fix: omit pagination token when all shard cursors are null

[PirosB3]

Fixes a pagination bug that occurs when pageToken is null across all shards. In that scenario the implementation serialized [null, null, ...] and returned it as the next page token, instead of omitting the cursor. Affects all multi-shard paginated endpoints: get_events, get_casts_by_parent, get_casts_by_mention, get_reactions_by_cast, get_reactions_by_target, get_links_by_target.

Result: clients receive duplicate results and keep paginating in a loop.

The HTTP layer had related issues that are also fixed here:

• EventsResponse was missing the nextPageToken field entirely, so HTTP clients couldn't paginate events at all.
• map_proto_messages_response_to_json_paged_response coerced None into Some("") instead of omitting the field.

Tests: adds test_get_casts_by_parent_returns_none_token_when_exhausted, which drains a query across two shards and asserts next_page_token is None once exhausted. Fails on main, passes on this branch.

[vercel]

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

Project	Deployment	Actions	Updated (UTC)
snapchain-docs	Ready	Preview, Comment	Apr 28, 2026 11:03pm

[github-actions]

Diff Coverage

Diff: origin/main...HEAD, staged and unstaged changes

• src/network/http_server.rs (89.7%): Missing lines 2967-2969
• src/network/server.rs (68.4%): Missing lines 1876,1880,1952,1956,2459,2463

Summary

• Total: 48 lines
• Missing: 9 lines
• Coverage: 81%

src/network/http_server.rs

  2963                 .events
  2964                 .iter()
  2965                 .map(|e| map_proto_hub_event_to_json_hub_event(e.clone()).unwrap())
  2966                 .collect(),
! 2967             next_page_token: events_response
! 2968                 .next_page_token
! 2969                 .map(|t| BASE64_STANDARD.encode(t)),
  2970         })
  2971     }
  2972 
  2973     async fn get_event_by_id(&self, req: EventRequest) -> Result<HubEvent, ErrorResponse> {

---

src/network/server.rs

  1872             .collect();
  1873 
  1874         let next_page_tokens: Vec<Option<Vec<u8>>> =
  1875             pages.into_iter().map(|page| page.next_page_token).collect();
! 1876         let next_page_token = serialize_page_token_if_any_shard_has_next(next_page_tokens)?;
  1877 
  1878         let response = MessagesResponse {
  1879             messages: combined_messages,
! 1880             next_page_token,
  1881         };
  1882 
  1883         Ok(Response::new(response))
  1884     }

---

  1948             .collect();
  1949 
  1950         let next_page_tokens: Vec<Option<Vec<u8>>> =
  1951             pages.into_iter().map(|page| page.next_page_token).collect();
! 1952         let next_page_token = serialize_page_token_if_any_shard_has_next(next_page_tokens)?;
  1953 
  1954         let response = MessagesResponse {
  1955             messages: combined_messages,
! 1956             next_page_token,
  1957         };
  1958 
  1959         Ok(Response::new(response))
  1960     }

---

  2455             .collect();
  2456 
  2457         let next_page_tokens: Vec<Option<Vec<u8>>> =
  2458             pages.into_iter().map(|page| page.next_page_token).collect();
! 2459         let next_page_token = serialize_page_token_if_any_shard_has_next(next_page_tokens)?;
  2460 
  2461         let response = MessagesResponse {
  2462             messages: combined_messages,
! 2463             next_page_token,
  2464         };
  2465 
  2466         Ok(Response::new(response))
  2467     }

---

[Copilot]

Pull request overview

Fixes a cross-shard pagination bug where an “exhausted” cursor was still being returned as a serialized [null, null, ...] token, causing clients to paginate indefinitely and receive duplicate results.

Changes:

• Add serialize_page_token_if_not_empty helper to omit next_page_token when all shards are exhausted (all per-shard tokens are None).
• Apply the helper across several multi-shard paginated RPC responses.
• Update HTTP JSON pagination token behavior to omit nextPageToken when absent, and add nextPageToken support for events; add a regression test for get_casts_by_parent.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 4 comments.

File	Description
src/network/server.rs	Introduces and uses a helper to avoid emitting a pagination token when all shards are exhausted.
src/network/server_tests.rs	Adds a regression test ensuring get_casts_by_parent returns None next_page_token when drained.
src/network/http_server.rs	Ensures nextPageToken is omitted (not "") when absent; adds events nextPageToken field support.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.