to be able to visualize inline plots.

The most important arguments are:

• data: the actual data to plot
• path: the output path for the plot. The format is automatically inferred by matplotlib, looking at the extension of the path. Put it to None to have the current plt object returned and no file written.
• mode: which type of plot to create (lines, bars, etc.). More details later.
• style: which graphical style to use. Can be:
  • sans-serif: classical sans-serif, Arial-like font.
  • serif: use Times New Roman. Good for IEEE papers. You must have Times New Roman installed.
  • latex: tells matplotlib to use latex to render text. You must have latex installed. FastPlot has a utlity function called fastplot.tex_escape() to easily escape Tex strings.
  • Note: if you want to use the Linux Libertine font (e.g., used in ACM papers), you can use the latex style and pass the argument rcParams={'text.latex.preamble'}: r'\usepackage{libertine}'} to the fastplot.plot() function.
  • Note 2: if you want to use Times New Roman with Latex (very good for IEEE papers), you can use the latex style and pass the argument rcParams={'text.latex.preamble': r'\usepackage{mathptmx}'} to the fastplot.plot() function.
  • Note 3: if you want to use Palladio with Latex (some Latex classes use it), you can use the latex style and pass the argument rcParams={'text.latex.preamble': r'\usepackage{mathpazo}'} to the fastplot.plot() function.

The modes are the type of plots fastplot allows to use. Some are simple (just a line), other are more advanced (bars, etc.).

• line: plot a simple line. data must be a two-sized tuple of lists, for x and y. E.g., ([x1,x2,x3],[y1,y2,y3])
• line_multi: plot multiple lines. Data must have the form [ (line1, ([x1,x2], [y1,y2])), (line2, ([x1,x2], [y1,y2]) ) ]. The names line1 and line2 are put in the legend.
• CDF: plot a CDF given the samples. data must be a list of scalars. Note: if your CDF is too coarse grained, you can increase the resolution increasing fastplot.NUM_BIN_CDF.
• CDF_multi: plot many CDFs given the samples. data must be a list of two-sized tuples like (name, [samples]). name is used in the legend.
• boxplot: plot a boxplot given the samples. data must be a list of two-sized tuples like (name, [samples]). name is used in the xticks labels.
• boxplot_multi: plot a boxplot given the samples, clustered in groups. data a pandas dataframe, where each cell is a list. A groups are defined by each row, elements of each groups by columns.
• timeseries: plot a time series. data must be a pandas series, with a DateTime index.
• timeseries_multi: plot many time series. data must be a list of two-sized tuples like (name, timeseries). name is used in the legend.
• timeseries_stacked: plot many time series, stacked. data must be a pandas dataframe, with a DateTime index. Each column will be plotted stacked to the others. Column names are used in the legend.
• bars: plot a bar plot. data must be a list of (name, value). name is used for the legend.
• bars_multi: plot grouped bars. data must be a padas dataframe. Each row is results in a group of bars, while columns determine bars within each group.
• bars_stacked: plot stacked bars. data must be a padas dataframe. Rows go on x axis, while each column is a level in the bars.
• callback: call a user function instead of plotting data. You must provide a function pointer in the callback argument, that will be called passing plt as paramenter in order to perform a user defined plot. No matter what you put in data.

Note: you can use callback mode to draw data with seaborn and profit from all the features of fastplot. See the examples.

fastplot provides utility functions to compute Lorenz curve and Gini Index.

To compute Gini index and Lorenz curve for a single set of samples call the fastplot.lorenz_gini() function, whose sole arguments is a set of samples. It returns (lorenz_x, lorenz_y), gini_index, that you can plot with the mode line.

To compute Gini index and Lorenz curve for multiple set of samples, call fastplot.lorenz_gini_multi(). It takes as input a list of two-sized tuples like (name, [samples]). It provides as output like: [ (name, ([x1,x2], [y1,y2])), (name, ([x1,x2], [y1,y2]) ) ]. The optional argument name_format="{} (GI={:0.2f})" is string format to transform names to include the value of the Gini index. The output of this function can be directly used as input to fastplot with mode line_multi

Arguments of the plot function are divided in many categories. Only core are mandatory.

Core

• data: the input data to plot
• path: the output path for the plot. The format is automatically inferred by matplotlib, looking at the extension of the path. Put it to None to have the current plt object returned and no file written.
• mode: which type of plot to create (lines, bars, etc.). More details later. Default line
• plot_args: an optional dictionary of arguments to pass the matplotlib plot() function. E.g., use plot_args={"markersize":0.5} to reduce the marker/point size.

Look

• style: which graphical style to use. Can be serif, sans-serif or latex. For latex, it enables matplotlib latex engine. Default sans-serif
• figsize: the size of the output figure. Default: (4,2.25)
• cycler: the style cycler to use. By default, it changes across main colors, and different linestyles. Change it if you want to change color, line, or point style. I provide some useful cycler in the code (see Cycler section). Default CYCLER_LINES
• fontsize: just the overall font size. Default 11
• dpi: DPI for the output image. Default 300
• classic_autolimit: Use classic autolimit feature (find 'nice' round numbers as view limits that enclosed the data limits), instead of v2 (sets the view limits to 5% wider than the data range). Default True
• rcParams: optional dictionary of rcParams to set before plotting

Grid

• grid: whether to display grid. Default False
• grid_which: ticks for the grid (major/minor). Default major
• grid_axis: axis for the grid. Default both
• grid_linestyle: gird linestyle. Default dotted
• grid_color: grid color. Default black

Scale

• xscale: scale for x axis (linear/log). Default linear
• yscale: scale for y axis (linear/log). Default linear

Axis

• xlim: Optional x limit, as a tuple (low, high). You can set low or high to None if you don't want to modify it.
• ylim: Optional y limit, as a tuple (low, high). You can set low or high to None if you don't want to modify it.
• xlabel: Label for x axis
• ylabel: Label for y axis
• xticks: Custom x ticks, in the form ([x1, x2, ...], [label1, label2, ...]). You can pass ([x1, x2, ...], None) if you just want to set the xticks, without specifying the labels.
• yticks: Custom y ticks, in the form ([y1, y2, ...], [label1, label2, ...]) You can pass ([y1, y2, ...], None) if you just want to set the yticks, without specifying the labels.
• xticks_rotate: Rotation for x ticks. Default 0
• yticks_rotate: Rotation for y ticks. Default 0
• xticks_fontsize: X ticks font size. Default medium
• yticks_fontsize: Y ticks font size. Default medium
• xtick_direction: xtick marker direction. Default 'in'
• xtick_width: xticks marker width. Default 1
• xtick_length: xticks marker length. Default 4
• ytick_direction: ytick marker direction. Default 'in'
• ytick_width: yticks marker width. Default 1
• ytick_length: yticks marker length. Default 4

Legend

• legend: whether to show the legend. Default False
• legend_loc: location for legend. Default best
• legend_ncol: number of columns. Default 1
• legend_fontsize: legend fontsize. Default medium
• legend_border: whether to show legend border. Default False
• legend_frameon: whether to show legend frame. Default True
• legend_fancybox: whether to use round corner on legend frame. Default False
• legend_alpha: transparency of legend frame. Default 1.0
• legend_args: an optional dictionary of arguments to pass the matplotlib legend() function. For example you might want to manually set legend location by passing legend_args={'bbox_to_anchor' : (0.9, 0.1)}.

Specific This arguments are specific for some modes.

• linewidth: linewidth when lines are used. Default: 1
• boxplot_sym: symbol for outliers in boxplots. Default ''
• boxplot_whis: whisker spec for boxplot. Default [5,95]
• boxplot_numerousness: plot the number of samples on the top-x axis. Default False
• boxplot_numerousness_fontsize: size of sample groups labels. Default x-small
• boxplot_numerousness_rotate: rotation for numerousness text. Default 0
• boxplot_palette: palette to be used with seaborn. Default is seaborn default
• boxplot_empty: Wether to plot boxplot without filling it. Default is False
• timeseries_format: format for printing dates in timeseries. Default %Y/%m/%d
• bars_width: width of bars when bars are plotted. Default 0.6
• callback: function to call instead of plotting, when mode=callback
• timeseries_stacked_right_legend_order: for timeseries_stacked, plot legend in the same order as colors are shown in the plot. Default is True.
• CDF_complementary: for CDF and CDF_multi, plot complementary CDF instead of the plain one. Default is False.

FastPlot provides simple cyclers to obtain nice lines, points, and linespoints in both color and B/W.

The list of available cyclers is:

• fastplot.CYCLER_LINES: Lines, with colors
• fastplot.CYCLER_LINESPOINTS: Linespoints, with colors
• fastplot.CYCLER_POINTS: Points, with colors
• fastplot.CYCLER_LINES_BLACK: Black lines
• fastplot.CYCLER_LINESPOINTS_BLACK: Black linespoints
• fastplot.CYCLER_POINTS_BLACK: Black points

line

x = range(11)
y=[4,150,234,465,745,612,554,43,565,987,154]
fastplot.plot((x, y),  'examples/1_line.png', xlabel = 'X', ylabel = 'Y')

line_multi

x = range(11)
y1=[4,150,234,465,645,612,554,43,565,987,154]
y2=[434,15,24,556,75,345,54,443,56,97,854]
fastplot.plot([ ('First', (x, y1) ), ('Second', (x, y2) )], 'examples/2_line_multi.png',
              mode='line_multi', xlabel = 'X', ylabel = 'Y', xlim = (-0.5,10.5),
              cycler = fastplot.CYCLER_LINESPOINTS, legend=True, legend_loc='upper left',
              legend_ncol=2)

CDF

fastplot.plot(np.random.normal(100, 30, 1000), 'examples/3_CDF.png', mode='CDF',
              xlabel = 'Data', style='latex')

CDF complementary

fastplot.plot(np.random.normal(100, 30, 1000), 'examples/3b_CCDF.png', mode='CDF', 
              CDF_complementary=True, xlabel = 'Data', style='latex')

CDF_multi

data = [ ('A', np.random.normal(100, 30, 1000)), ('B', np.random.normal(140, 50, 1000)) ]
plot_args={"markevery": [500]}
fastplot.plot(data , 'examples/4_CDF_multi.png', mode='CDF_multi', xlabel = 'Data', legend=True,
              cycler = fastplot.CYCLER_LINESPOINTS, plot_args=plot_args)

boxplot

data=[ ('A', np.random.normal(100, 30, 450)),
       ('B', np.random.normal(140, 50, 50)),
       ('C', np.random.normal(140, 50, 200))]
fastplot.plot( data,  'examples/5_boxplot.png', mode='boxplot', ylabel = 'Value',
               boxplot_numerousness=True, boxplot_empty=True, boxplot_numerousness_rotate=90)

boxplot_multi

data = pd.DataFrame(data=[ [np.random.normal(100, 30, 50),np.random.normal(110, 30, 50)],
                           [np.random.normal(90, 30, 50),np.random.normal(90, 30, 50)],
                           [np.random.normal(90, 30, 50),np.random.normal(80, 30, 50)],
                           [np.random.normal(80, 30, 50),np.random.normal(80, 30, 50)]],
                    columns=["Male","Female"], index = ["IT", "FR", "DE", "UK"] )
fastplot.plot( data,  'examples/5b_boxplot_multi.png', mode='boxplot_multi', ylabel = 'Value',
               boxplot_palette="muted", legend=True, legend_ncol=2, ylim=(0,None))

timeseries

rng = pd.date_range('1/1/2011', periods=480, freq='H')
ts = pd.Series(np.random.randn(len(rng)), index=rng)
fastplot.plot(ts ,  'examples/6_timeseries.png', mode='timeseries', ylabel = 'Value',
              style='latex', xticks_rotate=30, xticks_fontsize='small',
              xlim=(pd.Timestamp('1/1/2011'), pd.Timestamp('1/7/2011')))

timeseries_multi

rng = pd.date_range('1/1/2011', periods=480, freq='H')
ts = pd.Series(np.random.randn(len(rng)), index=rng) + 5
ts2 = pd.Series(np.random.randn(len(rng)), index=rng) + 10
fastplot.plot( [('One', ts), ('Two', ts2)] , 'examples/7_timeseries_multi.png',
               mode='timeseries_multi', ylabel = 'Value', xticks_rotate=30,
               legend = True, legend_loc='upper center', legend_ncol=2, legend_frameon=False,
               ylim = (0,None), xticks_fontsize='small')

timeseries_stacked

rng = pd.date_range('1/1/2011', periods=480, freq='H')
df = pd.DataFrame(np.random.uniform(3,4,size=(len(rng),2)), index=rng, columns=('One','Two'))
df = df.divide(df.sum(axis=1), axis=0)*100
fastplot.plot( df , 'examples/8_timeseries_stacked.png', mode='timeseries_stacked',
               ylabel = 'Value [%]', xticks_rotate=30, ylim=(0,100), legend=True,
               xlim=(pd.Timestamp('1/1/2011'), pd.Timestamp('1/7/2011')))

bars

data = [('First',3),('Second',2),('Third',7),('Four',6),('Five',5),('Six',4)]
fastplot.plot(data,  'examples/9_bars.png', mode = 'bars', ylabel = 'Value',
              xticks_rotate=30, style='serif', ylim = (0,10))

bars_multi

data = pd.DataFrame( [[2,5,9], [3,5,7], [1,6,9], [3,6,8], [2,6,8]],
                     index = ['One', 'Two', 'Three', 'Four', 'Five'],
                     columns = ['A', 'B', 'C'] )
fastplot.plot(data,  'examples/10_bars_multi.png', mode = 'bars_multi', style='latex',
              ylabel = 'Value', legend = True, ylim = (0,12), legend_ncol=3,
              legend_args={'markerfirst' : False})

bars_stacked

data = pd.DataFrame( [[2,5,9], [3,5,7], [1,6,9], [3,6,3], [2,6,2]],
                     index = ['One', 'Two', 'Three', 'Four', 'Five'],
                     columns = ['A', 'B', 'C'] )
fastplot.plot(data,  'examples/12_bars_stacked.png', mode = 'bars_stacked', style='serif',
              ylabel = 'Value', legend = True, xtick_length=0, legend_ncol=3, ylim = (0,25))

callback

x = range(11)
y=[120,150,234,465,745,612,554,234,565,888,154]
def my_callback(plt):
    plt.bar(x,y)
fastplot.plot(None,  'examples/11_callback.png', mode = 'callback', callback = my_callback,
              style='latex', xlim=(-0.5, 11.5), ylim=(0, 1000))

Lorenz Curves

data = [ ('A', np.random.chisquare(2, 1000)), ('B', np.random.chisquare(8, 1000)) ]
data = fastplot.lorenz_gini_multi(data)
fastplot.plot(data, 'examples/13_lorenz.png', mode='line_multi', legend=True, grid=True,
              xlabel = 'Samples [%]', ylabel = 'Share [%]', xlim=(0,1), ylim=(0,1))

seaborn

data = pd.DataFrame([(4,3),(5,4),(4,5),(8,6),(10,8),(3,1),(13,10),(9,7),(11,11)], columns=["x","y"])
def my_callback(plt):
     sns.regplot(x="x", y="y", data=data, ax=plt.gca())
fastplot.plot(None,  'examples/14_seaborn.png', mode = 'callback', callback = my_callback,
              style='latex', grid=True)