An Easy Way To Test sed Expressions

TL;DR

Make an `input.txt`` with some content and a `sed_runner.bash`` file with this:

sed_runner.bash

#!/bin/bash

cat input.txt | sed -E 's/\!\[Image: [^]]+\]\(([^\)]+)\)/<<\1|img>>/g'

Then setup watchexec to run the bash script every time it changes so you can see the output with this:

Code

watchexec --project-origin . -w sed_runner.bash ./sed_runner.bash

The Problem

Reducing friction and time to iteration is huge for me. I work best when the mechanics get out of the way.

I had a lot of clean up to do while transitioning my site from markdown/org-mode to neopolian. No surprise given that there' twenty years of files.

The main work was all done with rust's _nom__ parser. The clean up was done with the sed command line tool. I ended up with a bunch of commands like:

Code

sed -E -i "" 's/\!\[Image: [^]]+\]\(([^\)]+)\)/<<\1|img>>/g' a.txt

That turn's this:

<<aws-20120918--1724-02.jpg|img|caption: Image: aws-20120918--1724-02.jpg>>

Into this:

<</aws-20120918--1724-02.jpg|img>>

The Automation

Figuring out those expression took some doing, but I found a way to make it relatively painless.

My approach is to setup two files. One that has the input text of the string to work. Then I make a second file that uses _cat__ to send the input to set for testing.

That file looks something like this:

sed_runner.bash

#!/bin/bash

cat input.txt | sed -E 's/\!\[Image: [^]]+\]\(([^\)]+)\)/<<\1|img>>/g'

I `chmod u+x`bash` that file to turn it into an executable script. The last, and most important, step is to put a watcher on the script to run it every time it's saved. I use _watchexec__ for that. The command I run to start it is:

Code

watchexed --project-origin . -w sed_runner.bash ./sed_runner.bash

With all that setup it's just a matter of walking throught the sec expression to get things working by watching the output on the command line after every save.

retest the commands