Struct Cursor 

pub struct Cursor<'a, E, X> { /* private fields */ }

A forward-only keyset cursor for streaming large tables.

Rather than OFFSET (which the database must scan past, getting slower deeper in) or loading the whole table, the cursor remembers the last primary key it saw and asks for WHERE pk > ? ORDER BY pk LIMIT page_size. Every page is an index seek, so cost stays flat however deep you walk — and only one page is in memory at a time. This is the only iteration strategy that is also correct on D1-over-HTTP, which cannot hold a streaming server-side cursor.

let mut cur = repo::cursor::<User, _>(&exec, 500);
while let Some(page) = cur.next_page().await? {
    for user in page { /* ... */ }
}

Implementations

impl<'a, E, X> Cursor<'a, E, X>

where E: Entity, X: Executor,

pub fn after(self, pk: impl Into<Value>) -> Cursor<'a, E, X>

Resume from a known primary key — the next page starts strictly after pk. Lets a walk be checkpointed and continued across requests (e.g. one D1 HTTP call per invocation).

pub async fn next_page(&mut self) -> Result<Option<Vec<E>>, Error>

Fetch the next page. Returns None once the table is exhausted. A final short page (fewer than page_size rows) marks the end, so the following call returns None without a wasted round-trip.

pub fn checkpoint(&self) -> Option<&Value>

The primary key the next page will start strictly after, if a page has been read. Persist this to resume the walk later via after.