[ViewVC] Diff of: cvs/libev/ev.html

Comparing libev/ev.html (file contents):
Revision 1.64 by root, Fri Dec 7 16:49:49 2007 UTC vs.
Revision 1.74 by root, Sun Dec 9 19:42:57 2007 UTC

 <head>
 	<title>libev</title>
 	<meta name="description" content="Pod documentation for libev" />
 	<meta name="inputfile" content="&lt;standard input&gt;" />
 	<meta name="outputfile" content="&lt;standard output&gt;" />
-	<meta name="created" content="Fri Dec  7 17:49:47 2007" />
+	<meta name="created" content="Sun Dec  9 20:30:11 2007" />
 	<meta name="generator" content="Pod::Xhtml 1.57" />
 <link rel="stylesheet" href="http://res.tst.eu/pod.css"/></head>
 <body>
 <div class="pod">
 <!-- INDEX START -->
 </pre>
 </div>
 <h1 id="DESCRIPTION">DESCRIPTION</h1>
 <div id="DESCRIPTION_CONTENT">
+<p>The newest version of this document is also available as a html-formatted
+web page you might find easier to navigate when reading it for the first
+time: <a href="http://cvs.schmorp.de/libev/ev.html">http://cvs.schmorp.de/libev/ev.html</a>.</p>
 <p>Libev is an event loop: you register interest in certain events (such as a
 file descriptor being readable or a timeout occuring), and it will manage
 these event sources and provide your program with events.</p>
 <p>To do this, it must take more or less complete control over your process
 (or thread) by executing the <i>event loop</i> handler, and will then
 one iteration of the loop. This is useful if you are waiting for some
 external event in conjunction with something not expressible using other
 libev watchers. However, a pair of <code>ev_prepare</code>/<code>ev_check</code> watchers is
 usually a better approach for this kind of thing.</p>
 		<p>Here are the gory details of what <code>ev_loop</code> does:</p>
+<pre>   - Before the first iteration, call any pending watchers.
-<pre>   * If there are no active watchers (reference count is zero), return.
+   * If there are no active watchers (reference count is zero), return.
-   - Queue prepare watchers and then call all outstanding watchers.
+   - Queue all prepare watchers and then call all outstanding watchers.
    - If we have been forked, recreate the kernel state.
    - Update the kernel state with all outstanding changes.
    - Update the &quot;event loop time&quot;.
    - Calculate for how long to block.
    - Block the process, waiting for any events.
 	<dt>bool ev_is_pending (ev_TYPE *watcher)</dt>
 	<dd>
 		<p>Returns a true value iff the watcher is pending, (i.e. it has outstanding
 events but its callback has not yet been invoked). As long as a watcher
 is pending (but not active) you must not call an init function on it (but
-<code>ev_TYPE_set</code> is safe) and you must make sure the watcher is available to
+<code>ev_TYPE_set</code> is safe), you must not change its priority, and you must
-libev (e.g. you cnanot <code>free ()</code> it).</p>
+make sure the watcher is available to libev (e.g. you cannot <code>free ()</code>
+it).</p>
 	</dd>
 	<dt>callback ev_cb (ev_TYPE *watcher)</dt>
 	<dd>
 		<p>Returns the callback currently set on the watcher.</p>
 	</dd>
 invocation after new events have been received. This is useful, for
 example, to reduce latency after idling, or more often, to bind two
 watchers on the same event and make sure one is called first.</p>
 		<p>If you need to suppress invocation when higher priority events are pending
 you need to look at <code>ev_idle</code> watchers, which provide this functionality.</p>
+		<p>You <i>must not</i> change the priority of a watcher as long as it is active or
+pending.</p>
 		<p>The default priority used by watchers when no priority has been set is
 always <code>0</code>, which is supposed to not be too high and not be too low :).</p>
 		<p>Setting a priority outside the range of <code>EV_MINPRI</code> to <code>EV_MAXPRI</code> is
 fine, as long as you do not mind that the priority value you query might
 or might not have been adjusted to be within valid range.</p>
+	</dd>
+	<dt>ev_invoke (loop, ev_TYPE *watcher, int revents)</dt>
+	<dd>
+		<p>Invoke the <code>watcher</code> with the given <code>loop</code> and <code>revents</code>. Neither
+<code>loop</code> nor <code>revents</code> need to be valid as long as the watcher callback
+can deal with that fact.</p>
+	</dd>
+	<dt>int ev_clear_pending (loop, ev_TYPE *watcher)</dt>
+	<dd>
+		<p>If the watcher is pending, this function returns clears its pending status
+and returns its <code>revents</code> bitset (as if its callback was invoked). If the
+watcher isn't pending it does nothing and returns <code>0</code>.</p>
 	</dd>
 </dl>
 this situation even with a relatively standard program structure. Thus
 it is best to always use non-blocking I/O: An extra <code>read</code>(2) returning
 <code>EAGAIN</code> is far preferable to a program hanging until some data arrives.</p>
 <p>If you cannot run the fd in non-blocking mode (for example you should not
 play around with an Xlib connection), then you have to seperately re-test
-wether a file descriptor is really ready with a known-to-be good interface
+whether a file descriptor is really ready with a known-to-be good interface
 such as poll (fortunately in our Xlib example, Xlib already does this on
 its own, so its quite safe to use).</p>
 <dl>
 	<dt>ev_io_init (ev_io *, callback, int fd, int events)</dt>
 	<dt>ev_io_set (ev_io *, int fd, int events)</dt>
 but on wallclock time (absolute time). You can tell a periodic watcher
 to trigger &quot;at&quot; some specific point in time. For example, if you tell a
 periodic watcher to trigger in 10 seconds (by specifiying e.g. <code>ev_now ()
 + 10.</code>) and then reset your system clock to the last year, then it will
 take a year to trigger the event (unlike an <code>ev_timer</code>, which would trigger
-roughly 10 seconds later and of course not if you reset your system time
+roughly 10 seconds later).</p>
-again).</p>
 <p>They can also be used to implement vastly more complex timers, such as
-triggering an event on eahc midnight, local time.</p>
+triggering an event on each midnight, local time or other, complicated,
+rules.</p>
 <p>As with timers, the callback is guarenteed to be invoked only when the
 time (<code>at</code>) has been passed, but if multiple periodic timers become ready
 during the same loop iteration then order of execution is undefined.</p>
 <dl>
 	<dt>ev_periodic_init (ev_periodic *, callback, ev_tstamp at, ev_tstamp interval, reschedule_cb)</dt>
 	<dd>
 		<p>Lots of arguments, lets sort it out... There are basically three modes of
 operation, and we will explain them from simplest to complex:</p>
 		<p>
 			<dl>
-				<dt>* absolute timer (interval = reschedule_cb = 0)</dt>
+				<dt>* absolute timer (at = time, interval = reschedule_cb = 0)</dt>
 				<dd>
 					<p>In this configuration the watcher triggers an event at the wallclock time
 <code>at</code> and doesn't repeat. It will not adjust when a time jump occurs,
 that is, if it is to be run at January 1st 2011 then it will run when the
 system time reaches or surpasses this time.</p>
 				</dd>
-				<dt>* non-repeating interval timer (interval &gt; 0, reschedule_cb = 0)</dt>
+				<dt>* non-repeating interval timer (at = offset, interval &gt; 0, reschedule_cb = 0)</dt>
 				<dd>
 					<p>In this mode the watcher will always be scheduled to time out at the next
-<code>at + N * interval</code> time (for some integer N) and then repeat, regardless
+<code>at + N * interval</code> time (for some integer N, which can also be negative)
-of any time jumps.</p>
+and then repeat, regardless of any time jumps.</p>
 					<p>This can be used to create timers that do not drift with respect to system
 time:</p>
 <pre>   ev_periodic_set (&amp;periodic, 0., 3600., 0);
 </pre>
 full hour (UTC), or more correctly, when the system time is evenly divisible
 by 3600.</p>
 					<p>Another way to think about it (for the mathematically inclined) is that
 <code>ev_periodic</code> will try to run the callback in this mode at the next possible
 time where <code>time = at (mod interval)</code>, regardless of any time jumps.</p>
+					<p>For numerical stability it is preferable that the <code>at</code> value is near
+<code>ev_now ()</code> (the current time), but there is no range requirement for
+this value.</p>
 				</dd>
-				<dt>* manual reschedule mode (reschedule_cb = callback)</dt>
+				<dt>* manual reschedule mode (at and interval ignored, reschedule_cb = callback)</dt>
 				<dd>
 					<p>In this mode the values for <code>interval</code> and <code>at</code> are both being
 ignored. Instead, each time the periodic watcher gets scheduled, the
 reschedule callback will be called with the watcher as first, and the
 current time as second argument.</p>
 					<p>NOTE: <i>This callback MUST NOT stop or destroy any periodic watcher,
 ever, or make any event loop modifications</i>. If you need to stop it,
 return <code>now + 1e30</code> (or so, fudge fudge) and stop it afterwards (e.g. by
-starting a prepare watcher).</p>
+starting an <code>ev_prepare</code> watcher, which is legal).</p>
 					<p>Its prototype is <code>ev_tstamp (*reschedule_cb)(struct ev_periodic *w,
 ev_tstamp now)</code>, e.g.:</p>
 <pre>   static ev_tstamp my_rescheduler (struct ev_periodic *w, ev_tstamp now)
    {
      return now + 60.;
 	<dd>
 		<p>Simply stops and restarts the periodic watcher again. This is only useful
 when you changed some parameters or the reschedule callback would return
 a different time than the last time it was called (e.g. in a crond like
 program when the crontabs have changed).</p>
+	</dd>
+	<dt>ev_tstamp offset [read-write]</dt>
+	<dd>
+		<p>When repeating, this contains the offset value, otherwise this is the
+absolute point in time (the <code>at</code> value passed to <code>ev_periodic_set</code>).</p>
+		<p>Can be modified any time, but changes only take effect when the periodic
+timer fires or <code>ev_periodic_again</code> is being called.</p>
 	</dd>
 	<dt>ev_tstamp interval [read-write]</dt>
 	<dd>
 		<p>The current interval value. Can be modified any time, but changes only
 take effect when the periodic timer fires or <code>ev_periodic_again</code> is being
 are ready to run (it's actually more complicated: it only runs coroutines
 with priority higher than or equal to the event loop and one coroutine
 of lower priority, but only once, using idle watchers to keep the event
 loop from blocking if lower-priority coroutines are active, thus mapping
 low-priority coroutines to idle/background tasks).</p>
+<p>It is recommended to give <code>ev_check</code> watchers highest (<code>EV_MAXPRI</code>)
+priority, to ensure that they are being run before any other watchers
+after the poll. Also, <code>ev_check</code> watchers (and <code>ev_prepare</code> watchers,
+too) should not activate (&quot;feed&quot;) events into libev. While libev fully
+supports this, they will be called before other <code>ev_check</code> watchers did
+their job. As <code>ev_check</code> watchers are often used to embed other event
+loops those other event loops might be in an unusable state until their
+<code>ev_check</code> watcher ran (always remind yourself to coexist peacefully with
+others).</p>
 <dl>
 	<dt>ev_prepare_init (ev_prepare *, callback)</dt>
 	<dt>ev_check_init (ev_check *, callback)</dt>
 	<dd>
 		<p>Initialises and configures the prepare or check watcher - they have no
 parameters of any kind. There are <code>ev_prepare_set</code> and <code>ev_check_set</code>
 macros, but using them is utterly, utterly and completely pointless.</p>
 	</dd>
 </dl>
-<p>Example: To include a library such as adns, you would add IO watchers
+<p>There are a number of principal ways to embed other event loops or modules
-and a timeout watcher in a prepare handler, as required by libadns, and
+into libev. Here are some ideas on how to include libadns into libev
+(there is a Perl module named <code>EV::ADNS</code> that does this, which you could
+use for an actually working example. Another Perl module named <code>EV::Glib</code>
+embeds a Glib main context into libev, and finally, <code>Glib::EV</code> embeds EV
+into the Glib event loop).</p>
+<p>Method 1: Add IO watchers and a timeout watcher in a prepare handler,
-in a check watcher, destroy them and call into libadns. What follows is
+and in a check watcher, destroy them and call into libadns. What follows
-pseudo-code only of course:</p>
+is pseudo-code only of course. This requires you to either use a low
+priority for the check watcher or use <code>ev_clear_pending</code> explicitly, as
+the callbacks for the IO/timeout watchers might not have been called yet.</p>
 <pre>  static ev_io iow [nfd];
   static ev_timer tw;
   static void
   io_cb (ev_loop *loop, ev_io *w, int revents)
   {
-    // set the relevant poll flags
-    // could also call adns_processreadable etc. here
-    struct pollfd *fd = (struct pollfd *)w-&gt;data;
-    if (revents &amp; EV_READ ) fd-&gt;revents |= fd-&gt;events &amp; POLLIN;
-    if (revents &amp; EV_WRITE) fd-&gt;revents |= fd-&gt;events &amp; POLLOUT;
   }
   // create io watchers for each fd and a timer before blocking
   static void
   adns_prepare_cb (ev_loop *loop, ev_prepare *w, int revents)
     /* the callback is illegal, but won't be called as we stop during check */
     ev_timer_init (&amp;tw, 0, timeout * 1e-3);
     ev_timer_start (loop, &amp;tw);
-    // create on ev_io per pollfd
+    // create one ev_io per pollfd
     for (int i = 0; i &lt; nfd; ++i)
       {
         ev_io_init (iow + i, io_cb, fds [i].fd,
           ((fds [i].events &amp; POLLIN ? EV_READ : 0)
            | (fds [i].events &amp; POLLOUT ? EV_WRITE : 0)));
         fds [i].revents = 0;
-        iow [i].data = fds + i;
         ev_io_start (loop, iow + i);
       }
   }
   // stop all watchers after blocking
   adns_check_cb (ev_loop *loop, ev_check *w, int revents)
   {
     ev_timer_stop (loop, &amp;tw);
     for (int i = 0; i &lt; nfd; ++i)
+      {
+        // set the relevant poll flags
+        // could also call adns_processreadable etc. here
+        struct pollfd *fd = fds + i;
+        int revents = ev_clear_pending (iow + i);
+        if (revents &amp; EV_READ ) fd-&gt;revents |= fd-&gt;events &amp; POLLIN;
+        if (revents &amp; EV_WRITE) fd-&gt;revents |= fd-&gt;events &amp; POLLOUT;
+        // now stop the watcher
-      ev_io_stop (loop, iow + i);
+        ev_io_stop (loop, iow + i);
+      }
     adns_afterpoll (adns, fds, nfd, timeval_from (ev_now (loop));
+  }
+</pre>
+<p>Method 2: This would be just like method 1, but you run <code>adns_afterpoll</code>
+in the prepare watcher and would dispose of the check watcher.</p>
+<p>Method 3: If the module to be embedded supports explicit event
+notification (adns does), you can also make use of the actual watcher
+callbacks, and only destroy/create the watchers in the prepare watcher.</p>
+<pre>  static void
+  timer_cb (EV_P_ ev_timer *w, int revents)
+  {
+    adns_state ads = (adns_state)w-&gt;data;
+    update_now (EV_A);
+    adns_processtimeouts (ads, &amp;tv_now);
+  }
+  static void
+  io_cb (EV_P_ ev_io *w, int revents)
+  {
+    adns_state ads = (adns_state)w-&gt;data;
+    update_now (EV_A);
+    if (revents &amp; EV_READ ) adns_processreadable  (ads, w-&gt;fd, &amp;tv_now);
+    if (revents &amp; EV_WRITE) adns_processwriteable (ads, w-&gt;fd, &amp;tv_now);
+  }
+  // do not ever call adns_afterpoll
+</pre>
+<p>Method 4: Do not use a prepare or check watcher because the module you
+want to embed is too inflexible to support it. Instead, youc na override
+their poll function.  The drawback with this solution is that the main
+loop is now no longer controllable by EV. The <code>Glib::EV</code> module does
+this.</p>
+<pre>  static gint
+  event_poll_func (GPollFD *fds, guint nfds, gint timeout)
+  {
+    int got_events = 0;
+    for (n = 0; n &lt; nfds; ++n)
+      // create/start io watcher that sets the relevant bits in fds[n] and increment got_events
+    if (timeout &gt;= 0)
+      // create/start timer
+    // poll
+    ev_loop (EV_A_ 0);
+    // stop timer again
+    if (timeout &gt;= 0)
+      ev_timer_stop (EV_A_ &amp;to);
+    // stop io watchers again - their callbacks should have set
+    for (n = 0; n &lt; nfds; ++n)
+      ev_io_stop (EV_A_ iow [n]);
+    return got_events;
   }
 the callback model to a model using method callbacks on objects.</p>
 <p>To use it,</p>
 <pre>  #include &lt;ev++.h&gt;
 </pre>
-<p>(it is not installed by default). This automatically includes <cite>ev.h</cite>
+<p>This automatically includes <cite>ev.h</cite> and puts all of its definitions (many
-and puts all of its definitions (many of them macros) into the global
+of them macros) into the global namespace. All C++ specific things are
-namespace. All C++ specific things are put into the <code>ev</code> namespace.</p>
+put into the <code>ev</code> namespace. It should support all the same embedding
-<p>It should support all the same embedding options as <cite>ev.h</cite>, most notably
+options as <cite>ev.h</cite>, most notably <code>EV_MULTIPLICITY</code>.</p>
-<code>EV_MULTIPLICITY</code>.</p>
+<p>Care has been taken to keep the overhead low. The only data member the C++
+classes add (compared to plain C-style watchers) is the event loop pointer
+that the watcher is associated with (or no additional members at all if
+you disable <code>EV_MULTIPLICITY</code> when embedding libev).</p>
+<p>Currently, functions, and static and non-static member functions can be
+used as callbacks. Other types should be easy to add as long as they only
+need one additional pointer for context. If you need support for other
+types of functors please contact the author (preferably after implementing
+it).</p>
 <p>Here is a list of things available in the <code>ev</code> namespace:</p>
 <dl>
 	<dt><code>ev::READ</code>, <code>ev::WRITE</code> etc.</dt>
 	<dd>
 		<p>These are just enum values with the same values as the <code>EV_READ</code> etc.
 which is called <code>ev::sig</code> to avoid clashes with the <code>signal</code> macro
 defines by many implementations.</p>
 		<p>All of those classes have these methods:</p>
 		<p>
 			<dl>
-				<dt>ev::TYPE::TYPE (object *, object::method *)</dt>
+				<dt>ev::TYPE::TYPE ()</dt>
-				<dt>ev::TYPE::TYPE (object *, object::method *, struct ev_loop *)</dt>
+				<dt>ev::TYPE::TYPE (struct ev_loop *)</dt>
 				<dt>ev::TYPE::~TYPE</dt>
 				<dd>
-					<p>The constructor takes a pointer to an object and a method pointer to
+					<p>The constructor (optionally) takes an event loop to associate the watcher
-the event handler callback to call in this class. The constructor calls
+with. If it is omitted, it will use <code>EV_DEFAULT</code>.</p>
-<code>ev_init</code> for you, which means you have to call the <code>set</code> method
+					<p>The constructor calls <code>ev_init</code> for you, which means you have to call the
-before starting it. If you do not specify a loop then the constructor
+<code>set</code> method before starting it.</p>
-automatically associates the default loop with this watcher.</p>
+					<p>It will not set a callback, however: You have to call the templated <code>set</code>
+method to set a callback before you can start the watcher.</p>
+					<p>(The reason why you have to use a method is a limitation in C++ which does
+not allow explicit template arguments for constructors).</p>
 					<p>The destructor automatically stops the watcher if it is active.</p>
+				</dd>
+				<dt>w-&gt;set&lt;class, &amp;class::method&gt; (object *)</dt>
+				<dd>
+					<p>This method sets the callback method to call. The method has to have a
+signature of <code>void (*)(ev_TYPE &amp;, int)</code>, it receives the watcher as
+first argument and the <code>revents</code> as second. The object must be given as
+parameter and is stored in the <code>data</code> member of the watcher.</p>
+					<p>This method synthesizes efficient thunking code to call your method from
+the C callback that libev requires. If your compiler can inline your
+callback (i.e. it is visible to it at the place of the <code>set</code> call and
+your compiler is good :), then the method will be fully inlined into the
+thunking function, making it as fast as a direct C callback.</p>
+					<p>Example: simple class declaration and watcher initialisation</p>
+<pre>  struct myclass
+  {
+    void io_cb (ev::io &amp;w, int revents) { }
+  }
+  myclass obj;
+  ev::io iow;
+  iow.set &lt;myclass, &amp;myclass::io_cb&gt; (&amp;obj);
+</pre>
+				</dd>
+				<dt>w-&gt;set&lt;function&gt; (void *data = 0)</dt>
+				<dd>
+					<p>Also sets a callback, but uses a static method or plain function as
+callback. The optional <code>data</code> argument will be stored in the watcher's
+<code>data</code> member and is free for you to use.</p>
+					<p>The prototype of the <code>function</code> must be <code>void (*)(ev::TYPE &amp;w, int)</code>.</p>
+					<p>See the method-<code>set</code> above for more details.</p>
+					<p>Example:</p>
+<pre>  static void io_cb (ev::io &amp;w, int revents) { }
+  iow.set &lt;io_cb&gt; ();
+</pre>
 				</dd>
 				<dt>w-&gt;set (struct ev_loop *)</dt>
 				<dd>
 					<p>Associates a different <code>struct ev_loop</code> with this watcher. You can only
 do this when the watcher is inactive (and not pending either).</p>
 				</dd>
 				<dt>w-&gt;set ([args])</dt>
 				<dd>
 					<p>Basically the same as <code>ev_TYPE_set</code>, with the same args. Must be
-called at least once.  Unlike the C counterpart, an active watcher gets
+called at least once. Unlike the C counterpart, an active watcher gets
-automatically stopped and restarted.</p>
+automatically stopped and restarted when reconfiguring it with this
+method.</p>
 				</dd>
 				<dt>w-&gt;start ()</dt>
 				<dd>
-					<p>Starts the watcher. Note that there is no <code>loop</code> argument as the
+					<p>Starts the watcher. Note that there is no <code>loop</code> argument, as the
-constructor already takes the loop.</p>
+constructor already stores the event loop.</p>
 				</dd>
 				<dt>w-&gt;stop ()</dt>
 				<dd>
 					<p>Stops the watcher if it is active. Again, no <code>loop</code> argument.</p>
 				</dd>
     myclass ();
   }
   myclass::myclass (int fd)
-  : io   (this, &amp;myclass::io_cb),
-    idle (this, &amp;myclass::idle_cb)
   {
+    io  .set &lt;myclass, &amp;myclass::io_cb  &gt; (this);
+    idle.set &lt;myclass, &amp;myclass::idle_cb&gt; (this);
     io.start (fd, ev::READ);
   }
 </div>
 <h1 id="MACRO_MAGIC">MACRO MAGIC</h1>
 <div id="MACRO_MAGIC_CONTENT">
 <p>Libev can be compiled with a variety of options, the most fundemantal is
-<code>EV_MULTIPLICITY</code>. This option determines wether (most) functions and
+<code>EV_MULTIPLICITY</code>. This option determines whether (most) functions and
 callbacks have an initial <code>struct ev_loop *</code> argument.</p>
 <p>To make it easier to write programs that cope with either variant, the
 following macros are defined:</p>
 <dl>
 	<dt><code>EV_A</code>, <code>EV_A_</code></dt>
 		<p>Similar to the other two macros, this gives you the value of the default
 loop, if multiple loops are supported (&quot;ev loop default&quot;).</p>
 	</dd>
 </dl>
 <p>Example: Declare and initialise a check watcher, utilising the above
-macros so it will work regardless of wether multiple loops are supported
+macros so it will work regardless of whether multiple loops are supported
 or not.</p>
 <pre>  static void
   check_cb (EV_P_ ev_timer *w, int revents)
   {
     ev_check_stop (EV_A_ w);
 		<p>If undefined or defined to <code>1</code>, then all event-loop-specific functions
 will have the <code>struct ev_loop *</code> as first argument, and you can create
 additional independent event loops. Otherwise there will be no support
 for multiple event loops and there is no first event loop pointer
 argument. Instead, all functions act on the single default loop.</p>
+	</dd>
+	<dt>EV_MINPRI</dt>
+	<dt>EV_MAXPRI</dt>
+	<dd>
+		<p>The range of allowed priorities. <code>EV_MINPRI</code> must be smaller or equal to
+<code>EV_MAXPRI</code>, but otherwise there are no non-obvious limitations. You can
+provide for more priorities by overriding those symbols (usually defined
+to be <code>-2</code> and <code>2</code>, respectively).</p>
+		<p>When doing priority-based operations, libev usually has to linearly search
+all the priorities, so having many of them (hundreds) uses a lot of space
+and time, so using the defaults of five priorities (-2 .. +2) is usually
+fine.</p>
+		<p>If your embedding app does not need any priorities, defining these both to
+<code>0</code> will save some memory and cpu.</p>
 	</dd>
 	<dt>EV_PERIODIC_ENABLE</dt>
 	<dd>
 		<p>If undefined or defined to be <code>1</code>, then periodic timers are supported. If
 defined to be <code>0</code>, then they are not. Disabling them saves a few kB of
 <h1 id="COMPLEXITIES">COMPLEXITIES</h1>
 <div id="COMPLEXITIES_CONTENT">
 		<p>In this section the complexities of (many of) the algorithms used inside
 libev will be explained. For complexity discussions about backends see the
 documentation for <code>ev_default_init</code>.</p>
+		<p>All of the following are about amortised time: If an array needs to be
+extended, libev needs to realloc and move the whole array, but this
+happens asymptotically never with higher number of elements, so O(1) might
+mean it might do a lengthy realloc operation in rare cases, but on average
+it is much faster and asymptotically approaches constant time.</p>
 		<p>
 			<dl>
 				<dt>Starting and stopping timer/periodic watchers: O(log skipped_other_timers)</dt>
+				<dd>
+					<p>This means that, when you have a watcher that triggers in one hour and
+there are 100 watchers that would trigger before that then inserting will
+have to skip those 100 watchers.</p>
+				</dd>
 				<dt>Changing timer/periodic watchers (by autorepeat, again): O(log skipped_other_timers)</dt>
+				<dd>
+					<p>That means that for changing a timer costs less than removing/adding them
+as only the relative motion in the event queue has to be paid for.</p>
+				</dd>
 				<dt>Starting io/check/prepare/idle/signal/child watchers: O(1)</dt>
+				<dd>
+					<p>These just add the watcher into an array or at the head of a list.
-				<dt>Stopping check/prepare/idle watchers: O(1)</dt>
+=item Stopping check/prepare/idle watchers: O(1)</p>
+				</dd>
 				<dt>Stopping an io/signal/child watcher: O(number_of_watchers_for_this_(fd/signal/pid % EV_PID_HASHSIZE))</dt>
+				<dd>
+					<p>These watchers are stored in lists then need to be walked to find the
+correct watcher to remove. The lists are usually short (you don't usually
+have many watchers waiting for the same fd or signal).</p>
+				</dd>
 				<dt>Finding the next timer per loop iteration: O(1)</dt>
 				<dt>Each change on a file descriptor per loop iteration: O(number_of_watchers_for_this_fd)</dt>
+				<dd>
+					<p>A change means an I/O watcher gets started or stopped, which requires
+libev to recalculate its status (and possibly tell the kernel).</p>
+				</dd>
 				<dt>Activating one watcher: O(1)</dt>
+				<dt>Priority handling: O(number_of_priorities)</dt>
+				<dd>
+					<p>Priorities are implemented by allocating some space for each
+priority. When doing priority-based operations, libev usually has to
+linearly search all the priorities.</p>
+				</dd>
 			</dl>
 		</p>

Diff Legend

-–
+Removed lines
-+
+Added lines
-<
+Changed lines
->
+Changed lines

Comparing libev/ev.html (file contents): Revision 1.64 by root, Fri Dec 7 16:49:49 2007 UTC vs. Revision 1.74 by root, Sun Dec 9 19:42:57 2007 UTC

Diff Legend

Comparing libev/ev.html (file contents):
Revision 1.64 by root, Fri Dec 7 16:49:49 2007 UTC vs.
Revision 1.74 by root, Sun Dec 9 19:42:57 2007 UTC