Skip to content

alesr/tango

Repository files navigation

Tango

codecov

⚠️ Heads Up

This is an experimental/toy project. I built it to explore some ideas around matchmaking using concurrent patterns in Go. While it works, I wouldn't recommend using it in production. Feel free to poke around and maybe grab some ideas for your own projects.

How Tango Works

Flow Overview

When you start Tango, it spins up three main background processes:

  1. An operation processor that handles new players/hosts and other match-related operations
  2. A timeout checker that removes players with expired matching timeouts
  3. A stats updater that periodically collects matchup statistics

Note: The diagrams below only show the operation processor and timeout checker for simplicity.

                               ┌── Host → Creates new match
Player Enqueue → Operation ────┤
                               └── Player → Attempts to join existing match

Players can be either hosts (create matches) or joiners (look for matches). When a host joins:

  • A match is created immediately
  • Tango starts actively looking for suitable players to join this match

When a regular player joins, Tango keeps trying to find a suitable match until either:

  • A match is found
  • The player times out
  • The context is cancelled

Matchmaking Flow

%%{init: {'theme': 'base', 'themeVariables': { 'fontFamily': 'arial', 'fontSize': '16px', 'textColor': '#2A4365' }}}%%
flowchart TB
    subgraph Matchmaking
        Start[Start Tango] --> Queue{Player Queue}
        Queue -->|Host| CreateMatch[Create Match]
        Queue -->|Player| WorkerPool[Worker Pool]
        
        CreateMatch -->|From Pool| NewMatch[New Match]
        NewMatch --> SeekPlayers[Seek Players]
        
        WorkerPool -->|Attempts| FindMatch[Find Match]
        FindMatch -->|Success| JoinMatch[Join Match]
        FindMatch -->|No Match| Retry((Retry))
        Retry --> FindMatch
        
        SeekPlayers -->|Match Found| JoinMatch
        
        JoinMatch --> Stats[Update Stats]
    end

    style Start fill:#48BB78,color:#2F855A
    style Queue fill:#667EEA,color:#F7FAFC
    style WorkerPool fill:#9F7AEA,color:#F7FAFC
    style JoinMatch fill:#48BB78,color:#2F855A
    style Stats fill:#4299E1,color:#F7FAFC
    style Matchmaking fill:#EBF8FF,color:#2C5282
Loading

Cleanup Flow

%%{init: {'theme': 'base', 'themeVariables': { 'fontFamily': 'arial', 'fontSize': '16px', 'textColor': '#2A4365' }}}%%
flowchart TB
    subgraph Cleanup
        Timer[Timeout Checker] --> CheckPlayers{Check Players}
        CheckPlayers -->|Expired| RemovePlayer[Remove Player]
        CheckPlayers -->|Active| Continue[Continue Checking]
        
        RemovePlayer -->|Is Host| CleanMatch[Cleanup Match]
        RemovePlayer -->|Not Host| DeletePlayer[Delete Player]
        
        CleanMatch --> Events[Emit Events]
        CleanMatch --> ReturnMatchPool[Return Match to Pool]
        CleanMatch --> UpdateStats[Update Stats]
        
        DeletePlayer --> Events
        DeletePlayer --> UpdateStats
    end

    style Timer fill:#4299E1,color:#F7FAFC
    style ReturnMatchPool fill:#ED64A6,color:#F7FAFC
    style CleanMatch fill:#9F7AEA,color:#F7FAFC
    style DeletePlayer fill:#ED8936,color:#F7FAFC
    style Events fill:#48BB78,color:#2F855A
    style UpdateStats fill:#667EEA,color:#F7FAFC
    style Cleanup fill:#F0FFF4,color:#2C5282
Loading

Setup and Lifecycle

import "github.com/alesr/tango"

// Create a new Tango instance
tango := tango.New(
    tango.WithOperationBufferSize(100),
    tango.WithDefaultTimeout(5*time.Second),
    // More options available...
)

// Start the service
err := tango.Start()

// Shut it down!
err := tango.Shutdown(ctx)

Player Management

// Create a new player
player := tango.NewPlayer(
    "player-1",     // ID
    false,          // IsHost
    tango.Mode1v1,  // Game mode
    deadline,       // Timeout (how long it should look for a match)
    []string{"tag"} // Optional tags (not in use yet)
)

// Add to matchmaking queue
err := tango.Enqueue(ctx, player)

// Remove from system
err := tango.RemovePlayer("player-1")

Match Operations

// Get all active matches
matches := tango.ListMatches()

Stats

You can call the Stats() method at any time to retrieve match statistics:

stats, _ := tango.Stats(ctx)

fmt.Printf("Total Players: %d\n", stats.TotalPlayers)
fmt.Printf("Total Matches: %d\n", stats.TotalMatches)

Configuration Options

  • WithLogger: Custom logger for the service
  • WithOperationBufferSize: Size of the operation channel buffer
  • WithMatchBufferSize: Size of the match channel buffer
  • WithAttemptToJoinFrequency: How often to try matching players
  • WithCheckDeadlinesFrequency: How often to check for timeouts
  • WithDefaultTimeout: Default operation timeout

Game Modes

  • GameMode1v1: 1 host + 1 player
  • GameMode2v2: 1 host + 3 players
  • GameMode3v3: 1 host + 5 players

Future Improvements

Some cool ideas I'd like to explore:

  • Use tags for smarter matching (skill levels, regions, etc.)
  • Add match status notifications (websockets maybe?)
  • Match lifetime management (auto-cleanup after game ends)
  • Match history tracking
  • Support for tournament-style matchmaking
  • Custom matching rules
  • Support for teams

Got more ideas? I'd love to hear them!

Load Testing

To make sure Tango works well under different conditions, i've set up some load tests. These tests are there to simulate various scenarios and see how the system handles them.

Running Load Tests

You can run the load tests using this command:

make test-load

This will execute all the test cases in the loadtest package.

Configuring Load Tests

The load testing framework is flexible. You can adjust various parameters like the number of requests, concurrency level, duration, and more through a configuration struct:

type Config struct {
    Requests           int
    ConcurrentRequests int
    Duration           time.Duration
    StatsInterval      time.Duration
    RequestsPerSecond  float64
    Scenario           string
}

Example Scenarios

I have a few predefined scenarios to simulate different types of workloads:

  • High Concurrency: Simulates many concurrent requests with a specified rate limit.
  • Host Heavy: Focuses on creating matches where most players are hosts.

About

A matchmaking service written in Go

Resources

License

Stars

Watchers

Forks

Releases

No releases published