asobi_ops_matchmaker (asobi v0.72.6)

View Source

The matchmaking queue for the ops plane: one row per mode.

The rows come from asobi_matchmaker:snapshot/0, an ETS read that never touches the matchmaker's mailbox - the reason is on that function, and it is the whole design of this endpoint.

No ticket, player or property ever appears here. An operator asking about the queue is asking whether a mode is filling, which is a count and two waits; who is waiting is player data and belongs behind a capability, not behind the same bearer token the game client already holds.

Summary

Functions

project(Row)

Positive allowlist of the fields a queue row may carry off this endpoint.

rows/2

The per-mode rows and the order to page them in.

sortable()

Wire-name to column mapping this endpoint accepts in ?sort=.

summary/1

Queue-wide totals, reported alongside the page.

Functions

project(Row)

-spec project(map()) -> map().

Positive allowlist of the fields a queue row may carry off this endpoint.

rows/2

-spec rows(asobi_matchmaker:snapshot(), asobi_ops_params:params()) ->
              {ok, {[map()], asobi_ops_params:sort_spec()}} |
              {error, {unknown_sort, binary()} | {unknown_order, binary()}}.

The per-mode rows and the order to page them in.

Deepest queue first, since that is the mode in trouble. mode is unique across the rows and ends the order, so the offset window cannot repeat a row.

sortable()

-spec sortable() -> asobi_ops_params:sort_allowlist().

Wire-name to column mapping this endpoint accepts in ?sort=.

summary/1

-spec summary(asobi_matchmaker:snapshot()) -> map().

Queue-wide totals, reported alongside the page.

sampled_at and age_ms are part of the answer, not decoration: the counts are as old as the last matchmaker tick, and a console that renders them as live would show a queue that never quite settles.