A collection of random useful scripts - ariel faigon
Handy plotting command line utilities.
Note: requires 'cuts'
x plots a density chart of a uni-dimensional numeric vector
xy plots a scatter-plot of two numeric columns (x, y)
Documentation (man page style) is included, to get it, run:
$ perldoc x
Note that 'xy' doesn't exist in this repository. You can
create it by simply running this command in a shell:
$ ln x xy
in the directory where 'x' resides.
Handy cut and paste of columns (more powerful than 'cut')
See: https://arielf.github.io/cuts/
Python (using matplotlib) script to quickly plot 3D numeric data
from the command-line. Expects a *.tsv or *.csv input file, and
3 names (or 0-based integer indices) of columns to select from
the file and map them to the 3D (X, Y, Z) dimensions.
Supports many options to control how the plot looks like.
Emphasis on friendliness. Order of arguments doesn't matter.
Run without arguments for usage.
The default is to plot a contour map (topographical view)
of the Z dimension vs the X & Y dimensions.
For all options/parameters, prefix matches and regexp/abbreviations
are supported. For example a color-map parameter, can be
specified as any of:
colormap
cmap
cm
etc.
Examples:
# Get a usage summary:
xyz
# Visualize the R 3D volcano data-set:
xyz volcano.csv X Y Z
# Same + also modify the color-map & Z-axis color-resolution
xyz volcano.csv X Y Z cmap=jet zres=20
# Similar, but use integer indices (0,1,2) for column-selection,
# disable grid-lines (gl=0) and countour-lines (cl=0)
# increase Z-dimension resolution & add dots where data appears.
xyz 0 1 2 volcano.csv dotsize=0.01 zres=100 gr=0 cl=0
The data file volcano.csv (credit: Ross Ihaka)
is included here for convenience.
The following names can be symlinked to the same (xyz) script.
* If called as 'xyzb' will produce a bubble chart
* If called as 'xyzs' will produce a scatter plot
* If called as 'xyzp' will produce a polar plot (X-dim mapped to angle)
Print indices of column-names from a csv/tsv file with a header
Example:
colidx volcano.csv
My Find File.
Basically a wrapper around multiple stages of `locate` and `grep`.
Helps find files by name. Especially useful if you only vaguely
remember parts of the file name (say "jpe?g" to search for photos,
or a name of a person).
Smart about doing case-insensitive searches, supporting PCRE regexes, and more.
Examples:
mff katelin 'jpe?g'
mff beatles rubber.soul mp3
Handy bi-directional/smart universal time converter
- If it sees a Unix time_t (integer) - will convert to date
Example:
$ T 1443526029
Tue Sep 29 04:27:09 2015
- If it sees anything else - will convert to time_t
Example:
$ T week ago
1443381375
Input flexible: Works on either command-line-args OR stdin
percentage and cumulative-percentage summary of a list of items (one per line)
Usage example: summarize word frequency in /etc/passwd:
$ tr -cs 'A-Za-z' '\012' < /etc/passwd | sorted-count-with-pct
Sort data-file(s) descending by absolute value of some field.
Can define separator and field-number to sort by.
Use:
sort-by-abs -h
for a usage message
Calculates entropy of a file (or string) object by byte values.
One pass over the file/string to summarize frequencies of each
byte value and count total bytes.
Another loop over the (non-zero) frequencies to calculate the entropy.
Very simple: doesn't try to do N-gram/context/language-detection,
Only considers char-frequencies within the object as independent
probabilities.
Result is always normalized to [0 .. 1] range.
1.0 means highest-randomness.
Time a sequence of commands (each output line is timed separately).
e.g. you have a test-suite, normally invoked as 'make test' with many sub-tests and we want to get the sub-tests sorted by the time they took to run.
Example:
time-by-line make test
The output (trimmed for brevity):
[normal output comes first. And after everything completes:]
0.000006 test 42: OK
...
0.149873 test 127: OK
0.241587 test 134: OK
0.602354 test 126: OK
where the 1st column is number of seconds a line took to complete, and the lines are sorted by the time the individual sub-tests took to run.
Invocation is flexible. You can also use 'time-by-line' in a pipe:
make test |& time-by-line
There are 2 different ways to attribute times to lines:
- Attribute time to 1st line in a sequence (starting line)
- Attribute time to 2nd line in a sequence (ending line)
The default used by time-by-line is to attribute times to the ending-line
To change the default, simply pass - or -- as the 1st argument to time-by-line:
# Will attribute each time-slot duration to its 1st line:
time-by-line -- make test
# ditto:
make test |& time-by-line --
Kill high-CPU hogging processes by command+args regex & CPU consumption.
Buggy (or malicious) 3rd-party javascript code is often taking our browsers and computing resources hostage.
I used to find some processes on my desktop running at 100% CPU all night when I was not even noticing, because of bad javascript.
Involuntary CPU hogging might be caused by an infected ad, badly written code, "drive-by-scripting" leading to someone co-opting your computer to run crypto-mining, or much worse.
It makes you pay for electricity you don't want to, and multiplied by many desktops and browsers left running at night, is also very bad for our planet.
cpu-hog-killer is a simple script which I run from periodically
during the night. I make it run from cron every 13 minutes like this:
# Add this line (or similar) using 'crontab -e'
*/13 0-7 * * * ~/bin/cpu-hog-killer
It identifies CPU hogging processes, mainly inside Chrome or Firefox, and kills them instantly upon detection. Note that it doesn't kill processes taking CPU indiscriminately. It only kills those processes that you explicitly say you want killed based on a detailed regexp + CPU consumption parameters.
In the morning when I'm back, the worst case scenario is that I go to a browser tab and it says (e.g. for Firefox):
"Gah. Your tab just crashed."
I take notice of which site was misbehaving (based on its title) and the browser allows me (presenting a big button) to:
[Restore This Tab]
if I want to, so nothing is actually lost.
You can add more rules to the script to cover more apps that may be hogging your CPU when you don't want them to. Just look at the 'main' section (last few lines of the script) and add more rules as needed. Each rule looks like this:
process_list '<some pattern>' | terminate-hogs
For example, the Firefox rule which only kills one tab is:
process_list '[/]firefox -contentproc -childID.*tab$' | terminate-hogs
And for Chrome/chromium it is:
process_list '[/]chrom(e|ium) --type=(renderer|utility)' | terminate-hogs
I've seen (rare) cases where even KDE plasma was spinning
at 100% due to some bad widget/add-on, so there's a rule for that as well.
YMMV.
To restart a terminated plasma shell, simply run:
kstart5 plasmashell