Logging¶
How to Use the Logging System¶
Call logging functions like you would call printf()
:
#include <icetray/I3Logging.h>
void doSomething() {
int j=6;
log_debug("here is a debug message, j is %d", j);
}
};
Will output:
DEBUG (MyClass): here is a debug message, j is 6 (MyClass.h:14 in void MyClass::doSomething())
Each of the printf()
-style logging functions has an iostreams
-style analogue that can be
useful for easily formatting non-POD types. For example, both of following logging calls:
#include <icetray/I3Logging.h>
void doSomething() {
OMKey key(64, 26);
std::string name("InIceRawData")
log_debug("Looking for OMKey(%d,%d) in %s", key.GetString(), key.GetOM(), name.c_str());
log_debug_stream("Looking for " << key << " in " << name);
}
would yield identical output:
DEBUG (MyClass): Looking for OMKey(64,26) in InIceRawData (MyClass.h:14 in void MyClass::doSomething())
There are several logging levels available. Each has a different meaning:
Function |
Semantics |
---|---|
|
Only called for fatal errors, will throw an exception and data processing will stop. |
|
Non-fatal (recoverable) error. No exception thrown. |
|
Possible error conditions approaching…. |
|
Information to tell operator what’s going on. |
|
Information for system expert. |
|
Usable in inner loops, ok if tons of output is generated at this level. |
The “standard” logging level is WARN. If your module is working happily, no output at log_warn or higher should be visible.
Here are some guidelines and example statements at each level to help you choose the right logging level:
Function |
Example |
---|---|
|
(will cause exit) “Unable to open input file”, “No DB connection available” |
|
(won’t exit, but your module has a serious problem that needs attention) “Found malformed event data”, “divide by zero detected in module”. |
|
(non-serious errors, “non-standard” configuration detected) “OM key out of range”, “<1TB disk space remains”, “no AMANDA data in >5 minutes”…. |
|
(<< 1/event) Major transitions, new files open, “processed 1000 events” type statements, |
|
One or two lines output for each process call (per event, config, etc) per module |
|
line by line debugging is possible. |
Warning
If you compile in Release
mode (see CMAKE_BUILD_TYPE), log_trace and log_debug are “compiled out”
(removed by the C preprocessor) and will not be available, no
matter how you configure your logging output.
How to Add a Custom Logging Channel to your Module¶
You can add a logging channel just for your module, using the SET_LOGGER() macro:
I3SimpleModule.h:
class I3SimpleModule : public I3Module
{
public:
I3SimpleModule(const I3Context& context);
void Configure();
void Physics(I3FramePtr frame);
void Finish();
private:
std::string outputFilename_;
SET_LOGGER("I3SimpleModule");
};
That opens up a configurable logging channel called I3SimpleModule. This lets you configure logging for I3SimpleModule separately than the global logging preferences. For example, you can show log messages for I3SimpleModule at a different level (e.g. debug) than the global preferences. In addition, “I3SimpleModule” will be printed at the beginning of every log output and, for some logging backends, output from I3SimpleModule can be diverted to a different destination (e.g. to its own log file).
Configuring the logger’s output¶
Icetray logging messages can be routed to different places (e.g. the console or to a file) by choosing a specific logger module or changing the logging threshold either for a specific logging unit (e.g. MySimpleModule above) or the default for all logging units. Like most aspects of Icetray, this behavior can be controlled programmatically from Python as part of an Icetray script.
Setting the default logger¶
There are three logging backends provided with Icetray: I3NullLogger, I3PrintfLogger, and
the Python logging facility. The logger currently in use by Icetray is accessible from
Python at icetray.I3Logger.global_logger
and can be changed by assignment to an instance
of a different logging class. The equivalent C++ routine is the global function
SetIcetrayLogger
. The logger can be changed at any time – although it is usually
set at the beginning of an Icetray script after the icetray library has been loaded.
I3NullLogger¶
I3NullLogger is a logger that does not log anything. This can be useful if log messages
are never desired (although log_fatal()
calls will still cause program termination).
Logging can be disabled by:
icetray.I3Logger.global_logger = icetray.I3NullLogger()
I3PrintfLogger¶
I3PrintfLogger sends log messages to the standard error stream (usually the console), including decorative colors unless standard error has been redirected to a file. This is the default logging module. Error messages look like this:
ERROR (I3Tray): This I3Tray has already been executed. Ignoring second call to Execute() (I3Tray.cxx:442 in void I3Tray::Execute())
The first section in bold shows the level of the log message (ERROR) and the logging unit
(I3Tray) producing the error. The next part in normal text is the message passed to, in this
case, log_error()
, and the final sections in bold give the file, line number, and
function from which the logging system was called.
The default logger can be set to I3PrintfLogger by:
icetray.I3Logger.global_logger = icetray.I3PrintfLogger()
Python Logging¶
Icetray also includes a bridge to the python logging facility, which contains a number of handlers able to do things like write log messages to a rotating set of files, to the system logging daemon, to email, or to a web server. You can also define your own.
Several convenience functions are provided for standard uses of the logging facility either to write to standard error (like I3PrintfLogger), to rotating log files, or to the syslog daemon.
To use python logging to the console, do:
icetray.logging.console()
To use python logging to log files, do:
icetray.logging.rotating_files('/path/to/filename')
To log to syslogd (e.g. the Console
program on Mac OS X):
icetray.logging.syslog()
More complicated things can be done if desired using the Python logging infrastructure – see the Python documentation for details.
In addition to configuring logging from Python, all the C++ logging functions (log_info() etc.) are available from Python under icetray.logging. These functions take two arguments: a log message and optional logging unit (the default is the “Python” unit). For example:
icetray.logging.log_fatal("Things are irreparably broken!", "I3FooModule")
Setting Logging Thresholds¶
The logging threshold is set by default to LOG_WARN for all modules (i.e. only
log_warn()
and more severe conditions will be logged). If you want to change
this global threshold, this can be changed by I3Logger’s set_level()
method.
For example, to get more information, the global log threshold can be reduced to LOG_INFO:
icetray.set_log_level(icetray.I3LogLevel.LOG_INFO)
This can also be changed on a per-logging stream level. For example, to get extremely verbose information just from I3Tray, while leaving all other subsystems at their normal levels:
icetray.set_log_level_for_unit('I3Tray', icetray.I3LogLevel.LOG_TRACE)
Caveats¶
Be careful about logging statements in destructors. The rotating file logger is created in Python using the Python logger, and its destructor is called before the C++ module destructors. At that point, the logger defaults to the I3PrintfLogger in C++. For the time being we will make no effort to ensure the Python logger is destroyed after I3Modules.