Building Event Sourced Apps

Designing your software around a story!

What is Event Sourcing?

Instead of defining a global data model that fits all of your use cases think about the things your system does. Record the events that occur and then build multiple data models that fit your individual use cases.

Sounds very abstract? In this talk, Leif will elaborate on some side projects including use cases with the hope to encourage you to switch to more event-centric thinking.

Transcript

1. Building Event Sourced
   Apps
   

   View Slide

2. View Slide

3. View Slide

4. My Example
   

   View Slide

5. Every team plays every other team twice
   306 Matches
   

   View Slide

6. Home Team Away Team
   +3 +0
   +1 +1
   

   View Slide

7. View Slide

8. View Slide

9. Transformation
   Input
   Output
   ?
   

   View Slide

10. https://www.kaggle.com/mkhvalchik/soccer/data
    

    View Slide

11. Match
    Team1
    Team2
    League SoSo Soccer
    

    View Slide

12. sum(m.away_team_points) as points
    from
    (select *,
    case
    when away_team_goal > home_team_goal then 3
    when home_team_goal = away_team_goal then 1
    else 0
    end as away_team_points,
    case
    when away_team_goal > home_team_goal then 1
    else 0
    end as away_team_wins,
    case
    when away_team_goal = home_team_goal then 1
    else 0
    end as away_team_draws,
    case
    when away_team_goal < home_team_goal then 1
    else 0
    end as away_team_losses
    from matches) m
    join leagues l on m.league_id = l.id
    group by season,
    away_team_api_id) t
    join leagues l on t.league_id = l.id
    join teams te on te.api_id = t.team_id
    group by (t.team_id,
    t.season)
    order by points desc,
    goal_difference desc,
    goals_for desc;
    
    

    View Slide

13. def standings(season, league_id)
    Repo.all(m in Match, where: m.season == ^season and m.league_id == ^league_id)
    |> Map.merge(%{row.home_team_api_id => home_team_values(row) }, &add_up/3)
    |> Map.merge(%{row.away_team_api_id => away_team_values(row) }, &add_up/3)
    |> Enum.sort_by(fn s ->
    {s.points, s.goal_difference, s.goals_for}
    end, &>=/2)
    end
    defp home_team_values(row) do
    %{
    games: 1,
    wins: Rules.wins(row.home_team_goal, row.away_team_goal),
    draws: Rules.draws(row.home_team_goal, row.away_team_goal),
    losses: Rules.losses(row.home_team_goal, row.away_team_goal),
    goals_for: row.home_team_goal,
    goals_against: row.away_team_goal,
    points: Rules.points(row.home_team_goal, row.away_team_goal)
    }
    end
    defp add_up(_k, nil, v2), do: v2
    defp add_up(_k, v1, v2) when is_integer(v1), do: v1 + v2
    defp add_up(_k, v1, v2) when is_map(v1), do: Map.merge(v1, v2, &add_up/3)
    

    View Slide

14. Problem?
    

    View Slide

15. Simplicity?
    

    View Slide

16. Performance?
    

    View Slide

17. Change!
    

    View Slide

18. T1
    T2
    T3
    

    View Slide

19. There are only two
    hard things in
    Computer Science
    …
    

    View Slide

20. Rethink
    

    View Slide

21. It’s a Tool
    

    View Slide

22. It’s a Tool that
    helps you capture
    stuff that
    happened
    

    View Slide

23. MatchStarted
    MatchStarted
    MatchStarted
    TeamFounded
    TeamFounded
    TeamFounded
    TeamFounded
    TeamFounded
    TeamQualified
    TeamQualified
    TeamQualified
    MachEnded
    MachEnded
    MachEnded
    SeasonEnded
    SeasonStarted
    

    View Slide

24. View Slide

25. project %TeamQualifiedForSeason{} = qualified do
    Ecto.Multi.insert(multi, :standings, %Standing{
    season_id: qualified.season_id,
    team_api_id: qualified.team_api_id,
    team_long_name: qualified.team_long_name,
    league_id: qualified.league_id,
    sort_key: sort_key(0, 0, 0)
    })
    end
    project %MatchEnded{} = ended do
    season_id = id_from_stream_id(ended.season_id)
    home_team = Standing.by_team_and_season(ended.home_team_api_id, season_id)
    away_team = Standing.by_team_and_season(ended.away_team_api_id, season_id)
    home_team_diff = team_changeset(home_team, ended.home_team_goal, ended.away_team_goal)
    away_team_diff = team_changeset(away_team, ended.away_team_goal, ended.home_team_goal)
    multi |> Ecto.Multi.update(:home_team, home_team_changeset)
    |> Ecto.Multi.update(:away_team, away_team_changeset)
    end
    

    View Slide

26. View Slide

27. (Re)Tell
    (Re)Tell
    (Re)Tell
    Capture
    https://github.com/leifg/so_so_soccer
    

    View Slide

28. It’s a Tool that
    helps you capture
    events …
    

    View Slide

29. … and tell a story
    from them
    

    View Slide

30. “But I don’t like Football”
    

    View Slide

31. Let’s get some Calories
    

    View Slide

32. Demo Time
    

    View Slide

33. TrackMeal Meal MealTracked
    Command Aggregate Event
    ?
    

    View Slide

34. /api/do_stuff
    Command
    
    

    View Slide

35. /meals
    TrackMeal
    Frontend
    MealTracked
    Phx Channel
    

    View Slide

36. Hacker News Vanity
    

    View Slide

37. HnStory
    •StorySubmitted
    •StoryVotedOn
    •StoryHitFrontPage
    •StoryFrontPagePositionChanged
    

    View Slide

38. View Slide

39. View Slide

40. View Slide

41. Offer
    Price: £1000
    Buyer: E-Corp
    Status: “accepted”
    SELECT SUM(price) FROM offers
    WHERE status = ‘accepted’
    GROUP BY buyer_id;
    Offer
    Price: £1000
    Buyer: E-Corp
    Status: “paid”
    SELECT SUM(price) FROM offers
    WHERE status in (‘accepted’,
    ‘paid’) GROUP BY buyer_id;
    

    View Slide

42. Offer
    Price: £1000
    Buyer: E-Corp
    Status: “accepted”
    SELECT SUM(price) FROM offers
    WHERE status = ‘accepted’
    GROUP BY buyer_id;
    Offer
    Price: £1000
    Buyer: E-Corp
    Status: “rejected”
    

    View Slide

43. Event Sourcing is Great for Side Projects
    It’s also Great for Real World Applications
    leif.io
    If you know the business
    bit.ly/buildingconduit
    

    View Slide