nim-chronos/chronos/internal/asyncfutures.nim

1634 lines
55 KiB
Nim
Raw Normal View History

2018-05-16 08:22:34 +00:00
#
# Chronos
2018-05-16 08:22:34 +00:00
#
2021-05-06 07:49:55 +00:00
# (c) Copyright 2015 Dominik Picheta
# (c) Copyright 2018-2023 Status Research & Development GmbH
2018-05-16 08:22:34 +00:00
#
# Licensed under either of
# Apache License, version 2.0, (LICENSE-APACHEv2)
# MIT license (LICENSE-MIT)
## Features and utilities for `Future` that integrate it with the dispatcher
## and the rest of the async machinery
2023-10-24 14:21:07 +00:00
{.push raises: [].}
Raise tracking (#251) * Exception tracking v2 * some fixes * Nim 1.2 compat * simpler things * Fixes for libp2p * Fixes for strictException * better await exception check * Fix for template async proc * make async work with procTy * FuturEx is now a ref object type * add tests * update test * update readme * Switch to asyncraises pragma * Address tests review comments * Rename FuturEx to RaiseTrackingFuture * Fix typo * Split asyncraises into async, asyncraises * Add -d:chronosWarnMissingRaises * Add comment to RaiseTrackingFuture * Allow standalone asyncraises * CheckedFuture.fail type checking * First cleanup * Remove useless line * Review comments * nimble: Remove #head from unittest2 * Remove implict raises: CancelledError * Move checkFutureExceptions to asyncfutures2 * Small refacto * small cleanup * Complete in closure finally * cleanup tests, add comment * bump * chronos is not compatible with nim 1.2 anymore * re-add readme modifications * fix special exception handlers * also propagate excetion type in `read` * `RaiseTrackingFuture` -> `InternalRaisesFuture` Use internal naming scheme for RTF (this type should only be accessed via asyncraises) * use `internalError` for error reading * oops * 2.0 workarounds * again * remove try/finally for non-raising functions * Revert "remove try/finally for non-raising functions" This reverts commit 86bfeb5c972ef379a3bd34e4a16cd158a7455721. `finally` is needed if code returns early :/ * fixes * avoid exposing `newInternalRaisesFuture` in manual macro code * avoid unnecessary codegen for `Future[void]` * avoid reduntant block around async proc body * simplify body generation for forward declarations with comment but no body * avoid duplicate `gcsafe` annotiations * line info for return at end of async proc * expand tests * fix comments, add defer test --------- Co-authored-by: Jacek Sieka <jacek@status.im>
2023-10-17 12:18:14 +00:00
import std/[sequtils, macros]
import stew/base10
2023-10-24 14:21:07 +00:00
import ./[asyncengine, raisesfutures]
import ../[config, futures]
export
raisesfutures.Raising, raisesfutures.InternalRaisesFuture,
raisesfutures.init, raisesfutures.error, raisesfutures.readError
2023-10-24 14:21:07 +00:00
when chronosStackTrace:
import std/strutils
when defined(nimHasStacktracesModule):
import system/stacktraces
else:
const
reraisedFromBegin = -10
reraisedFromEnd = -100
template LocCreateIndex*: auto {.deprecated: "LocationKind.Create".} =
LocationKind.Create
template LocFinishIndex*: auto {.deprecated: "LocationKind.Finish".} =
LocationKind.Finish
template LocCompleteIndex*: untyped {.deprecated: "LocationKind.Finish".} =
LocationKind.Finish
2021-05-06 07:49:55 +00:00
func `[]`*(loc: array[LocationKind, ptr SrcLoc], v: int): ptr SrcLoc {.
deprecated: "use LocationKind".} =
case v
of 0: loc[LocationKind.Create]
of 1: loc[LocationKind.Finish]
else: raiseAssert("Unknown source location " & $v)
type
FutureStr*[T] = ref object of Future[T]
## Deprecated
gcholder*: string
FutureSeq*[A, B] = ref object of Future[A]
## Deprecated
gcholder*: seq[B]
FuturePendingError* = object of FutureError
## Error raised when trying to `read` a Future that is still pending
FutureCompletedError* = object of FutureError
## Error raised when trying access the error of a completed Future
2023-10-24 14:21:07 +00:00
SomeFuture = Future|InternalRaisesFuture
func raiseFuturePendingError(fut: FutureBase) {.
noinline, noreturn, raises: FuturePendingError.} =
raise (ref FuturePendingError)(msg: "Future is still pending", future: fut)
func raiseFutureCompletedError(fut: FutureBase) {.
noinline, noreturn, raises: FutureCompletedError.} =
raise (ref FutureCompletedError)(
msg: "Future is completed, cannot read error", future: fut)
# Backwards compatibility for old FutureState name
template Finished* {.deprecated: "Use Completed instead".} = Completed
template Finished*(T: type FutureState): FutureState {.
deprecated: "Use FutureState.Completed instead".} =
FutureState.Completed
proc newFutureImpl[T](loc: ptr SrcLoc): Future[T] =
let fut = Future[T]()
internalInitFutureBase(fut, loc, FutureState.Pending, {})
fut
proc newFutureImpl[T](loc: ptr SrcLoc, flags: FutureFlags): Future[T] =
let fut = Future[T]()
internalInitFutureBase(fut, loc, FutureState.Pending, flags)
fut
2023-10-24 14:21:07 +00:00
proc newInternalRaisesFutureImpl[T, E](
loc: ptr SrcLoc): InternalRaisesFuture[T, E] =
Raise tracking (#251) * Exception tracking v2 * some fixes * Nim 1.2 compat * simpler things * Fixes for libp2p * Fixes for strictException * better await exception check * Fix for template async proc * make async work with procTy * FuturEx is now a ref object type * add tests * update test * update readme * Switch to asyncraises pragma * Address tests review comments * Rename FuturEx to RaiseTrackingFuture * Fix typo * Split asyncraises into async, asyncraises * Add -d:chronosWarnMissingRaises * Add comment to RaiseTrackingFuture * Allow standalone asyncraises * CheckedFuture.fail type checking * First cleanup * Remove useless line * Review comments * nimble: Remove #head from unittest2 * Remove implict raises: CancelledError * Move checkFutureExceptions to asyncfutures2 * Small refacto * small cleanup * Complete in closure finally * cleanup tests, add comment * bump * chronos is not compatible with nim 1.2 anymore * re-add readme modifications * fix special exception handlers * also propagate excetion type in `read` * `RaiseTrackingFuture` -> `InternalRaisesFuture` Use internal naming scheme for RTF (this type should only be accessed via asyncraises) * use `internalError` for error reading * oops * 2.0 workarounds * again * remove try/finally for non-raising functions * Revert "remove try/finally for non-raising functions" This reverts commit 86bfeb5c972ef379a3bd34e4a16cd158a7455721. `finally` is needed if code returns early :/ * fixes * avoid exposing `newInternalRaisesFuture` in manual macro code * avoid unnecessary codegen for `Future[void]` * avoid reduntant block around async proc body * simplify body generation for forward declarations with comment but no body * avoid duplicate `gcsafe` annotiations * line info for return at end of async proc * expand tests * fix comments, add defer test --------- Co-authored-by: Jacek Sieka <jacek@status.im>
2023-10-17 12:18:14 +00:00
let fut = InternalRaisesFuture[T, E]()
internalInitFutureBase(fut, loc, FutureState.Pending, {})
fut
2023-10-24 14:21:07 +00:00
proc newInternalRaisesFutureImpl[T, E](
loc: ptr SrcLoc, flags: FutureFlags): InternalRaisesFuture[T, E] =
let fut = InternalRaisesFuture[T, E]()
internalInitFutureBase(fut, loc, FutureState.Pending, flags)
fut
proc newFutureSeqImpl[A, B](loc: ptr SrcLoc): FutureSeq[A, B] =
let fut = FutureSeq[A, B]()
internalInitFutureBase(fut, loc, FutureState.Pending, {})
fut
proc newFutureStrImpl[T](loc: ptr SrcLoc): FutureStr[T] =
let fut = FutureStr[T]()
internalInitFutureBase(fut, loc, FutureState.Pending, {})
fut
template newFuture*[T](fromProc: static[string] = "",
Raise tracking (#251) * Exception tracking v2 * some fixes * Nim 1.2 compat * simpler things * Fixes for libp2p * Fixes for strictException * better await exception check * Fix for template async proc * make async work with procTy * FuturEx is now a ref object type * add tests * update test * update readme * Switch to asyncraises pragma * Address tests review comments * Rename FuturEx to RaiseTrackingFuture * Fix typo * Split asyncraises into async, asyncraises * Add -d:chronosWarnMissingRaises * Add comment to RaiseTrackingFuture * Allow standalone asyncraises * CheckedFuture.fail type checking * First cleanup * Remove useless line * Review comments * nimble: Remove #head from unittest2 * Remove implict raises: CancelledError * Move checkFutureExceptions to asyncfutures2 * Small refacto * small cleanup * Complete in closure finally * cleanup tests, add comment * bump * chronos is not compatible with nim 1.2 anymore * re-add readme modifications * fix special exception handlers * also propagate excetion type in `read` * `RaiseTrackingFuture` -> `InternalRaisesFuture` Use internal naming scheme for RTF (this type should only be accessed via asyncraises) * use `internalError` for error reading * oops * 2.0 workarounds * again * remove try/finally for non-raising functions * Revert "remove try/finally for non-raising functions" This reverts commit 86bfeb5c972ef379a3bd34e4a16cd158a7455721. `finally` is needed if code returns early :/ * fixes * avoid exposing `newInternalRaisesFuture` in manual macro code * avoid unnecessary codegen for `Future[void]` * avoid reduntant block around async proc body * simplify body generation for forward declarations with comment but no body * avoid duplicate `gcsafe` annotiations * line info for return at end of async proc * expand tests * fix comments, add defer test --------- Co-authored-by: Jacek Sieka <jacek@status.im>
2023-10-17 12:18:14 +00:00
flags: static[FutureFlags] = {}): auto =
2018-05-16 08:22:34 +00:00
## Creates a new future.
##
## Specifying ``fromProc``, which is a string specifying the name of the proc
## that this future belongs to, is a good habit as it helps with debugging.
Raise tracking (#251) * Exception tracking v2 * some fixes * Nim 1.2 compat * simpler things * Fixes for libp2p * Fixes for strictException * better await exception check * Fix for template async proc * make async work with procTy * FuturEx is now a ref object type * add tests * update test * update readme * Switch to asyncraises pragma * Address tests review comments * Rename FuturEx to RaiseTrackingFuture * Fix typo * Split asyncraises into async, asyncraises * Add -d:chronosWarnMissingRaises * Add comment to RaiseTrackingFuture * Allow standalone asyncraises * CheckedFuture.fail type checking * First cleanup * Remove useless line * Review comments * nimble: Remove #head from unittest2 * Remove implict raises: CancelledError * Move checkFutureExceptions to asyncfutures2 * Small refacto * small cleanup * Complete in closure finally * cleanup tests, add comment * bump * chronos is not compatible with nim 1.2 anymore * re-add readme modifications * fix special exception handlers * also propagate excetion type in `read` * `RaiseTrackingFuture` -> `InternalRaisesFuture` Use internal naming scheme for RTF (this type should only be accessed via asyncraises) * use `internalError` for error reading * oops * 2.0 workarounds * again * remove try/finally for non-raising functions * Revert "remove try/finally for non-raising functions" This reverts commit 86bfeb5c972ef379a3bd34e4a16cd158a7455721. `finally` is needed if code returns early :/ * fixes * avoid exposing `newInternalRaisesFuture` in manual macro code * avoid unnecessary codegen for `Future[void]` * avoid reduntant block around async proc body * simplify body generation for forward declarations with comment but no body * avoid duplicate `gcsafe` annotiations * line info for return at end of async proc * expand tests * fix comments, add defer test --------- Co-authored-by: Jacek Sieka <jacek@status.im>
2023-10-17 12:18:14 +00:00
when declared(InternalRaisesFutureRaises): # injected by `asyncraises`
2023-10-24 14:21:07 +00:00
newInternalRaisesFutureImpl[T, InternalRaisesFutureRaises](
getSrcLocation(fromProc), flags)
Raise tracking (#251) * Exception tracking v2 * some fixes * Nim 1.2 compat * simpler things * Fixes for libp2p * Fixes for strictException * better await exception check * Fix for template async proc * make async work with procTy * FuturEx is now a ref object type * add tests * update test * update readme * Switch to asyncraises pragma * Address tests review comments * Rename FuturEx to RaiseTrackingFuture * Fix typo * Split asyncraises into async, asyncraises * Add -d:chronosWarnMissingRaises * Add comment to RaiseTrackingFuture * Allow standalone asyncraises * CheckedFuture.fail type checking * First cleanup * Remove useless line * Review comments * nimble: Remove #head from unittest2 * Remove implict raises: CancelledError * Move checkFutureExceptions to asyncfutures2 * Small refacto * small cleanup * Complete in closure finally * cleanup tests, add comment * bump * chronos is not compatible with nim 1.2 anymore * re-add readme modifications * fix special exception handlers * also propagate excetion type in `read` * `RaiseTrackingFuture` -> `InternalRaisesFuture` Use internal naming scheme for RTF (this type should only be accessed via asyncraises) * use `internalError` for error reading * oops * 2.0 workarounds * again * remove try/finally for non-raising functions * Revert "remove try/finally for non-raising functions" This reverts commit 86bfeb5c972ef379a3bd34e4a16cd158a7455721. `finally` is needed if code returns early :/ * fixes * avoid exposing `newInternalRaisesFuture` in manual macro code * avoid unnecessary codegen for `Future[void]` * avoid reduntant block around async proc body * simplify body generation for forward declarations with comment but no body * avoid duplicate `gcsafe` annotiations * line info for return at end of async proc * expand tests * fix comments, add defer test --------- Co-authored-by: Jacek Sieka <jacek@status.im>
2023-10-17 12:18:14 +00:00
else:
newFutureImpl[T](getSrcLocation(fromProc), flags)
template newInternalRaisesFuture*[T, E](fromProc: static[string] = ""): auto =
Raise tracking (#251) * Exception tracking v2 * some fixes * Nim 1.2 compat * simpler things * Fixes for libp2p * Fixes for strictException * better await exception check * Fix for template async proc * make async work with procTy * FuturEx is now a ref object type * add tests * update test * update readme * Switch to asyncraises pragma * Address tests review comments * Rename FuturEx to RaiseTrackingFuture * Fix typo * Split asyncraises into async, asyncraises * Add -d:chronosWarnMissingRaises * Add comment to RaiseTrackingFuture * Allow standalone asyncraises * CheckedFuture.fail type checking * First cleanup * Remove useless line * Review comments * nimble: Remove #head from unittest2 * Remove implict raises: CancelledError * Move checkFutureExceptions to asyncfutures2 * Small refacto * small cleanup * Complete in closure finally * cleanup tests, add comment * bump * chronos is not compatible with nim 1.2 anymore * re-add readme modifications * fix special exception handlers * also propagate excetion type in `read` * `RaiseTrackingFuture` -> `InternalRaisesFuture` Use internal naming scheme for RTF (this type should only be accessed via asyncraises) * use `internalError` for error reading * oops * 2.0 workarounds * again * remove try/finally for non-raising functions * Revert "remove try/finally for non-raising functions" This reverts commit 86bfeb5c972ef379a3bd34e4a16cd158a7455721. `finally` is needed if code returns early :/ * fixes * avoid exposing `newInternalRaisesFuture` in manual macro code * avoid unnecessary codegen for `Future[void]` * avoid reduntant block around async proc body * simplify body generation for forward declarations with comment but no body * avoid duplicate `gcsafe` annotiations * line info for return at end of async proc * expand tests * fix comments, add defer test --------- Co-authored-by: Jacek Sieka <jacek@status.im>
2023-10-17 12:18:14 +00:00
## Creates a new future.
##
## Specifying ``fromProc``, which is a string specifying the name of the proc
## that this future belongs to, is a good habit as it helps with debugging.
newInternalRaisesFutureImpl[T, E](getSrcLocation(fromProc))
2018-05-16 08:22:34 +00:00
template newFutureSeq*[A, B](fromProc: static[string] = ""): FutureSeq[A, B] {.deprecated.} =
## Create a new future which can hold/preserve GC sequence until future will
## not be completed.
2018-05-16 08:22:34 +00:00
##
## Specifying ``fromProc``, which is a string specifying the name of the proc
## that this future belongs to, is a good habit as it helps with debugging.
newFutureSeqImpl[A, B](getSrcLocation(fromProc))
2018-05-16 08:22:34 +00:00
template newFutureStr*[T](fromProc: static[string] = ""): FutureStr[T] {.deprecated.} =
## Create a new future which can hold/preserve GC string until future will
## not be completed.
##
## Specifying ``fromProc``, which is a string specifying the name of the proc
## that this future belongs to, is a good habit as it helps with debugging.
newFutureStrImpl[T](getSrcLocation(fromProc))
proc done*(future: FutureBase): bool {.deprecated: "Use `completed` instead".} =
## This is an alias for ``completed(future)`` procedure.
completed(future)
2019-06-20 20:30:41 +00:00
when chronosFutureTracking:
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
proc futureDestructor(udata: pointer) =
## This procedure will be called when Future[T] got completed, cancelled or
## failed and all Future[T].callbacks are already scheduled and processed.
let future = cast[FutureBase](udata)
if future == futureList.tail: futureList.tail = future.prev
if future == futureList.head: futureList.head = future.next
if not(isNil(future.next)): future.next.internalPrev = future.prev
if not(isNil(future.prev)): future.prev.internalNext = future.next
futureList.count.dec()
proc scheduleDestructor(future: FutureBase) {.inline.} =
callSoon(futureDestructor, cast[pointer](future))
proc checkFinished(future: FutureBase, loc: ptr SrcLoc) =
2018-05-16 08:22:34 +00:00
## Checks whether `future` is finished. If it is then raises a
2019-12-02 10:40:41 +00:00
## ``FutureDefect``.
2019-06-20 20:30:41 +00:00
if future.finished():
2019-03-29 09:53:24 +00:00
var msg = ""
msg.add("An attempt was made to complete a Future more than once. ")
msg.add("Details:")
msg.add("\n Future ID: " & Base10.toString(future.id))
msg.add("\n Creation location:")
msg.add("\n " & $future.location[LocationKind.Create])
msg.add("\n First completion location:")
msg.add("\n " & $future.location[LocationKind.Finish])
msg.add("\n Second completion location:")
msg.add("\n " & $loc)
when chronosStackTrace:
msg.add("\n Stack trace to moment of creation:")
msg.add("\n" & indent(future.stackTrace.strip(), 4))
msg.add("\n Stack trace to moment of secondary completion:")
msg.add("\n" & indent(getStackTrace().strip(), 4))
msg.add("\n\n")
2019-12-02 10:40:41 +00:00
var err = newException(FutureDefect, msg)
2019-03-29 09:53:24 +00:00
err.cause = future
raise err
else:
future.internalLocation[LocationKind.Finish] = loc
2018-05-16 08:22:34 +00:00
proc finish(fut: FutureBase, state: FutureState) =
# We do not perform any checks here, because:
# 1. `finish()` is a private procedure and `state` is under our control.
# 2. `fut.state` is checked by `checkFinished()`.
fut.internalState = state
fut.internalCancelcb = nil # release cancellation callback memory
for item in fut.internalCallbacks.mitems():
if not(isNil(item.function)):
callSoon(item)
item = default(AsyncCallback) # release memory as early as possible
fut.internalCallbacks = default(seq[AsyncCallback]) # release seq as well
2018-05-16 08:22:34 +00:00
when chronosFutureTracking:
scheduleDestructor(fut)
2018-05-16 08:22:34 +00:00
proc complete[T](future: Future[T], val: T, loc: ptr SrcLoc) =
if not(future.cancelled()):
checkFinished(future, loc)
doAssert(isNil(future.internalError))
future.internalValue = val
future.finish(FutureState.Completed)
2018-05-16 08:22:34 +00:00
template complete*[T](future: Future[T], val: T) =
## Completes ``future`` with value ``val``.
complete(future, val, getSrcLocation())
proc complete(future: Future[void], loc: ptr SrcLoc) =
if not(future.cancelled()):
checkFinished(future, loc)
doAssert(isNil(future.internalError))
future.finish(FutureState.Completed)
2018-05-16 08:22:34 +00:00
template complete*(future: Future[void]) =
## Completes a void ``future``.
complete(future, getSrcLocation())
proc failImpl(
future: FutureBase, error: ref CatchableError, loc: ptr SrcLoc) =
if not(future.cancelled()):
checkFinished(future, loc)
future.internalError = error
when chronosStackTrace:
future.internalErrorStackTrace = if getStackTrace(error) == "":
getStackTrace()
else:
getStackTrace(error)
future.finish(FutureState.Failed)
2018-05-16 08:22:34 +00:00
template fail*[T](
future: Future[T], error: ref CatchableError, warn: static bool = false) =
## Completes ``future`` with ``error``.
failImpl(future, error, getSrcLocation())
template fail*[T, E](
future: InternalRaisesFuture[T, E], error: ref CatchableError,
warn: static bool = true) =
checkRaises(future, E, error, warn)
failImpl(future, error, getSrcLocation())
template newCancelledError(): ref CancelledError =
(ref CancelledError)(msg: "Future operation cancelled!")
proc cancelAndSchedule(future: FutureBase, loc: ptr SrcLoc) =
if not(future.finished()):
2019-06-20 20:30:41 +00:00
checkFinished(future, loc)
future.internalError = newCancelledError()
when chronosStackTrace:
future.internalErrorStackTrace = getStackTrace()
future.finish(FutureState.Cancelled)
template cancelAndSchedule*(future: FutureBase) =
cancelAndSchedule(future, getSrcLocation())
proc tryCancel(future: FutureBase, loc: ptr SrcLoc): bool =
## Perform an attempt to cancel ``future``.
##
## NOTE: This procedure does not guarantee that cancellation will actually
## happened.
##
## Cancellation is the process which starts from the last ``future``
## descendent and moves step by step to the parent ``future``. To initiate
## this process procedure iterates through all non-finished ``future``
## descendents and tries to find the last one. If last descendent is still
## pending it will become cancelled and process will be initiated. In such
## case this procedure returns ``true``.
##
## If last descendent future is not pending, this procedure will be unable to
## initiate cancellation process and so it returns ``false``.
if future.cancelled():
return true
if future.finished():
return false
if not(isNil(future.internalChild)):
# If you hit this assertion, you should have used the `CancelledError`
# mechanism and/or use a regular `addCallback`
when chronosStrictFutureAccess:
doAssert future.internalCancelcb.isNil,
"futures returned from `{.async.}` functions must not use " &
"`cancelCallback`"
tryCancel(future.internalChild, loc)
else:
if not(isNil(future.internalCancelcb)):
future.internalCancelcb(cast[pointer](future))
if FutureFlag.OwnCancelSchedule notin future.internalFlags:
cancelAndSchedule(future, loc)
future.cancelled()
2019-06-20 20:30:41 +00:00
template tryCancel*(future: FutureBase): bool =
tryCancel(future, getSrcLocation())
2022-09-12 12:11:13 +00:00
2018-05-16 08:22:34 +00:00
proc clearCallbacks(future: FutureBase) =
future.internalCallbacks = default(seq[AsyncCallback])
2018-05-16 08:22:34 +00:00
proc addCallback*(future: FutureBase, cb: CallbackFunc, udata: pointer) =
2018-05-16 08:22:34 +00:00
## Adds the callbacks proc to be called when the future completes.
##
## If future has already completed then ``cb`` will be called immediately.
2019-03-15 00:43:51 +00:00
doAssert(not isNil(cb))
2019-06-20 20:30:41 +00:00
if future.finished():
2018-05-16 08:22:34 +00:00
callSoon(cb, udata)
else:
future.internalCallbacks.add AsyncCallback(function: cb, udata: udata)
2018-05-16 08:22:34 +00:00
proc addCallback*(future: FutureBase, cb: CallbackFunc) =
2018-05-16 08:22:34 +00:00
## Adds the callbacks proc to be called when the future completes.
##
## If future has already completed then ``cb`` will be called immediately.
2018-05-17 08:45:18 +00:00
future.addCallback(cb, cast[pointer](future))
2018-05-16 08:22:34 +00:00
proc removeCallback*(future: FutureBase, cb: CallbackFunc,
udata: pointer) =
## Remove future from list of callbacks - this operation may be slow if there
## are many registered callbacks!
2019-03-15 00:43:51 +00:00
doAssert(not isNil(cb))
# Make sure to release memory associated with callback, or reference chains
# may be created!
future.internalCallbacks.keepItIf:
it.function != cb or it.udata != udata
2018-05-16 08:22:34 +00:00
proc removeCallback*(future: FutureBase, cb: CallbackFunc) =
2018-05-17 08:45:18 +00:00
future.removeCallback(cb, cast[pointer](future))
2018-05-16 08:22:34 +00:00
proc `callback=`*(future: FutureBase, cb: CallbackFunc, udata: pointer) =
2018-05-16 08:22:34 +00:00
## Clears the list of callbacks and sets the callback proc to be called when
## the future completes.
##
## If future has already completed then ``cb`` will be called immediately.
##
## It's recommended to use ``addCallback`` or ``then`` instead.
2018-05-29 18:04:11 +00:00
# ZAH: how about `setLen(1); callbacks[0] = cb`
2018-05-16 08:22:34 +00:00
future.clearCallbacks
future.addCallback(cb, udata)
proc `callback=`*(future: FutureBase, cb: CallbackFunc) =
2018-05-16 08:22:34 +00:00
## Sets the callback proc to be called when the future completes.
##
## If future has already completed then ``cb`` will be called immediately.
`callback=`(future, cb, cast[pointer](future))
proc `cancelCallback=`*(future: FutureBase, cb: CallbackFunc) =
2019-06-20 20:30:41 +00:00
## Sets the callback procedure to be called when the future is cancelled.
##
## This callback will be called immediately as ``future.cancel()`` invoked and
## must be set before future is finished.
when chronosStrictFutureAccess:
doAssert not future.finished(),
"cancellation callback must be set before finishing the future"
future.internalCancelcb = cb
2019-06-20 20:30:41 +00:00
{.push stackTrace: off.}
2023-06-05 20:21:50 +00:00
proc futureContinue*(fut: FutureBase) {.raises: [], gcsafe.}
2023-06-05 20:21:50 +00:00
proc internalContinue(fut: pointer) {.raises: [], gcsafe.} =
let asFut = cast[FutureBase](fut)
GC_unref(asFut)
futureContinue(asFut)
2023-06-05 20:21:50 +00:00
proc futureContinue*(fut: FutureBase) {.raises: [], gcsafe.} =
# This function is responsible for calling the closure iterator generated by
# the `{.async.}` transformation either until it has completed its iteration
#
# Every call to an `{.async.}` proc is redirected to call this function
# instead with its original body captured in `fut.closure`.
while true:
# Call closure to make progress on `fut` until it reaches `yield` (inside
# `await` typically) or completes / fails / is cancelled
let next: FutureBase = fut.internalClosure(fut)
if fut.internalClosure.finished(): # Reached the end of the transformed proc
break
if next == nil:
raiseAssert "Async procedure (" & ($fut.location[LocationKind.Create]) &
") yielded `nil`, are you await'ing a `nil` Future?"
if not next.finished():
# We cannot make progress on `fut` until `next` has finished - schedule
# `fut` to continue running when that happens
GC_ref(fut)
next.addCallback(CallbackFunc(internalContinue), cast[pointer](fut))
# return here so that we don't remove the closure below
return
# Continue while the yielded future is already finished.
# `futureContinue` will not be called any more for this future so we can
# clean it up
fut.internalClosure = nil
fut.internalChild = nil
{.pop.}
when chronosStackTrace:
template getFilenameProcname(entry: StackTraceEntry): (string, string) =
when compiles(entry.filenameStr) and compiles(entry.procnameStr):
# We can't rely on "entry.filename" and "entry.procname" still being valid
# cstring pointers, because the "string.data" buffers they pointed to might
# be already garbage collected (this entry being a non-shallow copy,
# "entry.filename" no longer points to "entry.filenameStr.data", but to the
# buffer of the original object).
(entry.filenameStr, entry.procnameStr)
2021-05-06 07:49:55 +00:00
else:
($entry.filename, $entry.procname)
2018-05-16 08:22:34 +00:00
proc `$`(stackTraceEntries: seq[StackTraceEntry]): string =
try:
when defined(nimStackTraceOverride) and declared(addDebuggingInfo):
let entries = addDebuggingInfo(stackTraceEntries)
else:
let entries = stackTraceEntries
# Find longest filename & line number combo for alignment purposes.
var longestLeft = 0
for entry in entries:
let (filename, procname) = getFilenameProcname(entry)
if procname == "": continue
let leftLen = filename.len + len($entry.line)
if leftLen > longestLeft:
longestLeft = leftLen
var indent = 2
# Format the entries.
for entry in entries:
let (filename, procname) = getFilenameProcname(entry)
if procname == "":
if entry.line == reraisedFromBegin:
result.add(spaces(indent) & "#[\n")
indent.inc(2)
elif entry.line == reraisedFromEnd:
indent.dec(2)
result.add(spaces(indent) & "]#\n")
continue
let left = "$#($#)" % [filename, $entry.line]
result.add((spaces(indent) & "$#$# $#\n") % [
left,
spaces(longestLeft - left.len + 2),
procname
])
except ValueError as exc:
return exc.msg # Shouldn't actually happen since we set the formatting
# string
proc injectStacktrace(error: ref Exception) =
const header = "\nAsync traceback:\n"
2018-05-16 08:22:34 +00:00
var exceptionMsg = error.msg
if header in exceptionMsg:
# This is messy: extract the original exception message from the msg
# containing the async traceback.
let start = exceptionMsg.find(header)
exceptionMsg = exceptionMsg[0..<start]
2018-05-16 08:22:34 +00:00
var newMsg = exceptionMsg & header
2018-05-16 08:22:34 +00:00
let entries = getStackTraceEntries(error)
newMsg.add($entries)
2018-05-16 08:22:34 +00:00
newMsg.add("Exception message: " & exceptionMsg & "\n")
2018-05-16 08:22:34 +00:00
# # For debugging purposes
2021-05-06 07:49:55 +00:00
# newMsg.add("Exception type:")
# for entry in getStackTraceEntries(future.error):
# newMsg.add "\n" & $entry
error.msg = newMsg
2018-05-16 08:22:34 +00:00
2023-06-05 20:21:50 +00:00
proc internalCheckComplete*(fut: FutureBase) {.raises: [CatchableError].} =
2019-08-15 13:32:46 +00:00
# For internal use only. Used in asyncmacro
if not(isNil(fut.internalError)):
when chronosStackTrace:
injectStacktrace(fut.internalError)
raise fut.internalError
2019-08-15 13:32:46 +00:00
macro internalCheckComplete*(fut: InternalRaisesFuture, raises: typed) =
Raise tracking (#251) * Exception tracking v2 * some fixes * Nim 1.2 compat * simpler things * Fixes for libp2p * Fixes for strictException * better await exception check * Fix for template async proc * make async work with procTy * FuturEx is now a ref object type * add tests * update test * update readme * Switch to asyncraises pragma * Address tests review comments * Rename FuturEx to RaiseTrackingFuture * Fix typo * Split asyncraises into async, asyncraises * Add -d:chronosWarnMissingRaises * Add comment to RaiseTrackingFuture * Allow standalone asyncraises * CheckedFuture.fail type checking * First cleanup * Remove useless line * Review comments * nimble: Remove #head from unittest2 * Remove implict raises: CancelledError * Move checkFutureExceptions to asyncfutures2 * Small refacto * small cleanup * Complete in closure finally * cleanup tests, add comment * bump * chronos is not compatible with nim 1.2 anymore * re-add readme modifications * fix special exception handlers * also propagate excetion type in `read` * `RaiseTrackingFuture` -> `InternalRaisesFuture` Use internal naming scheme for RTF (this type should only be accessed via asyncraises) * use `internalError` for error reading * oops * 2.0 workarounds * again * remove try/finally for non-raising functions * Revert "remove try/finally for non-raising functions" This reverts commit 86bfeb5c972ef379a3bd34e4a16cd158a7455721. `finally` is needed if code returns early :/ * fixes * avoid exposing `newInternalRaisesFuture` in manual macro code * avoid unnecessary codegen for `Future[void]` * avoid reduntant block around async proc body * simplify body generation for forward declarations with comment but no body * avoid duplicate `gcsafe` annotiations * line info for return at end of async proc * expand tests * fix comments, add defer test --------- Co-authored-by: Jacek Sieka <jacek@status.im>
2023-10-17 12:18:14 +00:00
# For InternalRaisesFuture[void, (ValueError, OSError), will do:
# {.cast(raises: [ValueError, OSError]).}:
# if isNil(f.error): discard
# else: raise f.error
# TODO https://github.com/nim-lang/Nim/issues/22937
# we cannot `getTypeInst` on the `fut` - when aliases are involved, the
# generics are lost - so instead, we pass the raises list explicitly
Raise tracking (#251) * Exception tracking v2 * some fixes * Nim 1.2 compat * simpler things * Fixes for libp2p * Fixes for strictException * better await exception check * Fix for template async proc * make async work with procTy * FuturEx is now a ref object type * add tests * update test * update readme * Switch to asyncraises pragma * Address tests review comments * Rename FuturEx to RaiseTrackingFuture * Fix typo * Split asyncraises into async, asyncraises * Add -d:chronosWarnMissingRaises * Add comment to RaiseTrackingFuture * Allow standalone asyncraises * CheckedFuture.fail type checking * First cleanup * Remove useless line * Review comments * nimble: Remove #head from unittest2 * Remove implict raises: CancelledError * Move checkFutureExceptions to asyncfutures2 * Small refacto * small cleanup * Complete in closure finally * cleanup tests, add comment * bump * chronos is not compatible with nim 1.2 anymore * re-add readme modifications * fix special exception handlers * also propagate excetion type in `read` * `RaiseTrackingFuture` -> `InternalRaisesFuture` Use internal naming scheme for RTF (this type should only be accessed via asyncraises) * use `internalError` for error reading * oops * 2.0 workarounds * again * remove try/finally for non-raising functions * Revert "remove try/finally for non-raising functions" This reverts commit 86bfeb5c972ef379a3bd34e4a16cd158a7455721. `finally` is needed if code returns early :/ * fixes * avoid exposing `newInternalRaisesFuture` in manual macro code * avoid unnecessary codegen for `Future[void]` * avoid reduntant block around async proc body * simplify body generation for forward declarations with comment but no body * avoid duplicate `gcsafe` annotiations * line info for return at end of async proc * expand tests * fix comments, add defer test --------- Co-authored-by: Jacek Sieka <jacek@status.im>
2023-10-17 12:18:14 +00:00
let types = getRaisesTypes(raises)
types.copyLineInfo(raises)
for t in types:
t.copyLineInfo(raises)
if isNoRaises(types):
Raise tracking (#251) * Exception tracking v2 * some fixes * Nim 1.2 compat * simpler things * Fixes for libp2p * Fixes for strictException * better await exception check * Fix for template async proc * make async work with procTy * FuturEx is now a ref object type * add tests * update test * update readme * Switch to asyncraises pragma * Address tests review comments * Rename FuturEx to RaiseTrackingFuture * Fix typo * Split asyncraises into async, asyncraises * Add -d:chronosWarnMissingRaises * Add comment to RaiseTrackingFuture * Allow standalone asyncraises * CheckedFuture.fail type checking * First cleanup * Remove useless line * Review comments * nimble: Remove #head from unittest2 * Remove implict raises: CancelledError * Move checkFutureExceptions to asyncfutures2 * Small refacto * small cleanup * Complete in closure finally * cleanup tests, add comment * bump * chronos is not compatible with nim 1.2 anymore * re-add readme modifications * fix special exception handlers * also propagate excetion type in `read` * `RaiseTrackingFuture` -> `InternalRaisesFuture` Use internal naming scheme for RTF (this type should only be accessed via asyncraises) * use `internalError` for error reading * oops * 2.0 workarounds * again * remove try/finally for non-raising functions * Revert "remove try/finally for non-raising functions" This reverts commit 86bfeb5c972ef379a3bd34e4a16cd158a7455721. `finally` is needed if code returns early :/ * fixes * avoid exposing `newInternalRaisesFuture` in manual macro code * avoid unnecessary codegen for `Future[void]` * avoid reduntant block around async proc body * simplify body generation for forward declarations with comment but no body * avoid duplicate `gcsafe` annotiations * line info for return at end of async proc * expand tests * fix comments, add defer test --------- Co-authored-by: Jacek Sieka <jacek@status.im>
2023-10-17 12:18:14 +00:00
return quote do:
if not(isNil(`fut`.internalError)):
# This would indicate a bug in which `error` was set via the non-raising
# base type
raiseAssert("Error set on a non-raising future: " & `fut`.internalError.msg)
Raise tracking (#251) * Exception tracking v2 * some fixes * Nim 1.2 compat * simpler things * Fixes for libp2p * Fixes for strictException * better await exception check * Fix for template async proc * make async work with procTy * FuturEx is now a ref object type * add tests * update test * update readme * Switch to asyncraises pragma * Address tests review comments * Rename FuturEx to RaiseTrackingFuture * Fix typo * Split asyncraises into async, asyncraises * Add -d:chronosWarnMissingRaises * Add comment to RaiseTrackingFuture * Allow standalone asyncraises * CheckedFuture.fail type checking * First cleanup * Remove useless line * Review comments * nimble: Remove #head from unittest2 * Remove implict raises: CancelledError * Move checkFutureExceptions to asyncfutures2 * Small refacto * small cleanup * Complete in closure finally * cleanup tests, add comment * bump * chronos is not compatible with nim 1.2 anymore * re-add readme modifications * fix special exception handlers * also propagate excetion type in `read` * `RaiseTrackingFuture` -> `InternalRaisesFuture` Use internal naming scheme for RTF (this type should only be accessed via asyncraises) * use `internalError` for error reading * oops * 2.0 workarounds * again * remove try/finally for non-raising functions * Revert "remove try/finally for non-raising functions" This reverts commit 86bfeb5c972ef379a3bd34e4a16cd158a7455721. `finally` is needed if code returns early :/ * fixes * avoid exposing `newInternalRaisesFuture` in manual macro code * avoid unnecessary codegen for `Future[void]` * avoid reduntant block around async proc body * simplify body generation for forward declarations with comment but no body * avoid duplicate `gcsafe` annotiations * line info for return at end of async proc * expand tests * fix comments, add defer test --------- Co-authored-by: Jacek Sieka <jacek@status.im>
2023-10-17 12:18:14 +00:00
expectKind(types, nnkBracketExpr)
expectKind(types[0], nnkSym)
Raise tracking (#251) * Exception tracking v2 * some fixes * Nim 1.2 compat * simpler things * Fixes for libp2p * Fixes for strictException * better await exception check * Fix for template async proc * make async work with procTy * FuturEx is now a ref object type * add tests * update test * update readme * Switch to asyncraises pragma * Address tests review comments * Rename FuturEx to RaiseTrackingFuture * Fix typo * Split asyncraises into async, asyncraises * Add -d:chronosWarnMissingRaises * Add comment to RaiseTrackingFuture * Allow standalone asyncraises * CheckedFuture.fail type checking * First cleanup * Remove useless line * Review comments * nimble: Remove #head from unittest2 * Remove implict raises: CancelledError * Move checkFutureExceptions to asyncfutures2 * Small refacto * small cleanup * Complete in closure finally * cleanup tests, add comment * bump * chronos is not compatible with nim 1.2 anymore * re-add readme modifications * fix special exception handlers * also propagate excetion type in `read` * `RaiseTrackingFuture` -> `InternalRaisesFuture` Use internal naming scheme for RTF (this type should only be accessed via asyncraises) * use `internalError` for error reading * oops * 2.0 workarounds * again * remove try/finally for non-raising functions * Revert "remove try/finally for non-raising functions" This reverts commit 86bfeb5c972ef379a3bd34e4a16cd158a7455721. `finally` is needed if code returns early :/ * fixes * avoid exposing `newInternalRaisesFuture` in manual macro code * avoid unnecessary codegen for `Future[void]` * avoid reduntant block around async proc body * simplify body generation for forward declarations with comment but no body * avoid duplicate `gcsafe` annotiations * line info for return at end of async proc * expand tests * fix comments, add defer test --------- Co-authored-by: Jacek Sieka <jacek@status.im>
2023-10-17 12:18:14 +00:00
assert types[0].strVal == "tuple"
let ifRaise = nnkIfExpr.newTree(
nnkElifExpr.newTree(
quote do: isNil(`fut`.internalError),
Raise tracking (#251) * Exception tracking v2 * some fixes * Nim 1.2 compat * simpler things * Fixes for libp2p * Fixes for strictException * better await exception check * Fix for template async proc * make async work with procTy * FuturEx is now a ref object type * add tests * update test * update readme * Switch to asyncraises pragma * Address tests review comments * Rename FuturEx to RaiseTrackingFuture * Fix typo * Split asyncraises into async, asyncraises * Add -d:chronosWarnMissingRaises * Add comment to RaiseTrackingFuture * Allow standalone asyncraises * CheckedFuture.fail type checking * First cleanup * Remove useless line * Review comments * nimble: Remove #head from unittest2 * Remove implict raises: CancelledError * Move checkFutureExceptions to asyncfutures2 * Small refacto * small cleanup * Complete in closure finally * cleanup tests, add comment * bump * chronos is not compatible with nim 1.2 anymore * re-add readme modifications * fix special exception handlers * also propagate excetion type in `read` * `RaiseTrackingFuture` -> `InternalRaisesFuture` Use internal naming scheme for RTF (this type should only be accessed via asyncraises) * use `internalError` for error reading * oops * 2.0 workarounds * again * remove try/finally for non-raising functions * Revert "remove try/finally for non-raising functions" This reverts commit 86bfeb5c972ef379a3bd34e4a16cd158a7455721. `finally` is needed if code returns early :/ * fixes * avoid exposing `newInternalRaisesFuture` in manual macro code * avoid unnecessary codegen for `Future[void]` * avoid reduntant block around async proc body * simplify body generation for forward declarations with comment but no body * avoid duplicate `gcsafe` annotiations * line info for return at end of async proc * expand tests * fix comments, add defer test --------- Co-authored-by: Jacek Sieka <jacek@status.im>
2023-10-17 12:18:14 +00:00
quote do: discard
),
nnkElseExpr.newTree(
nnkRaiseStmt.newTree(
nnkDotExpr.newTree(fut, ident "internalError")
Raise tracking (#251) * Exception tracking v2 * some fixes * Nim 1.2 compat * simpler things * Fixes for libp2p * Fixes for strictException * better await exception check * Fix for template async proc * make async work with procTy * FuturEx is now a ref object type * add tests * update test * update readme * Switch to asyncraises pragma * Address tests review comments * Rename FuturEx to RaiseTrackingFuture * Fix typo * Split asyncraises into async, asyncraises * Add -d:chronosWarnMissingRaises * Add comment to RaiseTrackingFuture * Allow standalone asyncraises * CheckedFuture.fail type checking * First cleanup * Remove useless line * Review comments * nimble: Remove #head from unittest2 * Remove implict raises: CancelledError * Move checkFutureExceptions to asyncfutures2 * Small refacto * small cleanup * Complete in closure finally * cleanup tests, add comment * bump * chronos is not compatible with nim 1.2 anymore * re-add readme modifications * fix special exception handlers * also propagate excetion type in `read` * `RaiseTrackingFuture` -> `InternalRaisesFuture` Use internal naming scheme for RTF (this type should only be accessed via asyncraises) * use `internalError` for error reading * oops * 2.0 workarounds * again * remove try/finally for non-raising functions * Revert "remove try/finally for non-raising functions" This reverts commit 86bfeb5c972ef379a3bd34e4a16cd158a7455721. `finally` is needed if code returns early :/ * fixes * avoid exposing `newInternalRaisesFuture` in manual macro code * avoid unnecessary codegen for `Future[void]` * avoid reduntant block around async proc body * simplify body generation for forward declarations with comment but no body * avoid duplicate `gcsafe` annotiations * line info for return at end of async proc * expand tests * fix comments, add defer test --------- Co-authored-by: Jacek Sieka <jacek@status.im>
2023-10-17 12:18:14 +00:00
)
)
)
nnkPragmaBlock.newTree(
nnkPragma.newTree(
nnkCast.newTree(
newEmptyNode(),
nnkExprColonExpr.newTree(
ident"raises",
block:
var res = nnkBracket.newTree()
for r in types[1..^1]:
res.add(r)
res
)
),
),
ifRaise
)
proc readFinished[T: not void](fut: Future[T]): lent T {.
raises: [CatchableError].} =
# Read a future that is known to be finished, avoiding the extra exception
# effect.
internalCheckComplete(fut)
fut.internalValue
proc read*[T: not void](fut: Future[T] ): lent T {.raises: [CatchableError].} =
## Retrieves the value of `fut`.
##
## If the future failed or was cancelled, the corresponding exception will be
## raised.
##
## If the future is still pending, `FuturePendingError` will be raised.
if not fut.finished():
raiseFuturePendingError(fut)
fut.readFinished()
2019-08-15 13:32:46 +00:00
proc read*(fut: Future[void]) {.raises: [CatchableError].} =
## Checks that `fut` completed.
2018-05-16 08:22:34 +00:00
##
## If the future failed or was cancelled, the corresponding exception will be
## raised.
##
## If the future is still pending, `FuturePendingError` will be raised.
if not fut.finished():
raiseFuturePendingError(fut)
internalCheckComplete(fut)
Raise tracking (#251) * Exception tracking v2 * some fixes * Nim 1.2 compat * simpler things * Fixes for libp2p * Fixes for strictException * better await exception check * Fix for template async proc * make async work with procTy * FuturEx is now a ref object type * add tests * update test * update readme * Switch to asyncraises pragma * Address tests review comments * Rename FuturEx to RaiseTrackingFuture * Fix typo * Split asyncraises into async, asyncraises * Add -d:chronosWarnMissingRaises * Add comment to RaiseTrackingFuture * Allow standalone asyncraises * CheckedFuture.fail type checking * First cleanup * Remove useless line * Review comments * nimble: Remove #head from unittest2 * Remove implict raises: CancelledError * Move checkFutureExceptions to asyncfutures2 * Small refacto * small cleanup * Complete in closure finally * cleanup tests, add comment * bump * chronos is not compatible with nim 1.2 anymore * re-add readme modifications * fix special exception handlers * also propagate excetion type in `read` * `RaiseTrackingFuture` -> `InternalRaisesFuture` Use internal naming scheme for RTF (this type should only be accessed via asyncraises) * use `internalError` for error reading * oops * 2.0 workarounds * again * remove try/finally for non-raising functions * Revert "remove try/finally for non-raising functions" This reverts commit 86bfeb5c972ef379a3bd34e4a16cd158a7455721. `finally` is needed if code returns early :/ * fixes * avoid exposing `newInternalRaisesFuture` in manual macro code * avoid unnecessary codegen for `Future[void]` * avoid reduntant block around async proc body * simplify body generation for forward declarations with comment but no body * avoid duplicate `gcsafe` annotiations * line info for return at end of async proc * expand tests * fix comments, add defer test --------- Co-authored-by: Jacek Sieka <jacek@status.im>
2023-10-17 12:18:14 +00:00
proc readError*(fut: FutureBase): ref CatchableError {.raises: [FutureError].} =
## Retrieves the exception of the failed or cancelled `fut`.
2018-05-16 08:22:34 +00:00
##
## If the future was completed with a value, `FutureCompletedError` will be
## raised.
##
## If the future is still pending, `FuturePendingError` will be raised.
if not fut.finished():
raiseFuturePendingError(fut)
if isNil(fut.error):
raiseFutureCompletedError(fut)
fut.error
2018-05-16 08:22:34 +00:00
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
template taskFutureLocation(future: FutureBase): string =
let loc = future.location[LocationKind.Create]
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
"[" & (
if len(loc.procedure) == 0: "[unspecified]" else: $loc.procedure & "()"
) & " at " & $loc.file & ":" & $(loc.line) & "]"
template taskErrorMessage(future: FutureBase): string =
"Asynchronous task " & taskFutureLocation(future) &
" finished with an exception \"" & $future.error.name &
"\"!\nMessage: " & future.error.msg &
"\nStack trace: " & future.error.getStackTrace()
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
template taskCancelMessage(future: FutureBase): string =
"Asynchronous task " & taskFutureLocation(future) & " was cancelled!"
2018-05-16 08:22:34 +00:00
proc pollFor[F: Future | InternalRaisesFuture](fut: F): F {.raises: [].} =
# Blocks the current thread of execution until `fut` has finished, returning
# the given future.
#
# Must not be called recursively (from inside `async` procedures).
#
# See alse `awaitne`.
if not(fut.finished()):
var finished = false
# Ensure that callbacks currently scheduled on the future run before returning
proc continuation(udata: pointer) {.gcsafe.} = finished = true
fut.addCallback(continuation)
while not(finished):
poll()
fut
proc waitFor*[T: not void](fut: Future[T]): lent T {.raises: [CatchableError].} =
## Blocks the current thread of execution until `fut` has finished, returning
## its value.
##
## If the future failed or was cancelled, the corresponding exception will be
## raised.
##
## Must not be called recursively (from inside `async` procedures).
##
## See also `await`, `Future.read`
pollFor(fut).readFinished()
proc waitFor*(fut: Future[void]) {.raises: [CatchableError].} =
## Blocks the current thread of execution until `fut` has finished.
##
## If the future failed or was cancelled, the corresponding exception will be
## raised.
##
## Must not be called recursively (from inside `async` procedures).
##
## See also `await`, `Future.read`
pollFor(fut).internalCheckComplete()
proc asyncSpawn*(future: Future[void]) =
## Spawns a new concurrent async task.
##
## Tasks may not raise exceptions or be cancelled - a ``Defect`` will be
## raised when this happens.
##
## This should be used instead of ``discard`` and ``asyncCheck`` when calling
## an ``async`` procedure without ``await``, to ensure exceptions in the
## returned future are not silently discarded.
##
## Note, that if passed ``future`` is already finished, it will be checked
## and processed immediately.
doAssert(not isNil(future), "Future is nil")
proc cb(data: pointer) =
if future.failed():
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
raise newException(FutureDefect, taskErrorMessage(future))
elif future.cancelled():
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
raise newException(FutureDefect, taskCancelMessage(future))
if not(future.finished()):
# We adding completion callback only if ``future`` is not finished yet.
future.addCallback(cb)
else:
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
cb(nil)
proc asyncCheck*[T](future: Future[T]) {.
deprecated: "Raises Defect on future failure, fix your code and use" &
" asyncSpawn!".} =
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
## This function used to raise an exception through the `poll` call if
## the given future failed - there's no way to handle such exceptions so this
## function is now an alias for `asyncSpawn`
##
when T is void:
asyncSpawn(future)
else:
proc cb(data: pointer) =
if future.failed():
raise newException(FutureDefect, taskErrorMessage(future))
elif future.cancelled():
raise newException(FutureDefect, taskCancelMessage(future))
if not(future.finished()):
# We adding completion callback only if ``future`` is not finished yet.
future.addCallback(cb)
else:
cb(nil)
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
proc asyncDiscard*[T](future: Future[T]) {.
deprecated: "Use asyncSpawn or `discard await`".} = discard
## `asyncDiscard` will discard the outcome of the operation - unlike `discard`
## it also throws away exceptions! Use `asyncSpawn` if you're sure your
## code doesn't raise exceptions, or `discard await` to ignore successful
## outcomes
2019-03-15 00:43:51 +00:00
proc `and`*[T, Y](fut1: Future[T], fut2: Future[Y]): Future[void] {.
deprecated: "Use allFutures[T](varargs[Future[T]])".} =
2018-05-16 08:22:34 +00:00
## Returns a future which will complete once both ``fut1`` and ``fut2``
## finish.
2019-06-20 20:30:41 +00:00
##
## If cancelled, ``fut1`` and ``fut2`` futures WILL NOT BE cancelled.
var retFuture = newFuture[void]("chronos.`and`")
2018-05-16 08:22:34 +00:00
proc cb(data: pointer) =
2019-06-20 20:30:41 +00:00
if not(retFuture.finished()):
if fut1.finished() and fut2.finished():
2018-05-16 08:22:34 +00:00
if cast[pointer](fut1) == data:
2019-06-20 20:30:41 +00:00
if fut1.failed():
retFuture.fail(fut1.error)
else:
retFuture.complete()
2018-05-16 08:22:34 +00:00
else:
2019-06-20 20:30:41 +00:00
if fut2.failed():
retFuture.fail(fut2.error)
else:
retFuture.complete()
2018-05-16 08:22:34 +00:00
fut1.callback = cb
fut2.callback = cb
2019-06-20 20:30:41 +00:00
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
proc cancellation(udata: pointer) =
# On cancel we remove all our callbacks only.
if not(fut1.finished()):
fut1.removeCallback(cb)
if not(fut2.finished()):
fut2.removeCallback(cb)
2019-06-20 20:30:41 +00:00
retFuture.cancelCallback = cancellation
2018-05-16 08:22:34 +00:00
return retFuture
2023-10-24 14:21:07 +00:00
template orImpl*[T, Y](fut1: Future[T], fut2: Future[Y]): untyped =
2023-06-05 20:21:50 +00:00
var cb: proc(udata: pointer) {.gcsafe, raises: [].}
cb = proc(udata: pointer) {.gcsafe, raises: [].} =
2019-06-20 20:30:41 +00:00
if not(retFuture.finished()):
var fut = cast[FutureBase](udata)
if cast[pointer](fut1) == udata:
2018-05-16 08:22:34 +00:00
fut2.removeCallback(cb)
else:
fut1.removeCallback(cb)
if fut.failed():
2023-10-24 14:21:07 +00:00
retFuture.fail(fut.error, warn = false)
else:
retFuture.complete()
2019-06-20 20:30:41 +00:00
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
proc cancellation(udata: pointer) =
# On cancel we remove all our callbacks only.
if not(fut1.finished()):
fut1.removeCallback(cb)
if not(fut2.finished()):
fut2.removeCallback(cb)
2019-06-20 20:30:41 +00:00
if fut1.finished():
if fut1.failed():
2023-10-24 14:21:07 +00:00
retFuture.fail(fut1.error, warn = false)
else:
retFuture.complete()
return retFuture
if fut2.finished():
if fut2.failed():
2023-10-24 14:21:07 +00:00
retFuture.fail(fut2.error, warn = false)
else:
retFuture.complete()
return retFuture
fut1.addCallback(cb)
fut2.addCallback(cb)
retFuture.cancelCallback = cancellation
2018-05-16 08:22:34 +00:00
return retFuture
2023-10-24 14:21:07 +00:00
proc `or`*[T, Y](fut1: Future[T], fut2: Future[Y]): Future[void] =
## Returns a future which will complete once either ``fut1`` or ``fut2``
## finish.
##
## If ``fut1`` or ``fut2`` future is failed, the result future will also be
## failed with an error stored in ``fut1`` or ``fut2`` respectively.
##
## If both ``fut1`` and ``fut2`` future are completed or failed, the result
## future will depend on the state of ``fut1`` future. So if ``fut1`` future
## is failed, the result future will also be failed, if ``fut1`` future is
## completed, the result future will also be completed.
##
## If cancelled, ``fut1`` and ``fut2`` futures WILL NOT BE cancelled.
var retFuture = newFuture[void]("chronos.or")
orImpl(fut1, fut2)
proc all*[T](futs: varargs[Future[T]]): auto {.
deprecated: "Use allFutures(varargs[Future[T]])".} =
## Returns a future which will complete once all futures in ``futs`` finish.
2018-05-16 08:22:34 +00:00
## If the argument is empty, the returned future completes immediately.
##
## If the awaited futures are not ``Future[void]``, the returned future
## will hold the values of all awaited futures in a sequence.
##
2019-03-15 00:43:51 +00:00
## If the awaited futures *are* ``Future[void]``, this proc returns
## ``Future[void]``.
##
## Note, that if one of the futures in ``futs`` will fail, result of ``all()``
## will also be failed with error from failed future.
2019-06-20 20:30:41 +00:00
##
## TODO: This procedure has bug on handling cancelled futures from ``futs``.
## So if future from ``futs`` list become cancelled, what must be returned?
## You can't cancel result ``retFuture`` because in such way infinite
## recursion will happen.
2019-03-15 00:43:51 +00:00
let totalFutures = len(futs)
var completedFutures = 0
2018-05-16 08:22:34 +00:00
2019-03-15 00:43:51 +00:00
# Because we can't capture varargs[T] in closures we need to create copy.
var nfuts = @futs
2018-05-16 08:22:34 +00:00
2019-03-15 00:43:51 +00:00
when T is void:
var retFuture = newFuture[void]("chronos.all(void)")
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
proc cb(udata: pointer) =
2019-06-20 20:30:41 +00:00
if not(retFuture.finished()):
2018-05-16 08:22:34 +00:00
inc(completedFutures)
2019-06-20 20:30:41 +00:00
if completedFutures == totalFutures:
for nfut in nfuts:
if nfut.failed():
retFuture.fail(nfut.error)
break
if not(retFuture.failed()):
retFuture.complete()
for fut in nfuts:
fut.addCallback(cb)
2018-05-16 08:22:34 +00:00
2019-03-15 00:43:51 +00:00
if len(nfuts) == 0:
2018-05-16 08:22:34 +00:00
retFuture.complete()
return retFuture
else:
var retFuture = newFuture[seq[T]]("chronos.all(T)")
2019-03-15 00:43:51 +00:00
var retValues = newSeq[T](totalFutures)
2019-06-20 20:30:41 +00:00
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
proc cb(udata: pointer) =
2019-06-20 20:30:41 +00:00
if not(retFuture.finished()):
2019-03-15 00:43:51 +00:00
inc(completedFutures)
2019-06-20 20:30:41 +00:00
if completedFutures == totalFutures:
for k, nfut in nfuts:
if nfut.failed():
retFuture.fail(nfut.error)
break
else:
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
retValues[k] = nfut.value
2019-06-20 20:30:41 +00:00
if not(retFuture.failed()):
retFuture.complete(retValues)
for fut in nfuts:
fut.addCallback(cb)
2019-03-15 00:43:51 +00:00
if len(nfuts) == 0:
2018-05-16 08:22:34 +00:00
retFuture.complete(retValues)
return retFuture
proc oneIndex*[T](futs: varargs[Future[T]]): Future[int] {.
deprecated: "Use one[T](varargs[Future[T]])".} =
## Returns a future which will complete once one of the futures in ``futs``
## complete.
##
## If the argument is empty, the returned future FAILS immediately.
##
## Returned future will hold index of completed/failed future in ``futs``
## argument.
var nfuts = @futs
var retFuture = newFuture[int]("chronos.oneIndex(T)")
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
proc cb(udata: pointer) =
var res = -1
2019-06-20 20:30:41 +00:00
if not(retFuture.finished()):
var rfut = cast[FutureBase](udata)
for i in 0..<len(nfuts):
if cast[FutureBase](nfuts[i]) != rfut:
nfuts[i].removeCallback(cb)
else:
res = i
retFuture.complete(res)
for fut in nfuts:
fut.addCallback(cb)
if len(nfuts) == 0:
2019-06-25 07:50:56 +00:00
retFuture.fail(newException(ValueError, "Empty Future[T] list"))
return retFuture
proc oneValue*[T](futs: varargs[Future[T]]): Future[T] {.
deprecated: "Use one[T](varargs[Future[T]])".} =
## Returns a future which will finish once one of the futures in ``futs``
## finish.
##
## If the argument is empty, returned future FAILS immediately.
##
## Returned future will hold value of completed ``futs`` future, or error
## if future was failed.
var nfuts = @futs
var retFuture = newFuture[T]("chronos.oneValue(T)")
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
proc cb(udata: pointer) =
var resFut: Future[T]
2019-06-20 20:30:41 +00:00
if not(retFuture.finished()):
var rfut = cast[FutureBase](udata)
for i in 0..<len(nfuts):
if cast[FutureBase](nfuts[i]) != rfut:
nfuts[i].removeCallback(cb)
else:
resFut = nfuts[i]
2019-06-20 20:30:41 +00:00
if resFut.failed():
retFuture.fail(resFut.error)
else:
when T is void:
retFuture.complete()
else:
retFuture.complete(resFut.read())
for fut in nfuts:
fut.addCallback(cb)
if len(nfuts) == 0:
2019-06-25 07:50:56 +00:00
retFuture.fail(newException(ValueError, "Empty Future[T] list"))
2019-06-20 20:30:41 +00:00
return retFuture
proc cancelSoon(future: FutureBase, aftercb: CallbackFunc, udata: pointer,
2023-10-24 14:21:07 +00:00
loc: ptr SrcLoc) {.raises: [].} =
## Perform cancellation ``future`` and call ``aftercb`` callback when
## ``future`` become finished (completed with value, failed or cancelled).
##
## NOTE: Compared to the `tryCancel()` call, this procedure call guarantees
## that ``future``will be finished (completed with value, failed or cancelled)
## as quickly as possible.
proc checktick(udata: pointer) {.gcsafe, raises: [].} =
# We trying to cancel Future on more time, and if `cancel()` succeeds we
# return early.
if tryCancel(future, loc):
return
# Cancellation signal was not delivered, so we trying to deliver it one
# more time after one tick. But we need to check situation when child
# future was finished but our completion callback is not yet invoked.
if not(future.finished()):
internalCallTick(checktick)
proc continuation(udata: pointer) {.gcsafe.} =
# We do not use `callSoon` here because we was just scheduled from `poll()`.
if not(isNil(aftercb)):
aftercb(udata)
if future.finished():
# We could not schedule callback directly otherwise we could fall into
# recursion problem.
if not(isNil(aftercb)):
let loop = getThreadDispatcher()
loop.callbacks.addLast(AsyncCallback(function: aftercb, udata: udata))
return
future.addCallback(continuation)
# Initiate cancellation process.
if not(tryCancel(future, loc)):
# Cancellation signal was not delivered, so we trying to deliver it one
# more time after async tick. But we need to check case, when future was
# finished but our completion callback is not yet invoked.
if not(future.finished()):
internalCallTick(checktick)
template cancelSoon*(fut: FutureBase, cb: CallbackFunc, udata: pointer) =
cancelSoon(fut, cb, udata, getSrcLocation())
template cancelSoon*(fut: FutureBase, cb: CallbackFunc) =
cancelSoon(fut, cb, nil, getSrcLocation())
template cancelSoon*(fut: FutureBase, acb: AsyncCallback) =
cancelSoon(fut, acb.function, acb.udata, getSrcLocation())
template cancelSoon*(fut: FutureBase) =
cancelSoon(fut, nil, nil, getSrcLocation())
template cancel*(future: FutureBase) {.
deprecated: "Please use cancelSoon() or cancelAndWait() instead".} =
## Cancel ``future``.
cancelSoon(future, nil, nil, getSrcLocation())
2023-10-24 14:21:07 +00:00
proc cancelAndWait*(future: FutureBase, loc: ptr SrcLoc): Future[void] {.
async: (raw: true, raises: [CancelledError]).} =
## Perform cancellation ``future`` return Future which will be completed when
## ``future`` become finished (completed with value, failed or cancelled).
##
## NOTE: Compared to the `tryCancel()` call, this procedure call guarantees
## that ``future``will be finished (completed with value, failed or cancelled)
## as quickly as possible.
let retFuture = newFuture[void]("chronos.cancelAndWait(FutureBase)",
{FutureFlag.OwnCancelSchedule})
proc continuation(udata: pointer) {.gcsafe.} =
retFuture.complete()
if future.finished():
retFuture.complete()
else:
cancelSoon(future, continuation, cast[pointer](retFuture), loc)
retFuture
template cancelAndWait*(future: FutureBase): Future[void].Raising([CancelledError]) =
## Cancel ``future``.
cancelAndWait(future, getSrcLocation())
proc noCancel*[F: SomeFuture](future: F): auto = # async: (raw: true, raises: asyncraiseOf(future) - CancelledError
## Prevent cancellation requests from propagating to ``future`` while
## forwarding its value or error when it finishes.
##
## This procedure should be used when you need to perform operations which
## should not be cancelled at all cost, for example closing sockets, pipes,
## connections or servers. Usually it become useful in exception or finally
## blocks.
2023-10-24 14:21:07 +00:00
when F is InternalRaisesFuture:
type
E = F.E
InternalRaisesFutureRaises = E.remove(CancelledError)
let retFuture = newFuture[F.T]("chronos.noCancel(T)",
{FutureFlag.OwnCancelSchedule})
template completeFuture() =
if future.completed():
2023-10-24 14:21:07 +00:00
when F.T is void:
retFuture.complete()
else:
retFuture.complete(future.value)
elif future.failed():
2023-10-24 14:21:07 +00:00
when F is Future:
retFuture.fail(future.error, warn = false)
when declared(InternalRaisesFutureRaises):
when InternalRaisesFutureRaises isnot void:
retFuture.fail(future.error, warn = false)
else:
raiseAssert("Unexpected future state [" & $future.state & "]")
proc continuation(udata: pointer) {.gcsafe.} =
completeFuture()
if future.finished():
completeFuture()
else:
future.addCallback(continuation)
retFuture
2023-10-24 14:21:07 +00:00
proc allFutures*(futs: varargs[FutureBase]): Future[void] {.
async: (raw: true, raises: [CancelledError]).} =
## Returns a future which will complete only when all futures in ``futs``
## will be completed, failed or canceled.
##
## If the argument is empty, the returned future COMPLETES immediately.
##
## On cancel all the awaited futures ``futs`` WILL NOT BE cancelled.
var retFuture = newFuture[void]("chronos.allFutures()")
let totalFutures = len(futs)
var finishedFutures = 0
# Because we can't capture varargs[T] in closures we need to create copy.
var nfuts = @futs
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
proc cb(udata: pointer) =
if not(retFuture.finished()):
inc(finishedFutures)
if finishedFutures == totalFutures:
retFuture.complete()
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
proc cancellation(udata: pointer) =
# On cancel we remove all our callbacks only.
for i in 0..<len(nfuts):
if not(nfuts[i].finished()):
nfuts[i].removeCallback(cb)
for fut in nfuts:
if not(fut.finished()):
fut.addCallback(cb)
else:
inc(finishedFutures)
retFuture.cancelCallback = cancellation
if len(nfuts) == 0 or len(nfuts) == finishedFutures:
retFuture.complete()
retFuture
2023-10-24 14:21:07 +00:00
proc allFutures*[T](futs: varargs[Future[T]]): Future[void] {.
async: (raw: true, raises: [CancelledError]).} =
2022-09-12 12:11:13 +00:00
## Returns a future which will complete only when all futures in ``futs``
## will be completed, failed or canceled.
##
## If the argument is empty, the returned future COMPLETES immediately.
##
## On cancel all the awaited futures ``futs`` WILL NOT BE cancelled.
# Because we can't capture varargs[T] in closures we need to create copy.
var nfuts: seq[FutureBase]
for future in futs:
nfuts.add(future)
2022-09-12 12:11:13 +00:00
allFutures(nfuts)
2023-10-24 14:21:07 +00:00
proc allFinished*[F: SomeFuture](futs: varargs[F]): Future[seq[F]] {.
async: (raw: true, raises: [CancelledError]).} =
## Returns a future which will complete only when all futures in ``futs``
## will be completed, failed or canceled.
##
## Returned sequence will hold all the Future[T] objects passed to
## ``allFinished`` with the order preserved.
##
## If the argument is empty, the returned future COMPLETES immediately.
##
## On cancel all the awaited futures ``futs`` WILL NOT BE cancelled.
2023-10-24 14:21:07 +00:00
var retFuture = newFuture[seq[F]]("chronos.allFinished()")
let totalFutures = len(futs)
var finishedFutures = 0
var nfuts = @futs
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
proc cb(udata: pointer) =
if not(retFuture.finished()):
inc(finishedFutures)
if finishedFutures == totalFutures:
retFuture.complete(nfuts)
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
proc cancellation(udata: pointer) =
# On cancel we remove all our callbacks only.
for fut in nfuts.mitems():
if not(fut.finished()):
fut.removeCallback(cb)
for fut in nfuts:
if not(fut.finished()):
fut.addCallback(cb)
else:
inc(finishedFutures)
retFuture.cancelCallback = cancellation
if len(nfuts) == 0 or len(nfuts) == finishedFutures:
retFuture.complete(nfuts)
return retFuture
2023-10-24 14:21:07 +00:00
proc one*[F: SomeFuture](futs: varargs[F]): Future[F] {.
async: (raw: true, raises: [ValueError, CancelledError]).} =
## Returns a future which will complete and return completed Future[T] inside,
## when one of the futures in ``futs`` will be completed, failed or canceled.
##
## If the argument is empty, the returned future FAILS immediately.
##
## On success returned Future will hold finished Future[T].
##
## On cancel futures in ``futs`` WILL NOT BE cancelled.
2023-10-24 14:21:07 +00:00
var retFuture = newFuture[F]("chronos.one()")
if len(futs) == 0:
retFuture.fail(newException(ValueError, "Empty Future[T] list"))
return retFuture
# If one of the Future[T] already finished we return it as result
for fut in futs:
if fut.finished():
retFuture.complete(fut)
return retFuture
# Because we can't capture varargs[T] in closures we need to create copy.
var nfuts = @futs
2023-06-05 20:21:50 +00:00
var cb: proc(udata: pointer) {.gcsafe, raises: [].}
cb = proc(udata: pointer) {.gcsafe, raises: [].} =
if not(retFuture.finished()):
2023-10-24 14:21:07 +00:00
var res: F
var rfut = cast[FutureBase](udata)
for i in 0..<len(nfuts):
if cast[FutureBase](nfuts[i]) != rfut:
nfuts[i].removeCallback(cb)
else:
res = nfuts[i]
retFuture.complete(res)
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
proc cancellation(udata: pointer) =
# On cancel we remove all our callbacks only.
for i in 0..<len(nfuts):
if not(nfuts[i].finished()):
nfuts[i].removeCallback(cb)
for fut in nfuts:
fut.addCallback(cb)
retFuture.cancelCallback = cancellation
return retFuture
2023-10-24 14:21:07 +00:00
proc race*(futs: varargs[FutureBase]): Future[FutureBase] {.
async: (raw: true, raises: [ValueError, CancelledError]).} =
## Returns a future which will complete and return completed FutureBase,
## when one of the futures in ``futs`` will be completed, failed or canceled.
##
## If the argument is empty, the returned future FAILS immediately.
##
## On success returned Future will hold finished FutureBase.
##
## On cancel futures in ``futs`` WILL NOT BE cancelled.
let retFuture = newFuture[FutureBase]("chronos.race()")
if len(futs) == 0:
retFuture.fail(newException(ValueError, "Empty Future[T] list"))
return retFuture
# If one of the Future[T] already finished we return it as result
for fut in futs:
if fut.finished():
retFuture.complete(fut)
return retFuture
# Because we can't capture varargs[T] in closures we need to create copy.
var nfuts = @futs
2023-06-05 20:21:50 +00:00
var cb: proc(udata: pointer) {.gcsafe, raises: [].}
cb = proc(udata: pointer) {.gcsafe, raises: [].} =
if not(retFuture.finished()):
var res: FutureBase
var rfut = cast[FutureBase](udata)
for i in 0..<len(nfuts):
if nfuts[i] != rfut:
nfuts[i].removeCallback(cb)
else:
res = nfuts[i]
retFuture.complete(res)
exception tracking (#166) * exception tracking This PR adds minimal exception tracking to chronos, moving the goalpost one step further. In particular, it becomes invalid to raise exceptions from `callSoon` callbacks: this is critical for writing correct error handling because there's no reasonable way that a user of chronos can possibly _reason_ about exceptions coming out of there: the event loop will be in an indeterminite state when the loop is executing an _random_ callback. As expected, there are several issues in the error handling of chronos: in particular, it will end up in an inconsistent internal state whenever the selector loop operations fail, because the internal state update functions are not written in an exception-safe way. This PR turns this into a Defect, which probably is not the optimal way of handling things - expect more work to be done here. Some API have no way of reporting back errors to callers - for example, when something fails in the accept loop, there's not much it can do, and no way to report it back to the user of the API - this has been fixed with the new accept flow - the old one should be deprecated. Finally, there is information loss in the API: in composite operations like `poll` and `waitFor` there's no way to differentiate internal errors from user-level errors originating from callbacks. * store `CatchableError` in future * annotate proc's with correct raises information * `selectors2` to avoid non-CatchableError IOSelectorsException * `$` should never raise * remove unnecessary gcsafe annotations * fix exceptions leaking out of timer waits * fix some imports * functions must signal raising the union of all exceptions across all platforms to enable cross-platform code * switch to unittest2 * add `selectors2` which supercedes the std library version and fixes several exception handling issues in there * fixes * docs, platform-independent eh specifiers for some functions * add feature flag for strict exception mode also bump version to 3.0.0 - _most_ existing code should be compatible with this version of exception handling but some things might need fixing - callbacks, existing raises specifications etc. * fix AsyncCheck for non-void T
2021-03-24 09:08:33 +00:00
proc cancellation(udata: pointer) =
# On cancel we remove all our callbacks only.
for i in 0..<len(nfuts):
if not(nfuts[i].finished()):
nfuts[i].removeCallback(cb)
for fut in nfuts:
fut.addCallback(cb, cast[pointer](fut))
retFuture.cancelCallback = cancellation
return retFuture
when (chronosEventEngine in ["epoll", "kqueue"]) or defined(windows):
import std/os
proc waitSignal*(signal: int): Future[void] {.
async: (raw: true, raises: [AsyncError, CancelledError]).} =
var retFuture = newFuture[void]("chronos.waitSignal()")
var signalHandle: Opt[SignalHandle]
template getSignalException(e: OSErrorCode): untyped =
newException(AsyncError, "Could not manipulate signal handler, " &
"reason [" & $int(e) & "]: " & osErrorMsg(e))
proc continuation(udata: pointer) {.gcsafe.} =
if not(retFuture.finished()):
if signalHandle.isSome():
let res = removeSignal2(signalHandle.get())
if res.isErr():
retFuture.fail(getSignalException(res.error()))
else:
retFuture.complete()
proc cancellation(udata: pointer) {.gcsafe.} =
if not(retFuture.finished()):
if signalHandle.isSome():
let res = removeSignal2(signalHandle.get())
if res.isErr():
retFuture.fail(getSignalException(res.error()))
signalHandle =
block:
let res = addSignal2(signal, continuation)
if res.isErr():
retFuture.fail(getSignalException(res.error()))
Opt.some(res.get())
retFuture.cancelCallback = cancellation
retFuture
2023-10-24 14:21:07 +00:00
proc sleepAsync*(duration: Duration): Future[void] {.
async: (raw: true, raises: [CancelledError]).} =
## Suspends the execution of the current async procedure for the next
## ``duration`` time.
var retFuture = newFuture[void]("chronos.sleepAsync(Duration)")
let moment = Moment.fromNow(duration)
var timer: TimerCallback
proc completion(data: pointer) {.gcsafe.} =
if not(retFuture.finished()):
retFuture.complete()
proc cancellation(udata: pointer) {.gcsafe.} =
if not(retFuture.finished()):
clearTimer(timer)
retFuture.cancelCallback = cancellation
timer = setTimer(moment, completion, cast[pointer](retFuture))
return retFuture
proc sleepAsync*(ms: int): Future[void] {.
inline, deprecated: "Use sleepAsync(Duration)",
async: (raw: true, raises: [CancelledError]).} =
result = sleepAsync(ms.milliseconds())
proc stepsAsync*(number: int): Future[void] {.
async: (raw: true, raises: [CancelledError]).} =
## Suspends the execution of the current async procedure for the next
## ``number`` of asynchronous steps (``poll()`` calls).
##
## This primitive can be useful when you need to create more deterministic
## tests and cases.
doAssert(number > 0, "Number should be positive integer")
var
retFuture = newFuture[void]("chronos.stepsAsync(int)")
counter = 0
continuation: proc(data: pointer) {.gcsafe, raises: [].}
continuation = proc(data: pointer) {.gcsafe, raises: [].} =
if not(retFuture.finished()):
inc(counter)
if counter < number:
internalCallTick(continuation)
else:
retFuture.complete()
if number <= 0:
retFuture.complete()
else:
internalCallTick(continuation)
retFuture
proc idleAsync*(): Future[void] {.
async: (raw: true, raises: [CancelledError]).} =
## Suspends the execution of the current asynchronous task until "idle" time.
##
## "idle" time its moment of time, when no network events were processed by
## ``poll()`` call.
var retFuture = newFuture[void]("chronos.idleAsync()")
proc continuation(data: pointer) {.gcsafe.} =
if not(retFuture.finished()):
retFuture.complete()
proc cancellation(udata: pointer) {.gcsafe.} =
discard
retFuture.cancelCallback = cancellation
callIdle(continuation, nil)
retFuture
2023-10-24 14:21:07 +00:00
proc withTimeout*[T](fut: Future[T], timeout: Duration): Future[bool] {.
async: (raw: true, raises: [CancelledError]).} =
## Returns a future which will complete once ``fut`` completes or after
## ``timeout`` milliseconds has elapsed.
##
## If ``fut`` completes first the returned future will hold true,
## otherwise, if ``timeout`` milliseconds has elapsed first, the returned
## future will hold false.
var
retFuture = newFuture[bool]("chronos.withTimeout",
{FutureFlag.OwnCancelSchedule})
moment: Moment
timer: TimerCallback
timeouted = false
template completeFuture(fut: untyped): untyped =
if fut.failed() or fut.completed():
retFuture.complete(true)
else:
retFuture.cancelAndSchedule()
# TODO: raises annotation shouldn't be needed, but likely similar issue as
# https://github.com/nim-lang/Nim/issues/17369
proc continuation(udata: pointer) {.gcsafe, raises: [].} =
if not(retFuture.finished()):
if timeouted:
retFuture.complete(false)
return
if not(fut.finished()):
# Timer exceeded first, we going to cancel `fut` and wait until it
# not completes.
timeouted = true
fut.cancelSoon()
else:
# Future `fut` completed/failed/cancelled first.
if not(isNil(timer)):
clearTimer(timer)
fut.completeFuture()
# TODO: raises annotation shouldn't be needed, but likely similar issue as
# https://github.com/nim-lang/Nim/issues/17369
proc cancellation(udata: pointer) {.gcsafe, raises: [].} =
if not(fut.finished()):
if not isNil(timer):
clearTimer(timer)
fut.cancelSoon()
else:
fut.completeFuture()
if fut.finished():
retFuture.complete(true)
else:
if timeout.isZero():
retFuture.complete(false)
elif timeout.isInfinite():
retFuture.cancelCallback = cancellation
fut.addCallback(continuation)
else:
moment = Moment.fromNow(timeout)
retFuture.cancelCallback = cancellation
timer = setTimer(moment, continuation, nil)
fut.addCallback(continuation)
retFuture
proc withTimeout*[T](fut: Future[T], timeout: int): Future[bool] {.
inline, deprecated: "Use withTimeout(Future[T], Duration)".} =
withTimeout(fut, timeout.milliseconds())
2023-10-24 14:21:07 +00:00
proc waitImpl[F: SomeFuture](fut: F, retFuture: auto, timeout: Duration): auto =
var
moment: Moment
timer: TimerCallback
timeouted = false
template completeFuture(fut: untyped): untyped =
if fut.failed():
2023-10-24 14:21:07 +00:00
retFuture.fail(fut.error(), warn = false)
elif fut.cancelled():
retFuture.cancelAndSchedule()
else:
2023-10-24 14:21:07 +00:00
when type(fut).T is void:
retFuture.complete()
else:
retFuture.complete(fut.value)
proc continuation(udata: pointer) {.raises: [].} =
if not(retFuture.finished()):
if timeouted:
retFuture.fail(newException(AsyncTimeoutError, "Timeout exceeded!"))
return
if not(fut.finished()):
# Timer exceeded first.
timeouted = true
fut.cancelSoon()
else:
# Future `fut` completed/failed/cancelled first.
if not(isNil(timer)):
clearTimer(timer)
fut.completeFuture()
var cancellation: proc(udata: pointer) {.gcsafe, raises: [].}
cancellation = proc(udata: pointer) {.gcsafe, raises: [].} =
if not(fut.finished()):
if not(isNil(timer)):
clearTimer(timer)
fut.cancelSoon()
else:
fut.completeFuture()
if fut.finished():
fut.completeFuture()
else:
if timeout.isZero():
retFuture.fail(newException(AsyncTimeoutError, "Timeout exceeded!"))
elif timeout.isInfinite():
retFuture.cancelCallback = cancellation
fut.addCallback(continuation)
else:
moment = Moment.fromNow(timeout)
retFuture.cancelCallback = cancellation
timer = setTimer(moment, continuation, nil)
fut.addCallback(continuation)
retFuture
2023-10-24 14:21:07 +00:00
proc wait*[T](fut: Future[T], timeout = InfiniteDuration): Future[T] =
## Returns a future which will complete once future ``fut`` completes
## or if timeout of ``timeout`` milliseconds has been expired.
##
## If ``timeout`` is ``-1``, then statement ``await wait(fut)`` is
## equal to ``await fut``.
##
## TODO: In case when ``fut`` got cancelled, what result Future[T]
## should return, because it can't be cancelled too.
var
retFuture = newFuture[T]("chronos.wait()", {FutureFlag.OwnCancelSchedule})
waitImpl(fut, retFuture, timeout)
proc wait*[T](fut: Future[T], timeout = -1): Future[T] {.
inline, deprecated: "Use wait(Future[T], Duration)".} =
if timeout == -1:
wait(fut, InfiniteDuration)
elif timeout == 0:
wait(fut, ZeroDuration)
else:
wait(fut, timeout.milliseconds())
when defined(windows):
import ../osdefs
proc waitForSingleObject*(handle: HANDLE,
timeout: Duration): Future[WaitableResult] {.
raises: [].} =
## Waits until the specified object is in the signaled state or the
## time-out interval elapses. WaitForSingleObject() for asynchronous world.
let flags = WT_EXECUTEONLYONCE
var
retFuture = newFuture[WaitableResult]("chronos.waitForSingleObject()")
waitHandle: WaitableHandle = nil
proc continuation(udata: pointer) {.gcsafe.} =
doAssert(not(isNil(waitHandle)))
if not(retFuture.finished()):
let
ovl = cast[PtrCustomOverlapped](udata)
returnFlag = WINBOOL(ovl.data.bytesCount)
res = closeWaitable(waitHandle)
if res.isErr():
retFuture.fail(newException(AsyncError, osErrorMsg(res.error())))
else:
if returnFlag == TRUE:
retFuture.complete(WaitableResult.Timeout)
else:
retFuture.complete(WaitableResult.Ok)
proc cancellation(udata: pointer) {.gcsafe.} =
doAssert(not(isNil(waitHandle)))
if not(retFuture.finished()):
discard closeWaitable(waitHandle)
let wres = uint32(waitForSingleObject(handle, DWORD(0)))
if wres == WAIT_OBJECT_0:
retFuture.complete(WaitableResult.Ok)
return retFuture
elif wres == WAIT_ABANDONED:
retFuture.fail(newException(AsyncError, "Handle was abandoned"))
return retFuture
elif wres == WAIT_FAILED:
retFuture.fail(newException(AsyncError, osErrorMsg(osLastError())))
return retFuture
if timeout == ZeroDuration:
retFuture.complete(WaitableResult.Timeout)
return retFuture
waitHandle =
block:
let res = registerWaitable(handle, flags, timeout, continuation, nil)
if res.isErr():
retFuture.fail(newException(AsyncError, osErrorMsg(res.error())))
return retFuture
res.get()
retFuture.cancelCallback = cancellation
return retFuture
2023-10-24 14:21:07 +00:00
{.pop.} # Automatically deduced raises from here onwards
proc readFinished[T: not void; E](fut: InternalRaisesFuture[T, E]): lent T =
internalCheckComplete(fut, E)
fut.internalValue
proc read*[T: not void, E](fut: InternalRaisesFuture[T, E]): lent T = # {.raises: [E, FuturePendingError].}
## Retrieves the value of `fut`.
##
## If the future failed or was cancelled, the corresponding exception will be
## raised.
##
## If the future is still pending, `FuturePendingError` will be raised.
if not fut.finished():
raiseFuturePendingError(fut)
2023-10-24 14:21:07 +00:00
fut.readFinished()
2023-10-24 14:21:07 +00:00
proc read*[E](fut: InternalRaisesFuture[void, E]) = # {.raises: [E].}
## Checks that `fut` completed.
2023-10-24 14:21:07 +00:00
##
## If the future failed or was cancelled, the corresponding exception will be
## raised.
##
## If the future is still pending, `FuturePendingError` will be raised.
if not fut.finished():
raiseFuturePendingError(fut)
2023-10-24 14:21:07 +00:00
internalCheckComplete(fut, E)
2023-10-24 14:21:07 +00:00
proc waitFor*[T: not void; E](fut: InternalRaisesFuture[T, E]): lent T = # {.raises: [E]}
## Blocks the current thread of execution until `fut` has finished, returning
## its value.
2023-10-24 14:21:07 +00:00
##
## If the future failed or was cancelled, the corresponding exception will be
## raised.
##
## Must not be called recursively (from inside `async` procedures).
##
## See also `await`, `Future.read`
pollFor(fut).readFinished()
proc waitFor*[E](fut: InternalRaisesFuture[void, E]) = # {.raises: [E]}
## Blocks the current thread of execution until `fut` has finished.
##
## If the future failed or was cancelled, the corresponding exception will be
## raised.
##
## Must not be called recursively (from inside `async` procedures).
##
## See also `await`, `Future.read`
pollFor(fut).internalCheckComplete(E)
2023-10-24 14:21:07 +00:00
proc `or`*[T, Y, E1, E2](
fut1: InternalRaisesFuture[T, E1],
fut2: InternalRaisesFuture[Y, E2]): auto =
type
InternalRaisesFutureRaises = union(E1, E2)
let
retFuture = newFuture[void]("chronos.wait()", {FutureFlag.OwnCancelSchedule})
orImpl(fut1, fut2)
proc wait*(fut: InternalRaisesFuture, timeout = InfiniteDuration): auto =
type
T = type(fut).T
E = type(fut).E
InternalRaisesFutureRaises = E.prepend(CancelledError, AsyncTimeoutError)
let
retFuture = newFuture[T]("chronos.wait()", {FutureFlag.OwnCancelSchedule})
waitImpl(fut, retFuture, timeout)