Flitterite

Fork

Make Gold.to_image foldable.

Okay! This was the missing piece. Now to restore the big patch...

Duplicate and simplify row-displaying logic.

Remove world-record support.

For posterity: This is the ability to take somebody else's segments and
store them in a way that doesn't associate them with any particular
attempt. I don't like this because it's incompatible with the rest of
the data model, and also breaks an assumption about provenance; but
also, I think that it's unhealthy to compare oneself to the
world-record-holding runners constantly.

Fix a SQL query and add cumulative times per attempt.

Make row generation slightly easier.

Factor the ability to fetch golds for all segments.

Well, all segments for a given game and category. This query was already
used to make summaries, and it happens to be exactly the data that we
want to display in the UI too.

Add convenience functions for SQL SELECT statements.

Two new modules for managing queried state.

Meh.

Encapsulate prepared statements, caching them.

SQLite doesn't automatically cache prepared statements, and the OCaml
wrapper doesn't do it for us. We want to feel comfortable calling into
this storage module multiple times per run, but we want the statement to
be prepared and cached for us without keeping it in scope. (We can keep
the SQL string in scope because it's a pre-buildable global constant.)

Remove unused custom summary code.

Remove two more gettimeofday().

Somewhat factor timer state.

I might have to just hack and slash for the next commit.

Finish hacking on summary tool.

All of the queries seem to report accurate data. Summation of segments
across categories is working; if you run a game that branches into many
categories, then statistics are accumulated across all categories.

Summarize entire categories segment-by-segment.

Tighten the data model; start at "New Game".

I don't have a good way to customize this, but at least it exists.

Start factoring SQLite access.

Also, reuse the existing Welford implementation. The SQLite user-defined
function bindings worked perfectly on the first try, and this was much
easier than requiring a special SQLite with extra extension functions.

Start rebuilding summary tool.

Haven't ported the statistics yet. Need sleeps first.

Migrate heterogenous splits correctly.

Implement the bulk of the migration.

I think I've got the data model nailed down. Aside from a couple
optional fields for datetimes, which we'll want in the future, this is
basically complete.

One of my savefiles doesn't import. It uses different split titles in
its attempts, due to rerouting. This is exactly the sort of thing I
wanted to preserve, so I'll fix it.

Fulfill the flitter-new command.

Clean up docs.

Implement the first half of the migration script.

Create games, checkpoints, and routes. Don't upsert routes; if there's
multiple routes possible, then we'll want a flag or something to confirm
to the user that we haven't already imported this particular savefile.

Clean up display padding.

Some of it is still nasty, but we're at least not reinventing
left-padding arithmetic any longer.

Don't show all splits if there's not room.

Instead, take only some splits.

Lift some constants.

Start building a SQLite migration tool.

Depend on sqlite3.

Become a Nix flake.

And we have to use 22.11 because LWT is broken on unstable.

Clean up handling of optional values.