I love open source developers but there are times when I question their damn naming practices. I'm currently working with a giant data repository started as a "Z Standard" or "zstd" compressed file. And while I know that means "Z Standard", I can't help but look at it as "Z std". Oy.

Anyway. Zstd is a Facebook standard for data compression and it is strikingly effective. I've got over 100 gigs of JSON encoded data stored in a 13.7 gig file. Now I am aware that text compresses actually quite well but still 100 gigs in 13.7 gigs of space feels like wow.


If you're on a Mac then brew, as always, is your very best friend:

brew install zstd

Useful Command Lines

Assume that pol.zst is the name of the archive and it is located in your current directory.

Examining a handful of records:

zstd -cd pol.zst | head -n100

this dumps a stream of records out that are then fed into head which limits the quantity to 100.

The zstd -c and -d options mean:

-c     : force write to standard output, even if it is the console
-d     : decompression

Integrating the often useful jq (which just gets a single json element out):

zstd -cd pol.zst | jq '.timestamp'

And like all good *nix pipelines, this is composable (this example would extract the first 1000 records and then reduce them to only the comment element from the json):

zstd -cd pol.zst | head -n1000 | jq '.comment'

To count the total records in the zst file:

zstd -cd pol.zst | wc -l

Happily help is also available with:

zstd --help

Thank You's

Kudos to Facebook for another great bit of Open Source contributed back to the world. Also thanks to Grant Vousden-Dishington, the contributor of these command lines. He's been doing Zstd for a while; I'm the noob here.