tl;dr§

for FILE in *.md; do
	WORD_COUNT=$(sed '1 { /^---/ { :a N; /\n---/! ba; d} }' "${FILE}" | ws -w | awk '{ print $1 }')
	yq --front-matter="process" ".word_count=$WORD_COUNT" -i "${FILE}"
done

Motivation§

This is part of my ongoing project of using Obsidian as a read-it-later app. Since Omnivore (🪦2022-2024) and Pocket (🪦2007-2025) shut down, one right after the other, I decided I was done with SaaS to save my reading.

While procrastinating at work I got curious about what I could do with the data from my read-later files in Obsidian's new Bases feature, in order to data-mine myself Luxury Surveillance-style. I thought it would be nice to be able to sort my unread saved articles by word count/estimated reading time so I can get some of the shorter ones out of the way, but there's no built-in way to access and process file content with arbitrary JavaScript in Bases formulas AFAIK.

Thanks to an Obsidian forum post I found out I can add a property to my Web Clipper template to calculate word counts at snip-time, but what about my 8-month backlog of articles?

Parts of the problem§

Getting the word count of a file Solved by wc Print just the word count with wc -w input.md | awk '{ print $1 }' wc invocation (GNU Coreutils 9.10)

...but count only the file content, not the metadata Obsidian markdown files include YAML data in a header enclosed by hyphens (---). We can get just the body content with sed to pipe into wc:

sed '1 { /^---/ { :a N; /\n---/! ba; d} }' "${FILE}"

Kudos: text processing - How to remove YAML frontmatter from markdown files? - Stack Overflow

Add a property to the metadata We can use yq's --front-matter="process" flag to manipulate the frontmatter in-place.

yq --front-matter="process" ".word_count=100" -i input.md

Kudos: https://jonalmeida.com/til/yq-editing-frontmatter/

Script§

for FILE in *.md; do
	WORD_COUNT=$(sed '1 { /^---/ { :a N; /\n---/! ba; d} }' "${FILE}" | ws -w | awk '{ print $1 }')
	yq --front-matter="process" ".word_count=$WORD_COUNT" -i "${FILE}"
done

Interpolating the WORD_COUNT variable like that in the yq script is probably bad, but it worked well enough for me.