Commit Graph

453 Commits (f72eabd42aa32426c4db087ac325ace26bf5cde5)

Author SHA1 Message Date
Tyler Goodlet 1ab495a64d Map broken stream errs to transport closed; msgspec seems to be racy 2021-10-05 13:37:17 -04:00
Tyler Goodlet 562419c907 Convert actor UIDs to hashable tuples
`msgspec` sends python lists over the wire
(https://github.com/jcrist/msgspec/issues/30) which is fine and dandy
but we use them as lookup keys so we need to be sure we tuple-cast
first.
2021-10-05 13:37:17 -04:00
Tyler Goodlet 3facfb6d4c Fix log levels 2021-10-05 13:37:17 -04:00
Tyler Goodlet aa080543d0 Mypy fixes to enforce uid tuple 2021-10-05 13:37:17 -04:00
Tyler Goodlet b64396f708 Pkg `msgpec` as optional dep, load transport type if importable 2021-10-05 13:37:17 -04:00
Tyler Goodlet 96b3f94c72 Accept transport closed error during handshake and msg loop 2021-10-05 13:37:17 -04:00
Tyler Goodlet ecd8c4bc7e Drop happy eyeballs inf delay 2021-10-05 13:37:17 -04:00
Tyler Goodlet 112117c1fc Add our own "transport closed" signal
This change some super old (and bad) code from the project's very early
days. For some redic reason i must have thought masking `trio`'s
internal stream / transport errors and a TCP EOF as `StopAsyncIteration`
somehow a good idea. The reality is you probably
want to know the difference between an unexpected transport error
and a simple EOF lol. This begins to resolve that by adding our own
special `TransportClosed` error to signal the "graceful" termination of
a channel's underlying transport. Oh, and this builds on the `msgspec`
integration which helped shed light on the core issues here B)
2021-10-05 13:37:17 -04:00
Tyler Goodlet 95e35f3d60 Add streaming decode support for `msgspec`
Add a `tractor._ipc.MsgspecStream` type which can be swapped in for
`msgspec` serialization transparently. A small msg-length-prefix framing
is implemented as part of the type and we use
`tricycle.BufferedReceieveStream` to handle buffering logic for the
underlying transport.

Notes:
- had to force cast a few more list  -> tuple spots due to no native
  `tuple`decode-by-default in `msgspec`: https://github.com/jcrist/msgspec/issues/30
- the framing can be understood by this protobuf walkthrough:
  https://eli.thegreenplace.net/2011/08/02/length-prefix-framing-for-protocol-buffers
- `tricycle` becomes a new dependency
2021-10-05 13:37:17 -04:00
Tyler Goodlet e39ee3a9cc Always cast arbiter addr to tuple 2021-10-05 13:37:17 -04:00
Tyler Goodlet dda0b22870 Try out `msgspec` in our msgpack stream channel
Can only really use an encoder currently since there is no streaming api
in `msgspec` as of currently. See jcrist/msgspec#27.

Not sure if any encoding speedups are currently noticeable especially
without any validation going on yet XD.

First experiments toward #196
2021-10-05 13:37:17 -04:00
Tyler Goodlet 4079f02acf Cast to tuples for all uids explicitly 2021-10-05 13:37:17 -04:00
Tyler Goodlet 518a0d5e14 Add todo for log msg filename.. 2021-10-04 10:38:44 -04:00
Tyler Goodlet 8b416e6bba Stream and context api tweaks
- drop `shield` input to `MsgStream`
- check for cancel called prior to loading the feeder mem chan
  in `Context.open_stream()`
- warn on a timeout when trying to cancel a remote task from
  `Context.cancel()`
- drop noop endofchannel handler block
2021-10-04 10:38:44 -04:00
Tyler Goodlet bd31f47d5f Handle kbi in ctx blocks via `BaseException`
Fixes prior committed tests by more generally handling `BaseExcepion` in
context blocks. Left in the commented concrete list for reference.
2021-10-04 10:38:44 -04:00
Tyler Goodlet 4259738864 Flip to using the `trio` spawner on windows
Was able to try it manually on a windows 10 system and the debugger
works great!
2021-09-18 14:10:32 -04:00
Tyler Goodlet 1137a9e7ac Fix 404ed tokio urls 2021-09-02 21:12:54 -04:00
Tyler Goodlet 2745a2b1dc Solve first-recv-cancelled by recursive `.receive()` on wake 2021-09-02 21:12:54 -04:00
Tyler Goodlet d9bb52fe7b Store array `maxlen` in state singleton
The `collections.deque` takes care of array length truncation of values
for us implicitly but in the future we'll likely want this value exposed
to alternate array implementations. This patch is to provide for that as
well as make `mypy` happy since the `dequeu.maxlen` can also be `None`.
2021-09-02 21:12:54 -04:00
Tyler Goodlet 9258f79510 Don't wake sibling bcast consumers on a cancelled call 2021-09-02 21:12:54 -04:00
Tyler Goodlet 44ef26bb18 Shorten default feeder mem chan size to 64 2021-09-02 21:12:54 -04:00
Tyler Goodlet 7857a9ac6d Add `shield: bool` kwarg to `Portal.open_stream_from()` 2021-09-02 21:12:54 -04:00
Tyler Goodlet 63ec740e27 Add some bcaster ref sanity asserts around subscriptions 2021-09-02 21:12:54 -04:00
Tyler Goodlet 093e7d921c Instance ids are ints 2021-09-02 21:12:54 -04:00
Tyler Goodlet bec3f5999d Drop uuid4 keys, raise closed error on subscription after close 2021-09-02 21:12:54 -04:00
Tyler Goodlet a4cb0ef21f Fix `.receive()` re-assignment, drop `.clone()` 2021-09-02 21:12:54 -04:00
Tyler Goodlet 346b5d2eda Blade runner it
Get rid of all the (requirements for) clones of the underlying
receivable. We can just use a uuid generated key for each instance
(thinking now this can probably just be `id(self)`). I'm fully convinced
now that channel cloning is only a source of confusion and anti-patterns
when we already have nurseries to define resource lifetimes. There is no
benefit in particular when you allocate subscriptions using a context
manager (not sure why `trio.open_memory_channel()` doesn't enforce
this).

Further refinements:
- add a `._closed` state that will error the receiver on reuse
- drop module script section;  it's been moved to a real test
- call the "receiver" duck-type stub a new name
2021-09-02 21:12:54 -04:00
Tyler Goodlet 6c17c7367a Store handle to underlying channel's `.receive()`
This allows for wrapping an existing stream by re-assigning its receive
method to the allocated broadcaster's `.receive()` so as to avoid
expecting any original consumer(s) of the stream to now know about the
broadcaster; this instead mutates the stream to delegate to the new
receive call behind the scenes any time `.subscribe()` is called.

Add a `typing.Protocol` for so called "cloneable channels" until we
decide/figure out a better keying system for each subscription and
mask all undesired typing failures.
2021-09-02 21:12:54 -04:00
Tyler Goodlet 2d1c24112b Add subscription support to message streams
Add `ReceiveMsgStream.subscribe()` which allows allocating a broadcast
receiver around the stream for use by multiple actor-local consumer
tasks. Entering this context manager idempotently mutates the stream's
receive machinery which for now can not be undone. Move `.clone()` to
the receive stream type.

Resolves #204
2021-09-02 21:12:54 -04:00
Tyler Goodlet a12b1fc631 Drop optimization check, binance made its point 2021-09-02 21:12:54 -04:00
Tyler Goodlet ceed96aa3f Add common state delegate type for all consumers
For every set of broadcast receivers which pull from the same producer,
we need a singleton state for all of,
- subscriptions
- the sender ready event
- the queue

Add a `BroadcastState` dataclass for this and pass it to all
subscriptions. This makes the design much more like the built-in memory
channels which do something very similar with `MemoryChannelState`.

Use a `filter()` on the subs list in the sequence update step, plus some
other commented approaches we can try for speed.
2021-09-02 21:12:54 -04:00
Tyler Goodlet 6e78bcf898 Facepalm: use single `_subs` per clone set 2021-09-02 21:12:54 -04:00
Tyler Goodlet 4ad75a3287 Obviously keying on tasks isn't going to work
Using the current task as a subscription key fails horribly as soon as
you hand off new subscription receiver to another task you've spawned..

Instead use the underlying ``trio.abc.ReceiveChannel.clone()`` as a key
(so i guess we're assuming cloning is supported by the underlying?)
which makes this all work just like default mem chans. As a bonus, now
we can just close the underlying rx (which may be a clone) on
`.aclose()` and everything should just work in terms of the underlying
channels lifetime (i think?).

Change `.subscribe()` to be async since the receive channel type
interface only expects `.aclose()` and it actually ends up being
nicer for 3.9+ style `async with` parentheses style anyway.
2021-09-02 21:12:54 -04:00
Tyler Goodlet 64358f6525 Rename to broadcast mod, don't expect mem chan specifically 2021-09-02 21:12:54 -04:00
Tyler Goodlet 1af7dbb732 `Task` is hashable, so key on it 2021-09-02 21:12:54 -04:00
Tyler Goodlet 6a2c3da1bb Simplify api around receive channel
Buncha improvements:
- pass in the queue via constructor
- tracking over all underlying memory channel closure using cloning
- do it like `tokio` and set lagged consumers to the last sequence
  before raising
- copy the subs on first receiver wakeup for iteration instead of
  iterating the table directly (and being forced to skip the current
  tasks sequence increment)
- implement `.aclose()` to close the underlying clone for this task
- make `broadcast_receiver()` just take the recv chan since it doesn't
  need anything on the send side.
2021-09-02 21:12:54 -04:00
Tyler Goodlet 3817b4fb5e Ultra naive broadcast channel prototype 2021-09-02 21:12:54 -04:00
Tyler Goodlet 3208b67f57 Drop shielding on root lock acquire; seems to prevent hangs 2021-09-02 16:23:38 -04:00
Tyler Goodlet 61d2307e52 Unlock pdb tty on all possible net faults 2021-09-02 16:23:38 -04:00
Tyler Goodlet 79f0d6fda0 Attempt to avoid pdb lockups on channel breakage
Always try to release the root tty lock on broken connection errors.
2021-09-02 16:23:10 -04:00
Tyler Goodlet 4f166500d0 Add return type to debugger factory 2021-09-02 16:22:59 -04:00
Tyler Goodlet d906c81f14 Export portal type at top level 2021-09-02 16:22:59 -04:00
Tyler Goodlet 68d56d5df0 Try not masking SIGINT in child processes 2021-09-02 16:22:59 -04:00
Tyler Goodlet 497fa72c96 Add a SIGINT handler that kills the process tree
We're not actually using this but it's for reference if we do end up
needing it.

The std lib's `pdb` internals override SIGINT handling whenever one
enters the debugger repl. Force a handler that kills the tree if SIGINT
is triggered from the root actor, otherwise ignore it since supervised
children should be managed already. This resolves an issue with guest
mode where `pdb` causes SIGINTs to be swallowed resulting in the host
loop never terminating the process tree.
2021-09-02 16:22:02 -04:00
Tyler Goodlet b4d95e9543 Update docs to new close semantics 2021-09-02 08:24:18 -04:00
Tyler Goodlet af85d35685 Drop stream shielding; it was from a legacy design
The whole origin was not having an explicit open/close semantic for
streams. We have that now so this internal mechanic isn't needed and
further our streams become more correct by having `.aclose()` be
independent of cancellation.
2021-09-02 08:24:18 -04:00
Tyler Goodlet 7431e8ea01 Don't log cancelled inceptions seen by the root 2021-08-02 21:15:42 -04:00
Tyler Goodlet 82999801a6 Drop leftover noisy exception logging.. 2021-08-02 16:56:00 -04:00
Tyler Goodlet 674fbbc6b3 Docs and comments tidying 2021-08-01 10:44:13 -04:00
Tyler Goodlet 6006adc0de Hide `_invoke()` tb, move actor error to exceptions mod 2021-07-31 13:56:26 -04:00
Tyler Goodlet 0afa7f0f8e Fix lock context manager return type 2021-07-31 12:50:58 -04:00
Tyler Goodlet b3d28a1ee4 Drop debugger path and duplicate func from rebasing 2021-07-31 12:46:40 -04:00
Tyler Goodlet 09f00a5a00 Go back to only logging tbs on no debugger 2021-07-31 12:46:40 -04:00
Tyler Goodlet 44bfacc0c2 Comment hard-kill-sidestep for now since nursery version covers it? 2021-07-31 12:46:40 -04:00
Tyler Goodlet 551816e80d Solve the root-cancels-child-in-tty-lock race
Finally this makes a cancelled root actor nursery not clobber child
tasks which request and lock the root's tty for the debugger repl.

Using an edge triggered event which is set after all fifo-lock-queued
tasks are complete, we can be sure that no lingering child tasks are
going to get interrupted during pdb use and tty lock acquisition.
Further, even if new tasks do queue up to get the lock, the root will
incrementally send cancel msgs to each sub-actor only once the tty is
not locked by a (set of) child request task(s). Add shielding around all
the critical sections where the child attempts to allocate the lock from
the root such that it won't be disrupted from cancel messages from the
root after the acquire lock transaction has started.
2021-07-31 12:46:40 -04:00
Tyler Goodlet be1fcb2a5b Distinguish between a local pdb unlock and the tty unlock in root 2021-07-31 12:46:40 -04:00
Tyler Goodlet ef89ed947a Fix hard kill in debug mode; only do it when debug lock is empty 2021-07-31 12:46:40 -04:00
Tyler Goodlet 5b3894827f Move some infos to runtime level 2021-07-31 12:46:40 -04:00
Tyler Goodlet 0fdcfa0ba1 Move debugger wait inside OCA nursery 2021-07-31 12:46:40 -04:00
Tyler Goodlet 37a1897c47 Don't shield debugger status wait; it causes hangs 2021-07-31 12:46:40 -04:00
Tyler Goodlet 0f2a39a311 Catch and delay errors in the root if debugger is active 2021-07-31 12:46:40 -04:00
Tyler Goodlet 23a1622256 Don't kill root's immediate children when in debug
If the root calls `trio.Process.kill()` on immediate child proc teardown
when the child is using pdb, we can get stdstreams clobbering that
results in a pdb++ repl where the user can't see what's been typed. Not
killing such children on cancellation / error seems to resolve this
issue whilst still giving reliable termination. For now, code that
special path until a time it becomes a problem for ensuring zombie
reaps.
2021-07-31 12:46:40 -04:00
Tyler Goodlet 49d439b681 Add some brief todo notes on idea of shielded breakpoint 2021-07-31 12:46:40 -04:00
Tyler Goodlet 6f05f5d5e6 Wait for debugger lock task context termination 2021-07-31 12:46:40 -04:00
Tyler Goodlet b369b91357 Fix up var naming and typing 2021-07-31 12:46:40 -04:00
Tyler Goodlet 969bce3aa4 Use context for remote debugger locking
A context is the natural fit (vs. a receive stream) for locking the root
proc's tty usage via it's `.started()` sync point. Simplify the
`_breakpoin()` routine to be a simple async func instead of all this
"returning a coroutine" stuff from before we decided that
`tractor.breakpoint()` must be async. Use `runtime` level for locking
logging making it easier to trace.
2021-07-31 12:46:40 -04:00
Tyler Goodlet 443ebea165 Use "pdb" level logging in debug mode 2021-07-08 13:02:33 -04:00
Tyler Goodlet 25779d48a8 Define explicit adapter level methods for mypy 2021-07-08 12:51:35 -04:00
Tyler Goodlet fde52d2464 Mypy fixes 2021-07-08 12:48:34 -04:00
Tyler Goodlet 8c927d708d Change trace to transport level 2021-07-07 14:31:15 -04:00
Tyler Goodlet 31590e82a3 Flip "trace" level to "transport" level logging 2021-07-07 14:31:03 -04:00
Tyler Goodlet 2513c652c1 Go back to only logging crashes if no pdb gets engaged 2021-07-06 08:23:30 -04:00
Tyler Goodlet 9ddb654783 Avoid mutate on iterate race 2021-07-06 08:23:30 -04:00
Tyler Goodlet 7f86d63e77 Drop trip kwarg 2021-07-06 08:23:30 -04:00
Tyler Goodlet 12f987514d Don't enter debug on closed resource errors 2021-07-06 08:23:30 -04:00
Tyler Goodlet 98bbf8e0df Move join event trigger to direct exit path 2021-07-06 08:23:30 -04:00
Tyler Goodlet b1cd7fdedf Don't shield on root cancel it can causes hangs 2021-07-06 08:23:30 -04:00
Tyler Goodlet ef725c5972 Always hard kill sub-procs on teardown
Adds a new hard kill routine for the `trio` spawning backend.
2021-07-06 08:23:30 -04:00
Tyler Goodlet b21e2a6caa Add pre-stream open error conditions 2021-07-06 08:23:30 -04:00
Tyler Goodlet c6cdaf9c31 De-densify some code 2021-07-06 08:23:30 -04:00
Tyler Goodlet 91640facbc Always shield cancel the caller on cancel-causing-errors, add teardown logging 2021-07-06 08:23:30 -04:00
Tyler Goodlet c2484e88a1 First try: pack cancelled tracebacks and ship to caller 2021-07-06 08:23:30 -04:00
Tyler Goodlet 3423ea4011 Add temp warning msg for context cancel call 2021-07-06 08:23:29 -04:00
Tyler Goodlet af701c16ee Consider relaying context error via raised-in-scope-nursery task 2021-07-06 08:23:29 -04:00
Tyler Goodlet 1703171bea Set stream "end of channel" after shielded check!
Another face palm that was causing serious issues for code that is using
the `.shielded` feature..

Add a bunch more detailed comments for all this subtlety and hopefully
get it right once and for all. Also aggregated the `trio` errors that
should trigger closure inside `.aclose()`, hopefully that's right too.
2021-07-06 08:23:29 -04:00
Tyler Goodlet 3d633408fc Don't clobber msg loop mem chan on rx stream close
Revert this change since it really is poking at internals and doesn't
make a lot of sense. If the context is going to be cancelled then the
msg loop will tear down the feed memory channel when ready, we don't
need to be clobbering it and confusing the runtime machinery lol.
2021-07-06 08:23:29 -04:00
Tyler Goodlet 196dea80db Drop trailing comma 2021-07-06 08:23:29 -04:00
Tyler Goodlet 54916be601 Adjustments for non-frozen context dataclass change 2021-07-06 08:23:29 -04:00
Tyler Goodlet 1a69727b75 Fix exception typing 2021-07-06 08:23:29 -04:00
Tyler Goodlet 348148ff1e Explicitly formalize context/streaming teardown
Add clear teardown semantics for `Context` such that the remote side
cancellation propagation happens only on error or if client code
explicitly requests it (either by exit flag to `Portal.open_context()`
or by manually calling `Context.cancel()`).  Add `Context.result()`
to wait on and capture the final result from a remote context function;
any lingering msg sequence will be consumed/discarded.

Changes in order to make this possible:
- pass the runtime msg loop's feeder receive channel in to the context
  on the calling (portal opening) side such that a final 'return' msg
  can be waited upon using `Context.result()` which delivers the final
  return value from the callee side `@tractor.context` async function.
- always await a final result from the target context function in
  `Portal.open_context()`'s `__aexit__()` if the context has not
  been (requested to be) cancelled by client code on block exit.
- add an internal `Context._cancel_called` for context "cancel
  requested" tracking (much like `trio`'s cancel scope).
- allow flagging a stream as terminated using an internal
  `._eoc` flag which will mark the stream as stopped for iteration.
- drop `StopAsyncIteration` catching in `.receive()`; it does
  nothing.
2021-07-06 08:23:29 -04:00
Tyler Goodlet 73302d9d16 Specially raise a `ContextCancelled` for a task-context rpc 2021-07-06 08:23:29 -04:00
Tyler Goodlet 409f7f0d5a Expose streaming components at top level 2021-07-06 08:23:29 -04:00
Tyler Goodlet eb3662f981 Add a specially handled `ContextCancelled` error 2021-07-06 08:23:29 -04:00
Tyler Goodlet 39b9896a62 Only close recv chan if we get a ref 2021-07-06 08:23:29 -04:00
Tyler Goodlet 9a4244b9a6 Support no arg to `Context.started()` like trio 2021-07-06 08:23:29 -04:00
Tyler Goodlet a2e2f7e7a8 Only send stop msg if not received from far end 2021-07-06 08:23:29 -04:00
Tyler Goodlet 6559fb72aa Expose msg stream types at top level 2021-07-06 08:23:29 -04:00
Tyler Goodlet e311430d25 Be more pedantic with error handling 2021-07-06 08:23:29 -04:00
Tyler Goodlet 08eb6bd019 Fix typing 2021-07-06 08:23:29 -04:00
Tyler Goodlet 1f8966ba64 Support passing `shield` at stream contruction 2021-07-06 08:23:29 -04:00