Erlang (programming language)/Tutorials/Trap exit: Difference between revisions
imported>Eric Evers (New page: =Trapping Exit Signals= Trapping Exit Signals is one of the most elegant parts of erlang. Under certain conditions we can cause the error message of a crashing process to be turned into a...) |
imported>Tom Morris m (Erlang programming language/Tutorials/Trap exit moved to Erlang (programming language)/Tutorials/Trap exit) |
||
| (2 intermediate revisions by 2 users not shown) | |||
| Line 1: | Line 1: | ||
{{subpages}} | |||
=Trapping Exit Signals= | =Trapping Exit Signals= | ||
| Line 12: | Line 13: | ||
... | ... | ||
When trap_exit==true then monitor | When trap_exit==true then monitor behaves like a system process and receives exit messages from its workers. Normally, a monitor process will exit with its linked worker. After trap_exit=true, if the worker exits then the monitor receives an {EXIT, Pid, Why} message. If the monitor wishes, it could restart the worker or do some other needed action based on the type of exit condition. | ||
In the example code we take the log of 10, 1 and 0. Of course log(10) and log(1) have values but log(0) is undefined and causes the worker process to exit with a | In the example code we take the log of 10, 1 and 0. Of course log(10) and log(1) have values but log(0) is undefined and causes the worker process to exit with a | ||
Latest revision as of 06:07, 8 August 2009
| The metadata subpage is missing. You can start it via filling in this form or by following the instructions that come up after clicking on the [show] link to the right. | |||
|---|---|---|---|
|
Trapping Exit Signals
Trapping Exit Signals is one of the most elegant parts of erlang. Under certain conditions we can cause the error message of a crashing process to be turned into a regular message to another process that is watching. In the example program we construct a monitor-worker relationship. A monitor process is watching a worker process. First we spawn the monitor, then we spawn the worker. We need to link the two processes. Then we change the process flag for trap_exit to true.
start() -> spawn( monitor... spawn( worker...
monitor() -> ... link(worker). process_flag(trap_exit, true). ...
When trap_exit==true then monitor behaves like a system process and receives exit messages from its workers. Normally, a monitor process will exit with its linked worker. After trap_exit=true, if the worker exits then the monitor receives an {EXIT, Pid, Why} message. If the monitor wishes, it could restart the worker or do some other needed action based on the type of exit condition.
In the example code we take the log of 10, 1 and 0. Of course log(10) and log(1) have values but log(0) is undefined and causes the worker process to exit with a badarith(bad arithmatic) error.
Sample Program
%% ---
%% trap_exit_test
%% a simple demo of using trap_exit
%%---
-module(trap_exit_test).
-compile(export_all).
start() ->
Pid = spawn(trap_exit_test, worker, []),
register(work, Pid),
spawn(trap_exit_test, monitor, [Pid]),
work ! {log,10}, sleep(1000),
work ! {log,1}, sleep(1000),
work ! {log,0}.
monitor(Pid) ->
link(Pid),
process_flag(trap_exit, true),
receive
{'EXIT', _Pid, Why} ->
io:format("worker exit: ~w
~n",[Why]),
exit(monitor_normal_exit)
end,
monitor(Pid).
worker() ->
receive
{log, Arg} ->
Output = math:log(Arg),
io:format("log of ~w = ",[Arg]),
io:format("~w ~n",[Output])
after 5000 ->
exit(time_out)
end,
worker().
sleep(T) ->
receive
after T -> ok
end.
Sample Output
1> c(trap_exit_test).
{ok,trap_exit_test}
2> trap_exit_test:start().
log of 10 = 2.30259
log of 1 = 0.00000e+0
=ERROR REPORT==== 12-Jul-2008::10:40:53 ===
Error in process <0.37.0> with exit value:
{badarith,[{math,log,[0]},{trap_exit_test,worker,0}]}
worker exit:
{badarith,[{math,log,[0]},{trap_exit_test,worker,0}]}
{log,0}