jackonelli / wednesday-world-cup Goto Github PK

The current migrations create table games before table teams, forcing the removal on foreign keys dependencies of games to teams.

I think the easiest way to do this is to remove all migrations and re-add them in correct order.

• Struct in core
• Db table
• CLI for add/remove
• UI reads default better's bet and updates on input.

When saving preds, the preds coming from the UI is the official version of the truth. Any conflicting pred in the database should be updated. Currently we just keep adding predictions.

UI currently lists first played, then unplayed games. It would be nice to group them together, sorted on date or whatever.

Learn how to benchmark, did it a bit for AOC but learn properly.

Follow the set of rules proposed here: https://rust-lang.github.io/api-guidelines/

Especially for core

Low hanging fruit are sections documentation and naming.

Follow diesel conventions for table names and keys, foreign keys. Make life simpler.

• group_game_map --> group_game_maps
• random foreign key names: id for prim. key, table name for foreign key (but singular.), e.g. foreign key id in table games should be game_id

Reminder to properly set up a license. I would like it to have GPL, but that seems scary.

Port this (or some other stuff to the seed code)

https://codepen.io/jimmyhayek/pen/yJkdEB/

Implementation of playoff logic.

• Populate first round from groups, nb. must be able to handle partially finished groups. No, the order is strict even with no games played. In practice, we will always have a group order, although it may be random (Fifa rules) or deterministic, indep. of played games (Uefa ranking).
• Playoff representation (List of lists? Tree? Linked list?)
• Play/unplay games.
• Propagate result, i.e. populate the next round with the winners from the previous.

Not a bug technically, the group order can be random. But it is a bit flickery when the list shifts as you enter scores.

Really any function that pre-allocates a team struct could be made this way instead.

Original source code comment for this trait.

This could perhaps be more efficient with the cool group by
https://docs.rs/itertools/0.10.0/itertools/trait.Itertools.html#method.into_grouping_map
i.e. don't fold but map all games to (team_id, stat), then eff. fold to team stats map.

Beware though:

• All team id's must be present in the final map. If no played games involve a team it will not appear in the map. This extra check might negate the perf. increase.
• Must benchmark.

Sources

• https://seed-rs.org/
• https://github.com/seed-rs/seed-quickstart-webpack
• https://erwabook.com/intro/index.html
• https://github.com/seed-rs/seed-rs-realworld

This type of application is ideal for fuzzing since we have so many invariants that we (believe) are correctly encoded in the type system.
This blog post:
https://tiemoko.com/blog/diff-fuzz/
is a nice introduction to fuzzing and I would like to try it out, as a learning exercise.

In the module core/src/pred_score.rs I have defined a simple version of the prediction score.
This module is pedagogically commented to explain all the rust quirks that go in to creating another score,
with the intent of making this a good first issue if someone else wants a crack at it.

It's mostly for this intro purpose that a new score should be created; it's not necessary for a first working app but it's simple, isolated and well documented.
In the end, it's still useful since one of the motivations for doing this is to have more control over the scoring process.

Remove all panics except the non-reachable ones, which instead should be documented.

When doing a POST command (save_preds) in the UI the server first routes to a OPTIONS route, fails that and then does the post. Doesn't seem to have any impact but it is annoying.

The input basically follows a regex: "\d+-\d+" which evaluates greedily.
Having written "9-", if the next character is a digit, the page is uploaded with the score.
Basically, you cannot give the away team a two-digit goal count. You can work around it by adding a dummy non-digit and the prepend with the correct score. Silly, but it works.