commits
This branch wasn't quite ready yet. In particular, issues came up with
subscribeRepos, which is pretty load-bearing for the Go codebase.
This reverts commit 5b368e30b8dc6ccb8a9d0bcfafe0a519a40f6351, reversing
changes made to 7ad48e1c510fe3d4b683281b1e285887d2ea4b11.
This is at a point where the new lexgen code roughly reproduces the
current ("legacy") codegen behavior. This means it can replace the
previous `lex/*.go` implementation. This version is somewhat more
complete and correct about the lexicon language in general, not just the
features we have used in the `com.atproto.*` and `app.bsky.*` lexicons
to date.
Things I might try to hit before merging:
- [ ] add some serialization tests (`./testing/`): records with
strongRef, recordWithMedia, unions, etc
- [x] resolve some more XXX and TODO
For later follow-up:
- [ ] API server handlers (for query and procedure), and ability to
disable client method output
- [ ] alternative CBOR library support (eg, annotations for `go-dasl`)
- [ ] ability to configure more package mappings (not just the indigo
pages)
- [ ] string, boolean, and integer query params as pointers
(configurable?)
- [ ] more correctness around `$type` serialization
- [ ] validation methods (for limits, etc)
- [ ] option to break out large query param lists as a struct
This would supersede #682 and is based on the other existing ghcr
workflows. This was [inspired by a
question](https://discord.com/channels/1097580399187738645/1411924152297984081/1416794679222993047)
asking if bluesky's relay implementation was published publicly as a
docker image.
I was trying to follow dropped events, and realised this metric is not
being written.
Signed-off-by: claudecodering <claudecoder@outlook.com>
Includes types from:
- https://github.com/bluesky-social/atproto/pull/4297
- https://github.com/bluesky-social/atproto/pull/4317
Updates style of output code, but should have no behavioral or API
impact.
Relay is not starting because initial-seq-number is always 0, resulting
from typecast from int to int64 fails. Changed flag type of
initial-seq-number fom int to Int64.
default-account-limit and new-hosts-per-day-limit had the same issue
fixes #1176
these are some small changes *before* getting in a refactor of the
lexgen tool.
This is to get things sycnhronized with upstream `atproto` repo while
preparing for lexgen workflow updates.
Small fix to the SQLite carstore -- the first record in the cache will
have ID 0, there's no reason to ignore it. Naturally, this case happens
a lot in testing and stuff 'cause the first attempt to get a revision on
an empty database will fail
This reverts commit 5b368e30b8dc6ccb8a9d0bcfafe0a519a40f6351, reversing
changes made to 7ad48e1c510fe3d4b683281b1e285887d2ea4b11.
This is at a point where the new lexgen code roughly reproduces the
current ("legacy") codegen behavior. This means it can replace the
previous `lex/*.go` implementation. This version is somewhat more
complete and correct about the lexicon language in general, not just the
features we have used in the `com.atproto.*` and `app.bsky.*` lexicons
to date.
Things I might try to hit before merging:
- [ ] add some serialization tests (`./testing/`): records with
strongRef, recordWithMedia, unions, etc
- [x] resolve some more XXX and TODO
For later follow-up:
- [ ] API server handlers (for query and procedure), and ability to
disable client method output
- [ ] alternative CBOR library support (eg, annotations for `go-dasl`)
- [ ] ability to configure more package mappings (not just the indigo
pages)
- [ ] string, boolean, and integer query params as pointers
(configurable?)
- [ ] more correctness around `$type` serialization
- [ ] validation methods (for limits, etc)
- [ ] option to break out large query param lists as a struct