feat(tournament): implement complete bracket generation - fix #448

[iyanumajekodunmi756]

Implements the four supported tournament formats as real rounds and matches instead of empty stubs that silently produced nothing.

New module

backend/src/service/bracket_generator.rs (~640 lines)

• BracketGenerator::generate(bracket_type, participants) -> GeneratedBracket dispatches to one of four pure-function algorithms.
• BracketGenerator::persist(tournament_id, bracket) inserts all rounds and matches inside a single Postgres transaction.
• BracketGenerator::generate_next_swiss_round(tournament_id) reads current standings (3 / 1 / 0 scoring) and pairs by descending score, honoring the no-repeat rule, with a Bye handed to the lowest available odd player if N is odd.

Algorithms

• Single Elimination: standard '1 vs N' seeding so the top two seeds can only meet in the final; byes auto-assigned to top seeds when N is not a power of two.
• Double Elimination: round-number sentinel scheme so the (tournament_id, round_number) UNIQUE index is never violated:
  • Winners Bracket: 1..=ceil(log2 N)
  • Losers Bracket: 101..=(101 + 2W - 2) (empty match shells; the
    advance orchestrator populates them after each WB round completes)
  • Grand Final: 201 (empty until LB winner is known)
  • Bracket Reset: 202 (empty unless LB winner takes GF)
• Round Robin: circle / polygon method. Position 0 is fixed and the
  remaining N-1 players rotate clockwise each round; deterministic pairings
  and proper Bye handling for odd N.
• Swiss: round 1 paired (1 vs N/2+1, 2 vs N/2+2, ...); subsequent rounds
  generated lazily via generate_next_swiss_round using current standings
  and the no-repeat pairing rule.

Round-number sentinels (100, 200, ...) keep WB / LB / GF in a single tournament_rounds table without collisions.

Wiring

backend/src/service/tournament_service.rs (~700 lines, was 2311 corrupted)

The previous file had a broken sqlx::query! macro, two duplicated
TournamentLeaderboardEntry / TournamentAnalyticsResponse structs, and
unpaired braces that prevented cargo check from compiling. Rewrote keeping
the working CRUD / payment / lifecycle helpers and delegating bracket
generation to BracketGenerator.

• generate_tournament_bracket fetches active participants in seed order, calls BracketGenerator::generate, then BracketGenerator::persist - all in one transaction.
• advance_swiss_round exposes the lazy Swiss round generator for round-by-round execution.
• get_tournament_bracket reads back the persisted rounds and matches.
• Removed the broken dashboard / leaderboard / analytics methods that are outside the scope of [BACKEND] - Double elimination, round robin, and Swiss bracket generation are unimplemented stubs #448 and were not referenced outside the service file (grep verified against server/, contract/, frontend/).

backend/src/service/mod.rs
Added pub mod bracket_generator; and removed a pre-existing duplicate
pub mod tournament_service; declaration.

Tests

11 unit tests in bracket_generator.rs cover:
* standard seeding order invariants
* single-elim with byes for non-power-of-two N
* double-elim round-number scheme and uniqueness
* round-robin: even count balanced, odd count bye rotation, no intra-round repeats
* swiss round 1: top-vs-bottom half for even N, lowest-seeded Bye for odd
* invalid input (zero / one participant) does not panic

Migrations

No new migration required. The bracket generator emits empty match Vecs
for LB / GF shells and writes only round rows for them; the existing
tournament_rounds / tournament_matches schema is sufficient and the
player1_id NOT NULL FK constraint is preserved.

References: #448

Summary

Type of change

• Bug fix
• New feature
• Breaking change
• Refactor / cleanup
• Docs / config only

Related issues

Closes #

Changes

Testing

• Unit tests pass (cargo test / npm test)
• Contracts tests pass (cargo test in contracts/)
• Manually tested locally
• Migration tested against a fresh DB

Checklist

• Code follows project conventions
• No secrets or PII committed
• Migrations are reversible (.down.sql exists)
• Contract changes are backward-compatible or versioned
• CI passes

closes #448

[vercel]

Someone is attempting to deploy a commit to the paul joseph's projects Team on Vercel.

A member of the Team first needs to authorize it.

[drips-wave]

@iyanumajekodunmi756 Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits