Here are some shell scripts and niceties I use to make my life easier. They are not offered as packages; they are meant to be copied around.
I use hist.sh, which I am pretty sure was authored by Carl Shapiro, for
profiling things. It provides a text-based histogram for whatever lines are
piped to it.
#!/bin/sh
# Likely originally by Carl Shapiro
sort |
uniq -c |
sort -n |
sed 's/^ *//' |
awk 'BEGIN { OFS="\t"} { sum += $1; print sum, $1, substr($0, 1 + index($0, " ")) }' |
sort -n -r |
awk 'BEGIN { IFS=OFS="\t"; print "cum", "pct", "freq" }$1 > max { max = $1 }{ print int($1/max*100), int($2/max*100), $2, $3 }'
I’ll modify a program like python to print out (for example) comma-separated
triples %s,%s,%s of slot name, left argument type, and right argument type.
Then I will invoke it like ./modified-python script.py | hist.sh and the
output looks like:
$ hist < err
cum pct freq
100 59 87 NB_ADD,str,str
40 16 24 NB_AND,int,int
23 10 15 NB_INPLACE_ADD,str,str
13 6 9 NB_ADD,int,int
7 2 3 NB_OR,int,int
5 1 2 NB_SUBTRACT,int,int
4 1 2 NB_MULTIPLY,str,int
2 0 1 NB_REMAINDER,str,tuple
2 0 1 NB_LSHIFT,int,int
1 0 1 NB_ADD,tuple,tuple
0 0 1 NB_ADD,bytes,bytes
$
Which indicates to me that str + str is the most frequently occurring use of
+ in my script (for example).
I use run.py, which was adapted from a Python program written by
Matthias Braun, to help orchestrate shell stuff from Python. It’s a wrapper
around subprocess.run that adds a verbose mode, more enforceable
timeouts, and a PTY option.
I copy this into projects frequently.
It’s pretty easy to generate Ninja files and BYOB1 and I wrote a blog post about that. Then I discovered someone’s very small single-file Ninja reimplementation in Python and mashed them up into a little demo.
This relies on ninja_syntax.py which is copied from the Ninja project.
Sometimes you have a bunch of tasks to run and want quick-n-dirty
Buck/Bazel-type output. For that, I wrote lines.py and a blog
post about it.
Build Your Own Bazel ↩