Skip to content

Latest commit

 

History

History
58 lines (45 loc) · 3.25 KB

export-timeline-to-sqlite.md

File metadata and controls

58 lines (45 loc) · 3.25 KB
Raw

Export a Mastodon timeline to SQLite

I've been playing around with the Mastodon timelines API. It's pretty fun!

I'm running my own instance which means I don't feel limited by politeness in terms of rate limits - if I'm just using my own instance's resources I feel fine hammering it.

Many instances (including mine) expose a public timeline API. You can use that to request all "local" posts on the server - since my server only has one user (me) that's all of my stuff.

That API starts here: https://fedi.simonwillison.net/api/v1/timelines/public?local=1

If you fetch it through curl -i to include headers you'll spot some interesting headers:

$ curl -s -i 'https://fedi.simonwillison.net/api/v1/timelines/public?local=1'
...
X-RateLimit-Limit: 300
X-RateLimit-Remaining: 294
X-RateLimit-Reset: 2022-11-05T00:40:00.971772Z
Cache-Control: no-store
Vary: Accept, Accept-Encoding, Origin
Link: <https://fedi.simonwillison.net/api/v1/timelines/public?local=1&max_id=109277272598366124>; rel="next", <https://fedi.simonwillison.net/api/v1/timelines/public?local=1&min_id=109288139761544382>; rel="prev"

The rate limit stuff is cool, but the thing that instantly caught my eye was the Link: header. It looks like Mastodon implements cursor-based pagination in that header.

I just so happen to have written a tool for working with that kind of header! paginate-json, which fetches every page by following those rel="next" links until they run out.

Here's how to download everything from that feed and save it to a single newline-delimited JSON file (the --nl option):

paginate-json 'https://fedi.simonwillison.net/api/v1/timelines/public?local=1' --nl > /tmp/simon.json

Output:

https://fedi.simonwillison.net/api/v1/timelines/public?local=1
20
https://fedi.simonwillison.net/api/v1/timelines/public?local=1&max_id=109277272598366124
20
https://fedi.simonwillison.net/api/v1/timelines/public?local=1&max_id=109276505589113334
20
https://fedi.simonwillison.net/api/v1/timelines/public?local=1&max_id=109276498877771086
13
https://fedi.simonwillison.net/api/v1/timelines/public?local=1&max_id=109276387855048126
0

And here's how to load the resulting file into a SQLite database using sqlite-utils, which also supports newline-delimited JSON:

cat /tmp/simon.json | sqlite-utils insert /tmp/simon.db posts - --pk id --nl --replace

The --pk id option sets the primary key on the newly created posts table to id. The --replace option means any records that clash with an existing ID will be replaced - this means I can safely run the command multiple times.

I can even combine the two commands like this:

paginate-json 'https://fedi.simonwillison.net/api/v1/timelines/public?local=1' --nl \
  | sqlite-utils insert /tmp/simon.db posts - --pk id --nl --replace

The result is a SQLite database!

Here it is in Datasette Lite as a demo.