hissyfit can be used at the end of Unix pipes (or read from file) to draw
terminal histograms and bar charts for quickly gauging numerical data and count
data - see screenshots and link to raw output below.
No need to fire up R or Python. It's just one script. Example usage with some random data is below. Sadly github markdown does not allow setting of line-height, hence the stripy look. It looks better in the terminal (screen shots below).
for i in {1..100000}; do echo $(( (RANDOM+RANDOM+RANDOM)/3)); done | \
hissyfit --cv=0/32768/64/16 --x=0%4000 --d=5461/10922/21845/27306
▂ ▂█▅
▆▄█████▆▇▇
▃███████████▄▂▂
▄▆███████████████▃
▂██████████████████▃
▂████████████████████▅
▆██████████████████████▃
▂████████████████████████▇
▅██████████████████████████▆
▁▆████████████████████████████▇▂
▃████████████████████████████████▂
▃██████████████████████████████████▇
▃▇████████████████████████████████████▇▅
▄█████████████████████████████████████████▇▅
▂▅█████████████████████████████████████████████▇▅▁
▁▁▃▅▇██████████████████████████████████████████████████▇▅▃▂▁
____.____,____.____,____Q____,__Q_.____Q____.____,____.____,____
| | ◆ | ◆ | | | ◆ | ◆| |
0 4000 8000 12000 16000 20000 24000 28000 32000
5461 10922 21845 27306
▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭▭
ND=100000 hmin=0 hmax=32768 dmin=359 dmax=32391
topbin(i=33,n=3635,p=3.6,x~16640) bw=512 qr=(12559 16425 20276)
Division bin sizes
Left Right Count Pct Sum Pct
359 5461.3 2085 2.08 2085 2.08
5461.3 10922.7 14396 14.40 16481 16.48
10922.7 21845.3 66780 66.78 83261 83.26
21845.3 27306.6 14685 14.69 97946 97.95
27306.6 32391 2054 2.05 100000 100.00
All hissyfit parameters here are optional.
The --cv (canvas) parameter can set up to four parameters simultaneously: min and max histogram
endpoints, the number of bins, and the height of the histogram.
The --x parameter sets ticks for display. When given the format N%w it results in all ticks that are w
apart, anchored at position N, and fit in the range of data shown. Alternatively, it can be supplied
with the format x1[/x2]* (a list of numbers separated by slashes) to set each tick explicitly.
The --d parameter defines variable-sized super-bins for which hissyfit will
output a table of counts and percentages. This option can be useful after initial inspection
of a histogram to gauge the size of different modes. In this example the division values were
chosen to represent the four points { μ ± σ, μ ± 2σ } where μ and σ are the expected mean and deviation
for a sum of three uniform distributions on [0,32768].
The default number of bins is 100. With no parameters given hissyfit output looks like this:
for i in {1..100000}; do echo $(( (RANDOM+RANDOM+RANDOM)/3)); done | hissyfit
█▁ ▅▃▃
▆█▇▇▅██▅▆███ ▁▃
█ ▇████████████████▅▅
▇██▆███████████████████▄
▃███████████████████████████
▃▂████████████████████████████▄▆
▁████████████████████████████████▇
███████████████████████████████████▂
▁█████████████████████████████████████▁▁
▂████████████████████████████████████████▁▁
████████████████████████████████████████████▂
▃▅██████████████████████████████████████████████▁
▄█████████████████████████████████████████████████▆
▃▅████████████████████████████████████████████████████▆
▂▆████████████████████████████████████████████████████████▇▂
▆████████████████████████████████████████████████████████████▆▂
▁▄▆▇███████████████████████████████████████████████████████████████▇▃
▃▅██████████████████████████████████████████████████████████████████████▆▅▁
▁▂▃▇▇███████████████████████████████████████████████████████████████████████████▇▅▃
▂▁▃▆▆████████████████████████████████████████████████████████████████████████████████████▆▅▄▄▂
375.0 ___,____.____,____.____,____.___Q,____.____,Q___.____,__Q_.____,____.____,____.____,__ 32289.0
ND=100000 hmin=375 hmax=32289 dmin=375 dmax=32289
topbin(i=50,n=2254,p=2.3,x~16172.43) bw=319.14 qr=(12519 16388 20225)
For numerical data expected input is a stream of numbers, one per line.
Options are min and max
histogram endpoints (--min --max, simultaneously acts as data sub-selection),
height/resolution (--height in lines, resolution is 8 times larger),
cumulative tallying (--cumulative),
the number of bins (--bins),
and user-specified glyphs (--stairs).
It is also possible to read input that already specifies the histogram heights (--histin).
For bar charts hissyfit accepts the output of uniq -c (that is, each line has <count> <label>)
to draw bar charts for categorical data (see fourth screenshot), when given the --tallyin option.
In this case the output is in the form of horizontal bars with labels in
front and counts after.
To sort by label (or otherwise), simply use unix sort before uniq -c, e.g.
sort | uniq -c | hissyfit.
To sort by count, use sort | uniq -c | sort -n | hissyfit.
Histograms and bar charts are drawn with Unicode block glyphs by default. Use
--plain to disable this and use --stairs to provide your own glyphs.
Sticking the output within <pre></pre> tags is a cheap way of getting
histograms in HTML. Below are some screenshots; the corresponding raw text
(and display) in HTML/<pre> can be seen either as a github
preview,
right in this repo
(source) or
here at micans.org.
Hissyfit is just one script. Projects with more code and much greater capabilities are
- Jp - terminal plots from JSON (or CSV) data
- Termgraph - basic graphs in the terminal
- YouPlot - plots in the terminal
- bashplotlib - plots in the terminal
- plotext - plot directly in the terminal
- distribution - character-based graphs in the terminal
