Creating publication-quality figures with Python

Supported figure types: bar, stacked, scatter, line, subplot, func

Xianwei Zhang / https://people.cs.pitt.edu/~xianeizhang/

• Workflow
• How to run?
• Example of figures
• Checklist

Workflow

• shell file fetches data paths

  top-level shell file '*.sh' gets data file paths from cfg file 'cfgs/para.cfg';

• shell file scan the specified data path

  within the data path, '*.sh' scans to find data file (*.dat) and figure property file (*.cfg);

• draw

  '*.sh' then calls the corresponding python file to draw;

• python calling format

  python python_file data_file figure_cfg_file section_name_in_figure_cfg_file e.g.,

  python $PY_DIR/bar.py $data $bar_CFG_DIR/bar.cfg “read ratio figure”
  

• generated figure files (*.eps/*.png/*.pdf, …) are saved into the data file directory;

  Note: file name is automatically created by appending figure extension to data file

• provide the ‘*.pdf’ (by default, *.eps are converted to *.pdf) file path to your latex file.

  e.g.,

  \includegraphics[width=3in]{figures/rst/DATA/{stat_energyEDP.dat}.pdf}
  

How to run?

• prepare your data in the required format (see example data files or *.dat.format files;

  if you are collecting data with scripts, then it is easy to follow the format;

• write up shell file and specify data path

  edit ‘cfgs/para.cfg’ and shell file to locate your data files;

• update the figure cfg file (e.g., figure size, color, pattern and gap, etc);

  Note: I tried to make as many settings be not hard-coded. However, to avoid messy config, many others (e.g., major/minor ticks and figure dpi, etc) are still hard coded in the python files. If you need different settings of such parameters, then please hack the python files;

• set figure properties, e.g., height/width, pattern, color, ...

  specify which figure cfg file (bar.cfg/stacked.cfg/y2.cfg) to use and set the section (e.g., "xxx figure") for each figure.

The top *.sh file can be written as follow

 1 #!/bin/bash
 2 
 3 #-- directory of path cfg file, i.e., para.cfg
 4 CFG_DIR="./cfgs"
 5 #-- directory of python files
 6 PY_DIR="./pythons"
 7 
 8 #-- include para.cfg to specify data file path
 9 . $CFG_DIR/para.cfg
10 
11 scatter_CFG_DIR=$scatter_DATA_DIR #-- figure cfg file is put together with data file
12 
13 #-- scan data files in the directory
14 for data in $(find $scatter_DATA_DIR -mindepth 1 -maxdepth 1 -type f -name "*.dat")
15 do
16         #--figure of memory comparison
17         if [[ $data == *memory_comp.dat* ]]; then
18                 echo $data
19                 python $PY_DIR/scatter.py $data $scatter_CFG_DIR/scatter.cfg "memory comp figure"
20         fi
21         [ -f ${data}.eps ] && `epstopdf ${data}.eps && rm ${data}.eps`
22 done

Example of figures

Note that PNG images are for this README only, the output figure is in PDF.

• historic bar

• stacked bar

• double y axis

• line

• log line (axis in log scale)

• sub plot (multiple sub plots in one single figure)

• scatter

• function plot (plot using some curve functions instead of data samples)

Checklist

• README: this file
• *.figures.sh : shell file to call python
• pythons/: python graph files
• cfgs/: data path configuration files
• DATA/: source data files
• example paper/: main figures in the paper were created with those scripts

1. historic bar

   • sh: bar.figures.sh
   • py: bar.py
   • path cfg: cfgs/para.cfg
   • draw cfg: bar.test.DATA/bar.cfg
   • data: bar.test.DATA/*.dat
   • out: bar.test.DATA/*.pdf

2. stacked bar

   • sh: stacked.figures.sh
   • py: stacked/stacked_[12].py
   • path cfg: cfgs/para.cfg
   • draw cfg: stacked.test.DATA/stacked.cfg
   • data: stacked.test.DATA/*.dat
   • out: stacked.test.DATA/*.pdf

3. double y axis

   • sh: y2.figures.sh
   • py: y2.py
   • path cfg: cfgs/para.cfg
   • draw cfg: cfgs/y2.cfg
   • data: y2.test.DATA/*.dat
   • out: y2.test.DATA/*.pdf

4. line (will be updated later)

   • sh: {line/log_line/subp_line}.figures.sh

   • py: {line/usual_line/log_line/subp_line}.py

   • path cfg: cfgs/para.cfg

   • draw cfg:

     line.test.DATA/line.cfg

     line.test.DATA/usual_line.cfg

     log_line.test.DATA/log_line.cfg

     subp_line.test.DATA/subp_line.cfg

   • data: *.test.DATA/*.dat

   • out: *.test.DATA/*.pdf

5. func (will be improved later)

   • sh: func.figures.sh
   • py: func/{leak/rc/tret_20nm/tret_60nm}.py
   • path cfg: cfgs/para.cfg
   • draw cfg: func.test.DATA/{leak/rc/tret}.cfg
   • data: func.test.DATA/*.dat
   • out: func.test.DATA/*.pdf