Name
msys.runInPool — Run a function in a separate threadpool
Synopsis
msys.runInPool(pool, closure, ...);
pool: mixed closure: mixed ...: mixed, optional
Description
This function runs a process in a separate threadpool. This can be done in a number of ways.
Note
The first parameter passed to this function may be a thread pool name (string) or a thread pool identifier (integer).
The following example runs closure in the specified pool, returning the value(s) from the closure to the calling script. The calling script is suspended pending completion of the closure. Calling msys.runInPool in this way is useful for moving processing off the scheduler thread for some blocking work.
status, result = msys.runInPool(pool, closure)
Return values are the same as pcall; the first return value is a boolean indicating whether the function was called successfully or not. If it is true then the call was successful and the remaining return value(s) are those from the closure. If the status value is false, the second return value holds an error message.
When the third parameter is not specified, it is equivalent to the function call: msys.runInPool(pool, closure, false);, the boolean indicating that the function specified by closure will not be run asynchronously.
To run a closure asynchronously from the execution of the calling script explicitly specify the third parameter as true:
msys.runInPool(pool, closure, true)
Due to the mechanics of the system, the closure will not start executing until the calling script has returned to the scheduler (which typically means at the end of each validation phase, or when the top level frame on the Lua call stack unwinds). When called in this way no results are returned . The calling script will continue executing the next statement immediately; it will not wait for the closure to begin or for execution to complete.
You can also run a closure asynchronously and execute a function after that closure is complete. To do this call msys.runInPool in the following way:
msys.runInPool(pool, closure, completion_closure)
When the closure returns, completion_closure will be called on the same OS level thread as closure.
The completion function can be run in the same threadpool, or in the scheduler thread. You might want to have the completion function in the scheduler thread to schedule an event, for example.
If the third argument of msys.runInPool is a function, then it will be used as the completion function. An optional fourth argument allows the completion mode to be specified—if it's set to msys.core.ECTP_COMPLETE_CALLBACK_SCHED then the completion function will be called in the scheduler thread; otherwise it will be called in the threadpool thread.
Note
msys.core.ECTP_COMPLETE_CALLBACK_SCHED is the only completion mode that can be passed as the fourth argument. All other values will be ignored, and msys.core.ECTP_COMPLETE_CALLBACK will be used.
status, res = msys.runInPool("IO",
function()
-- do some IO intensive work here, such as interrogate the message contents
return "done"; -- passed back to the 'res' variable
end,
function()
-- Completion function
print("Done!");
end,
msys.core.ECTP_COMPLETE_CALLBACK_SCHED -- Complete in the scheduler
);
In any of the preceding examples, the closure parameter may be a function reference or a function defined inline. For example:
do_heavy_lifting = function()
print("doing heavy lifting")
return true
end
st, res = msys.runInPool('IO', do_heavy_lifting)
An inline function passed as a parameter would be as follows:
st, res = msys.runInPool('IO',
function()
print("doing heavy lifting")
return true
end
)
New data sharing features that are available in threaded Lua are as follows:
-
_TLS– session-local storage -
_OSTLS– OS-level thread-local storage -
using
self– When running validation functions in Lua, if the callout has a validate_context parameter, the "self" parameter can be used as a table to store session-local variables. This is similar to the vctx dictionaries, but these values can be any Lua value.
...
-- we want to validate that we ran on a different OS thread, so
-- set a marker we can use in our output
_OSTLS.async_t_marker = 42;
function mod:validate_data(msg, ac, vctx)
msys.runInPool('CPU', function ()
print("CPU", "in cpu", _OSTLS.async_t_marker);
end, true);
test('validate_data');
return msys.core.VALIDATE_CONT;
end
function mod:validate_data_spool(msg, ac, vctx)
msys.runInPool('CPU',
function ()
print("CB", "cb data spool", _OSTLS.async_t_marker);
end,
function ()
print("CB", "cb completing", _OSTLS.async_t_marker);
end
)
...
end
Because this function is in the msys namespace, an explicit require is not necessary.