grain.social/grain

Switch repository

Loading…

What changes

The archive. social.grain.story.getArchive lists an account's stories, expired or not, the latest first, paged by a cursor of the last story's createdAt and URI, limit 1 to 100. It leaves out what every list of stories does: an account not shown, a story with a label that hides (grain's or its author's), and an account that blocks, or is blocked by, the viewer. Each is a storyView, its viewer.viewed and viewer.fav filled for the viewer.

What was watched. social.grain.story.markViewed keeps the stories the caller watched in the index's story_view (index/stories.rs), which getStories, getStory, getArchive and getAuthors already read. The apps call it with a service token, the web with its cookie from a page of the site. Marking one twice changes nothing; a story the index does not hold, deleted since it was read, is passed over. Each call lets go of views more than a week old, as a story stands a day.

One of the pulls closing what the clients call and grain-serve did not serve; the table of them all is in gaps/graph's body (Accounts mute and unmute, and list whom they block and mute).

Why

The web and both apps call both, and grain-serve answered 501. knotless has neither, so this follows grain's own reads: the list shape, cursor and filters of read/story.rs.

Differences from the old server

• An archive's cursor is the last story's time and URI, not its time alone, so two stories made in the same millisecond are not lost between pages.
• The archive leaves out stories with a label that hides, and accounts the viewer blocks or is blocked by, as getStories does; the old archive left out only accounts gone at their host or taken down.
• An actor the index does not know is refused (400 Profile not found), as every query naming an actor is here; the old one answered an empty list.
• Old views were pruned when the server started; here each markViewed lets go of those a week old.
• stories must be at-URIs, 1 to 100, as before; a URI that is not one is refused before anything is kept, where the old server checked only at://.

How it was checked

crates/grain-serve/tests/stories.rs, against the reference PDS for the service tokens: the archive in order, expired flagged, a labelled story left out, paged, empty for an account taken down, an unknown or missing actor refused; marking twice and a story not held, viewer.viewed in getStories and getArchive, unviewedCount in getAuthors, nothing watched for no one, and the input's bounds refused. index/stories.rs has a unit test of keeping and letting go, and the plan check covers the archive's query. The rest is the runner's.

Shortcomings

• Views stay when an account is deleted until gaps/deletes purges them.

🤖 Generated with Claude Code

This adds getArchive and markViewed. getArchive reuses the existing listed() and unblocked filters and the page and after cursor helpers. markViewed writes to the existing story_view table with INSERT OR IGNORE on the (did, subject) primary key. It skips URIs with no matching story and prunes views older than a week, all in one transaction. The input is validated before anything is written, and the service-token and cookie auth goes through viewer::writer. I found no bugs. Tests cover paging, expiry flags, labelled and taken-down accounts, idempotent marking, and the input bounds.

The minor point is the lib.rs module doc, where the rewrapped paragraph leaves one over-long line. I haven't read the checks, so I don't know whether they pass.

Carried forward: the change this pull makes at 32b56f7be8fa23e6e0d3e30eab0f2f2b3aed06d5 is the one reviewed at 92f7c8c5a258e9d6f009e7728d18ef13cccda4a5, so that review stands. What is new comes from its base.

The over-long line in lib.rs's module doc: rewrapped at 1e8d010.

The only change since my last review is the rewrap of the module doc in crates/grain-serve/src/lib.rs. That was the one point I raised, and it's done. The rest of the pull is the same getArchive and markViewed change I already reviewed and found no bugs in. The other changes in the head come from the merge of feature/rust, not from this pull. The four checks (test, rust, harness, check) were still queued when I looked, so I can't say they pass.

Landed on feature/rust at fc5f686 through the batch pull Seven reviewed pulls land on feature/rust together, checked once, checked there as a whole. This pull's head is in feature/rust's history.