Copyright | Will Thompson Iñaki García Etxebarria and Jonas Platte |
---|---|
License | LGPL-2.1 |
Maintainer | Iñaki García Etxebarria |
Safe Haskell | None |
Language | Haskell2010 |
The Cond
struct is an opaque data structure that represents a
condition. Threads can block on a Cond
if they find a certain
condition to be false. If other threads change the state of this
condition they signal the Cond
, and that causes the waiting
threads to be woken up.
Consider the following example of a shared variable. One or more threads can wait for data to be published to the variable and when another thread publishes the data, it can signal one of the waiting threads to wake up to collect the data.
Here is an example for using GCond to block a thread until a condition is satisfied:
C code
gpointer current_data = NULL; GMutex data_mutex; GCond data_cond; void push_data (gpointer data) { g_mutex_lock (&data_mutex); current_data = data; g_cond_signal (&data_cond); g_mutex_unlock (&data_mutex); } gpointer pop_data (void) { gpointer data; g_mutex_lock (&data_mutex); while (!current_data) g_cond_wait (&data_cond, &data_mutex); data = current_data; current_data = NULL; g_mutex_unlock (&data_mutex); return data; }
Whenever a thread calls pop_data()
now, it will wait until
current_data is non-Nothing
, i.e. until some other thread
has called push_data()
.
The example shows that use of a condition variable must always be
paired with a mutex. Without the use of a mutex, there would be a
race between the check of currentData
by the while loop in
pop_data()
and waiting. Specifically, another thread could set
currentData
after the check, and signal the cond (with nobody
waiting on it) before the first thread goes to sleep. Cond
is
specifically useful for its ability to release the mutex and go
to sleep atomically.
It is also important to use the condWait
and condWaitUntil
functions only inside a loop which checks for the condition to be
true. See condWait
for an explanation of why the condition may
not be true even after it returns.
If a Cond
is allocated in static storage then it can be used
without initialisation. Otherwise, you should call condInit
on it and condClear
when done.
A Cond
should only be accessed via the g_cond_ functions.
Synopsis
- newtype Cond = Cond (ManagedPtr Cond)
- newZeroCond :: MonadIO m => m Cond
- condBroadcast :: (HasCallStack, MonadIO m) => Cond -> m ()
- condClear :: (HasCallStack, MonadIO m) => Cond -> m ()
- condInit :: (HasCallStack, MonadIO m) => Cond -> m ()
- condSignal :: (HasCallStack, MonadIO m) => Cond -> m ()
- condWait :: (HasCallStack, MonadIO m) => Cond -> Mutex -> m ()
- condWaitUntil :: (HasCallStack, MonadIO m) => Cond -> Mutex -> Int64 -> m Bool
Exported types
Memory-managed wrapper type.
Instances
Eq Cond Source # | |
ManagedPtrNewtype Cond Source # | |
Defined in GI.GLib.Structs.Cond toManagedPtr :: Cond -> ManagedPtr Cond # | |
BoxedPtr Cond Source # | |
Defined in GI.GLib.Structs.Cond boxedPtrCopy :: Cond -> IO Cond # boxedPtrFree :: Cond -> IO () # | |
CallocPtr Cond Source # | |
Defined in GI.GLib.Structs.Cond boxedPtrCalloc :: IO (Ptr Cond) # | |
tag ~ 'AttrSet => Constructible Cond tag Source # | |
Defined in GI.GLib.Structs.Cond |
Methods
Overloaded methods
broadcast
:: (HasCallStack, MonadIO m) | |
=> Cond |
|
-> m () |
If threads are waiting for cond
, all of them are unblocked.
If no threads are waiting for cond
, this function has no effect.
It is good practice to lock the same mutex as the waiting threads
while calling this function, though not required.
clear
:: (HasCallStack, MonadIO m) | |
=> Cond |
|
-> m () |
init
:: (HasCallStack, MonadIO m) | |
=> Cond |
|
-> m () |
Initialises a Cond
so that it can be used.
This function is useful to initialise a Cond
that has been
allocated as part of a larger structure. It is not necessary to
initialise a Cond
that has been statically allocated.
To undo the effect of condInit
when a Cond
is no longer
needed, use condClear
.
Calling condInit
on an already-initialised Cond
leads
to undefined behaviour.
Since: 2.32
signal
:: (HasCallStack, MonadIO m) | |
=> Cond |
|
-> m () |
If threads are waiting for cond
, at least one of them is unblocked.
If no threads are waiting for cond
, this function has no effect.
It is good practice to hold the same lock as the waiting thread
while calling this function, though not required.
wait
:: (HasCallStack, MonadIO m) | |
=> Cond |
|
-> Mutex |
|
-> m () |
Atomically releases mutex
and waits until cond
is signalled.
When this function returns, mutex
is locked again and owned by the
calling thread.
When using condition variables, it is possible that a spurious wakeup
may occur (ie: condWait
returns even though condSignal
was
not called). It's also possible that a stolen wakeup may occur.
This is when condSignal
is called, but another thread acquires
mutex
before this thread and modifies the state of the program in
such a way that when condWait
is able to return, the expected
condition is no longer met.
For this reason, condWait
must always be used in a loop. See
the documentation for Cond
for a complete example.
waitUntil
:: (HasCallStack, MonadIO m) | |
=> Cond |
|
-> Mutex |
|
-> Int64 |
|
-> m Bool | Returns: |
Waits until either cond
is signalled or endTime
has passed.
As with condWait
it is possible that a spurious or stolen wakeup
could occur. For that reason, waiting on a condition variable should
always be in a loop, based on an explicitly-checked predicate.
True
is returned if the condition variable was signalled (or in the
case of a spurious wakeup). False
is returned if endTime
has
passed.
The following code shows how to correctly perform a timed wait on a
condition variable (extending the example presented in the
documentation for Cond
):
C code
gpointer pop_data_timed (void) { gint64 end_time; gpointer data; g_mutex_lock (&data_mutex); end_time = g_get_monotonic_time () + 5 * G_TIME_SPAN_SECOND; while (!current_data) if (!g_cond_wait_until (&data_cond, &data_mutex, end_time)) { // timeout has passed. g_mutex_unlock (&data_mutex); return NULL; } // there is data for us data = current_data; current_data = NULL; g_mutex_unlock (&data_mutex); return data; }
Notice that the end time is calculated once, before entering the loop and reused. This is the motivation behind the use of absolute time on this API -- if a relative time of 5 seconds were passed directly to the call and a spurious wakeup occurred, the program would have to start over waiting again (which would lead to a total wait time of more than 5 seconds).
Since: 2.32