commits
move to using `ConfigValue` which acts like a `LazyLock`
but supports async initializers (mostly)
also rename INDEX_USER to USER_DID and add support for
overriding the PDS url, repo export url, and (later)
the subscribe repos url
this should prepare us for backfill later and allow strongrefs etc bc thats just like. sane lol
also remove stupid car rev nonsense bc lowk was stupid lol
its only on startup the backfill overhead is NAWT that high (like one second to parse and save on my pc + only on startup its _nawt_ that bad. bottleneck is network atp)
optomizations to reduce downloaded car size (mst diffs) may be used later but i cba rn lol
i also need to figure out some rev logic to dedupe repo ops :3
theres no other functionality except the basic car backfill lol
- remove sqlx cache bc i aint usin it rn (will add it back l8r for docker build)
- document how im going to do backfill/etc
- add meta table with a rev column. this is to store repo rev for diffing/etc
we use `cargo sqlx prepare` with offline mode as the database can be offline when docker starts, which leads to builds failing
move to using `ConfigValue` which acts like a `LazyLock`
but supports async initializers (mostly)
also rename INDEX_USER to USER_DID and add support for
overriding the PDS url, repo export url, and (later)
the subscribe repos url
also remove stupid car rev nonsense bc lowk was stupid lol
its only on startup the backfill overhead is NAWT that high (like one second to parse and save on my pc + only on startup its _nawt_ that bad. bottleneck is network atp)
optomizations to reduce downloaded car size (mst diffs) may be used later but i cba rn lol
i also need to figure out some rev logic to dedupe repo ops :3