[sui-rpc-api] symbolic resume for the scalar lanes 5/5

[wlmyng]

Description

While applying an after cursor, the intra-tx path converts an after-Item cursor into Excluded(coordinate), while the scalar u64 path eagerly computes Included(coordinate + 1). This PR makes both uniformly symbolic, and pushes the +1/u64::MAX handling to ScanBounds<u64>::to_range. This shrinks the ScanCursor trait to coordinate().

In practice, the fetch interval is invariant: to_range projects Excluded(n) and Included(n + 1) to the same Range<u64>. However, the bounds are also read by two frame decisions, which now see the raw coordinate instead of the successor:

1. ScanBounds::is_empty checks whether a cursor emptied the window and thus should set CursorBound. Included(n + 1) vs Excluded(n + 1) reads as empty, but Excluded(n) vs Excluded(n + 1) does not, so exactly-adjacent bounds no longer set CursorBound eagerly at resolution — the zero-row drain produces the frame instead. This is the current event/intra-tx behavior.
2. The admission comparison lower_bound_gte used to see Included(n + 1), which ties a window bound at the same coordinate and admits the cursor; Excluded(n) ranks below the tie, so a cursor coinciding with the window's own bound is now rejected outright.

Wire-visible changes outlined below, over window [10, 20):

scenario	old → new	via
asc, after = Item@19 (window's last item)	CursorBound at the request cursor → CheckpointBound at the window end	1
desc, after = Item@9 (one below window)	CursorBound at the cursor → the window's natural frame	2
desc, after = Item@14, before = Item@15	frame at before@15 → after@14 (the before's emptying claim no longer fires, so the after's stop-side stamp stands)	1
desc, after = Item@19	unchanged — the stop-side stamp never asks about emptiness	—
asc, after = Item@14, before = Item@15	unchanged — the asc stop side (before) stamps and claims the same coordinate	—

Beyond-window cursors still empty at resolution, and u64::MAX handling is unchanged
(the saturation now lives in to_range).

Test plan

• to_range_collapses_excluded_lo_at_successor
• raw_after_item_echo_resumes_like_successor_boundary
• tx_after_item_at_window_edge_defers_to_drain
• tx_after_item_at_window_edge_descending_keeps_stamped_terminal
• tx_after_item_below_window_start_descending_is_noop
• test_list_transactions_after_before_adjacent_empty_results (pre-flip preservation pin)
• test_list_transactions_after_last_item_ends_at_window_bound (cargo simtest -p sui-e2e-tests list_ledger_history)

Release notes

• Protocol
• Nodes (Validators and Full nodes)
• gRPC
• JSON-RPC
• GraphQL
• CLI
• Rust SDK

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
sui-docs	Ready	Preview	Aug 20, 2026 11:47pm

2 Skipped Deployments

Project	Deployment	Actions	Updated (UTC)
multisig-toolkit	Ignored	Preview	Aug 20, 2026 11:47pm
sui-kiosk	Ignored	Preview	Aug 20, 2026 11:47pm

[nickvikeras]

looks good to me. Thanks again for cleaning this up. Once you have handled the empty-range discrepancy we discussed over slack lmk and I'll take one last look.