4. Belle II Python Interface

The Belle II Software has an extensive interface to the Python 3 scripting language: All configuration and steering is done via python and in principle also simple algorithms can be implemented directly in python. All main functions are implemented in a module called basf2 and most people will just start their steering file or script with

import basf2
main = basf2.Path()

4.1. Modules and Paths

A typical data processing chain consists of a linear arrangement of smaller processing blocks, called Modules. Their tasks vary from simple ones like reading data from a file to complex tasks like the full detector simulation or the tracking.

In basf2 all work is done in modules, which means that even the reading of data from disk and writing it back is done in modules. They live in a Path, which corresponds to a container where the modules are arranged in a strict linear order. The specific selection and arrangement of modules depend on the user’s current task. When processing data, the framework executes the modules of a path, starting with the first one and proceeding with the module next to it. The modules are executed one at a time, exactly in the order in which they were placed into the path.

Modules can have conditions attached to them to steer the processing flow depending of the outcome of the calculation in each module.

The data, to be processed by the modules, is stored in a common storage, the DataStore. Each module has read and write access to the storage. In addition there’s also non-event data, the so called conditions, which will be loaded from a central conditions database and are available in the DBStore.


Fig. 4.1 Schematic view of the processing flow in the Belle II Software

Usually each script needs to create a new path using Path(), add all required modules in the correct order and finally call process() on the fully configured path.


Preparing a Path and adding Modules to it does not execute anything, it only prepares the computation which is only done when process is called.

The following functions are all related to the handling of modules and paths:


Creates a new path and returns it. You can also instantiate basf2.Path directly.

basf2.register_module(name_or_module, shared_lib_path=None, logLevel=None, debugLevel=None, **kwargs)[source]

Register the module ‘name’ and return it (e.g. for adding to a path). This function is intended to instantiate existing modules. To find out which modules exist you can run basf2 -m and to get details about the parameters for each module you can use basf2 -m {modulename}

Parameters can be passed directly to the module as keyword parameters or can be set later using Module.param

>>> module = basf2.register_module('EventInfoSetter', evtNumList=100, logLevel=LogLevel.ERROR)
>>> module.param("evtNumList", 100)
  • name_or_module – The name of the module type, may also be an existing Module instance for which parameters should be set

  • shared_lib_path (str) – An optional path to a shared library from which the module should be loaded

  • logLevel (LogLevel) – indicates the minimum severity of log messages to be shown from this module. See Module.set_log_level

  • debugLevel (int) – Number indicating the detail of debug messages, the default level is 100. See Module.set_debug_level

  • kwargs – Additional parameters to be passed to the module.


You can also use Path.add_module() directly, which accepts the same name, logging and module parameter arguments. There is no need to register the module by hand if you will add it to the path in any case.

basf2.set_module_parameters(path, name=None, type=None, recursive=False, **kwargs)[source]

Set the given set of parameters for all modules in a path which have the given name (see Module.set_name)

Usage is similar to register_module() but this function will not create new modules but just adjust parameters for modules already in a Path

>>> set_module_parameters(path, "Geometry", components=["PXD"], logLevel=LogLevel.WARNING)
  • path (basf2.Path) – The path to search for the modules

  • name (str) – Then name of the module to set parameters for

  • type (str) – The type of the module to set parameters for.

  • recursive (bool) – if True also look in paths connected by conditions or Path.for_each()

  • kwargs – Named parameters to be set for the module, see register_module()

basf2.print_params(module, print_values=True, shared_lib_path=None)[source]

This function prints parameter information

  • module – Print the parameter information of this module

  • print_values – Set it to True to print the current values of the parameters

  • shared_lib_path – The path of the shared library from which the module was loaded

basf2.print_path(path, defaults=False, description=False, indentation=0, title=True)[source]

This function prints the modules in the given path and the module parameters. Parameters that are not set by the user are suppressed by default.

  • defaults – Set it to True to print also the parameters with default values

  • description – Set to True to print the descriptions of modules and parameters

  • indentation – an internal parameter to indent the whole output (needed for outputting sub-paths)

  • title – show the title string or not (defaults to True)

basf2.process(path, max_event=0)[source]

Start processing events using the modules in the given basf2.Path object.

Can be called multiple times in one steering file (some restrictions apply: modules need to perform proper cleanup & reinitialisation, if Geometry is involved this might be difficult to achieve.)

When used in a Jupyter notebook this function will automatically print a nice progress bar and display the log messages in an advanced way once the processing is complete.


This also means that in a Jupyter Notebook, modifications to class members or global variables will not be visible after processing is complete as the processing is performed in a subprocess.

To restore the old behavior you can use basf2.core.process() which will behave exactly identical in Jupyter notebooks as it does in normal python scripts

from basf2 import core
  • path – The path with which the processing starts

  • max_event – The maximal number of events which will be processed, 0 for no limit

Changed in version release-03-00-00: automatic Jupyter integration

4.1.1. The Module Object

Unless you develop your own module in Python you should always instantiate new modules by calling register_module or Path.add_module.

class basf2.Module

Base class for Modules.

A module is the smallest building block of the framework. A typical event processing chain consists of a Path containing modules. By inheriting from this base class, various types of modules can be created. To use a module, please refer to Path.add_module(). A list of modules is available by running basf2 -m or basf2 -m package, detailed information on parameters is given by e.g. basf2 -m RootInput.

The ‘Module Development’ section in the manual provides detailed information on how to create modules, setting parameters, or using return values/conditions: https://confluence.desy.de/display/BI/Software+Basf2manual#Module_Development

available_params()list :

Return list of all module parameters as ModuleParamInfo instances

beginRun()None :

This function is called by the processing just before a new run of data is processed. Modules can override this method to perform actions which are run dependent

description()str :

Returns the description of this module.

endRun()None :

This function is called by the processing just after a new run of data is processed. Modules can override this method to perform actions which are run dependent

event()None :

This function is called by the processing once for each event.Modules should override this method to perform actions during event processing

get_all_condition_paths()list :

Return a list of all conditional paths set for this module using if_value, if_true or if_false

get_all_conditions()list :

Return a list of all conditional path expressions set for this module using if_value, if_true or if_false

has_condition()bool :

Return true if a conditional path has been set for this module using if_value, if_true or if_false

has_properties((int)properties)bool :

Allows to check if the module has the given properties out of ModulePropFlags set.

>>> if module.has_properties(ModulePropFlags.PARALLELPROCESSINGCERTIFIED):
>>>     ...

properties (int) – bitmask of ModulePropFlags to check for.

if_false(condition_path, after_condition_path=AfterConditionPath.END)

Sets a conditional sub path which will be executed after this module if the return value of the module evaluates to False. This is equivalent to calling if_value with expression=\"<1\"

if_true(condition_path, after_condition_path=AfterConditionPath.END)

Sets a conditional sub path which will be executed after this module if the return value of the module evaluates to True. It is equivalent to calling if_value with expression=\">=1\"

if_value(expression, condition_path, after_condition_path=AfterConditionPath.END)

Sets a conditional sub path which will be executed after this module if the return value set in the module passes the given expression.

Modules can define a return value (int or bool) using setReturnValue(), which can be used in the steering file to split the Path based on this value, for example

>>> module_with_condition.if_value("<1", another_path)

In case the return value of the module_with_condition for a given event is less than 1, the execution will be diverted into another_path for this event.

You could for example set a special return value if an error occurs, and divert the execution into a path containing RootOutput if it is found; saving only the data producing/produced by the error.

After a conditional path has executed, basf2 will by default stop processing the path for this event. This behaviour can be changed by setting the after_condition_path argument.

  • expression (str) – Expression to determine if the conditional path should be executed. This should be one of the comparison operators <, >, <=, >=, ==, or != followed by a numerical value for the return value

  • condition_path (Path) – path to execute in case the expression is fulfilled

  • after_condition_path (AfterConditionPath) – What to do once the condition_path has been executed.

initialize()None :

This function is called by the processing just once before processing any data is processed. Modules can override this method to perform some actions at startup once all parameters are set

name()str :

Returns the name of the module. Can be changed via set_name(), use type() for identifying a particular module class.

package()str :

Returns the package this module belongs to.

param(key, value=None)

This method can be used to set module parameters. There are two ways of calling this function:

  1. With two arguments where the first is the name of the parameter and the second is the value.

    >>> module.param("parameterName", "parameterValue")
  2. Or with just one parameter which is a dictionary mapping multiple parameter names to their values

    >>> module.param({"parameter1": "value1", "parameter2": True})
return_value((int)value)None :

Set a return value. Can be used by custom modules to set the return value used to determine if conditional paths are executed

set_abort_level((int)abort_level)None :

Set the log level which will cause processing to be aborted. Usually processing is only aborted for FATAL messages but with this function it’s possible to set this to a lower value


abort_level (LogLevel) – log level which will cause processing to be aborted.

set_debug_level((int)debug_level)None :

Set the debug level for this module. Debug messages with a higher level will be suppressed. This function has no visible effect if the log level is not set to DEBUG


debug_level (int) – Maximum debug level for messages to be displayed.

set_log_info((int)arg2, (int)log_info)None :

Set a LogInfo configuration object for this module to determine how log messages should be formatted

set_log_level((int)log_level)None :

Set the log level for this module. Messages below that level will be suppressed


log_level (LogLevel) – Minimum level for messages to be displayed

set_name((str)name)None :

Set custom name, e.g. to distinguish multiple modules of the same type.

>>> path.add_module('EventInfoSetter')
>>> ro = path.add_module('RootOutput', branchNames=['EventMetaData'])
>>> ro.set_name('RootOutput_metadata_only')
>>> print(path)
[EventInfoSetter -> RootOutput_metadata_only]
set_property_flags((int)property_mask)None :

Set module properties in the form of an OR combination of ModulePropFlags.

terminate()None :

This function is called by the processing once after all data is processed. Modules can override this method to perform some cleanup at shutdown. The terminate functions of all modules are called in reverse order of the initialize calls.

type()str :

Returns the type of the module (i.e. class name minus ‘Module’)

4.1.2. The Path Object

class basf2.Path

Implements a path consisting of Module and/or Path objects (arranged in a linear order).

See also



Does this Path contain a module of the given type?

>>> path = basf2.Path()
>>> 'RootInput' in path
>>> path.add_module('RootInput')
>>> 'RootInput' in path
add_independent_path(skim_path, ds_ID='', merge_back_event=None)

Add given path at the end of this path and ensure all modules there do not influence the main DataStore. You can thus use modules in skim_path to clean up e.g. the list of particles, save a skimmed uDST file, and continue working with the unmodified DataStore contents outside of skim_path.

  • ds_ID – can be specified to give a defined ID to the temporary DataStore, otherwise, a random name will be generated.

  • merge_back_event – is a list of object/array names (of event durability) that will be merged back into the main path.

add_module(module, logLevel=None, debugLevel=None, **kwargs)

Add given module (either object or name) at the end of this path. All unknown arguments are passed as module parameters.

>>> path = create_path()
>>> path.add_module('EventInfoSetter', evtNumList=100, logLevel=LogLevel.ERROR)
<pybasf2.Module at 0x1e356e0>
>>> path = create_path()
>>> eventinfosetter = register_module('EventInfoSetter')
>>> path.add_module(eventinfosetter)
<pybasf2.Module at 0x2289de8>

Insert another path at the end of this one. For example,

>>> path.add_module('A')
>>> path.add_path(otherPath)
>>> path.add_module('B')

would create a path [ A -> [ contents of otherPath ] -> B ].)


path (Path) – path to add to this path

do_while(path, condition='<1', max_iterations=10000)

Similar to add_path() this will execute a path at the current position but it will repeat execution of this path as long as the return value of the last module in the path fulfills the given condition.

This is useful for event generation with special cuts like inclusive particle generation.

See also

Module.if_value for an explanation of the condition expression.

  • path (basf2.Path) – sub path to execute repeatedly

  • condition (str) – condition on the return value of the last module in path. The execution will be repeated as long as this condition is fulfilled.

  • max_iterations (int) – Maximum number of iterations per event. If this number is exceeded the execution is aborted.

for_each(loop_object_name, array_name, path)

Similar to add_path(), this will execute the given path at the current position, but in each event it will execute it once for each object in the given StoreArray arrayName. It will create a StoreObject named loop_object_name of same type as array which will point to each element in turn for each execution.

This has the effect of calling the event() methods of modules in path for each entry in arrayName.

The main use case is to use it after using the RestOfEventBuilder on a ParticeList, where you can use this feature to perform actions on only a part of the event for a given list of candidates:

>>> path.for_each('RestOfEvent', 'RestOfEvents', roe_path)

You can read this as

“for each RestOfEvent in the array of “RestOfEvents”, execute roe_path

For example, if ‘RestOfEvents’ contains two elements then roe_path will be executed twice and during the execution a StoreObjectPtr ‘RestOfEvent’ will be available, which will point to the first element in the first execution, and the second element in the second execution.

See also

A working example of this for_each RestOfEvent is to build a veto against photons from \(\pi^0\to\gamma\gamma\). It is described in How to Veto.


This feature is used by both the Flavor Tagger and Full event interpretation algorithms.

Changes to existing arrays / objects will be available to all modules after the for_each(), including those made to the loop object itself (it will simply modify the i’th item in the array looped over.)

StoreArrays / StoreObjects (of event durability) created inside the loop will be removed at the end of each iteration. So if you create a new particle list inside a for_each() path execution the particle list will not exist for the next iteration or after the for_each() is complete.

  • loop_object_name (str) – The name of the object in the datastore during each execution

  • array_name (str) – The name of the StoreArray to loop over where the i-th element will be available as loop_object_name during the i-th execution of path

  • path (basf2.Path) – The path to execute for each element in array_name


Returns an ordered list of all modules in this path.

4.2. Logging

The Logging system of the Belle II Software is rather flexible and allows extensive configurations. In the most simple case a call to set_log_level is all that is needed to set the minimum severity of messages to be printed. However in addition to this global log level one can set the log level for specific packages and even for individual modules separately. The existing log levels are defined as

class basf2.LogLevel

Class for all possible log levels


The lowest possible severity meant for expert only information and disabled by default. In contrast to all other log levels DEBUG messages have an additional numeric indication of their priority called the debug_level to allow for different levels of verbosity.

The agreed values for debug_level are

  • 0-9 for user code. These numbers are reserved for user analysis code and may not be used by any part of basf2.

  • 10-19 for analysis package code. The use case is that a user wants to debug problems in analysis jobs with the help of experts.

  • 20-29 for simulation/reconstruction code.

  • 30-39 for core framework code.


The default maximum debug level which will be shown when running basf2 --debug without any argument for --debug is 10


Used for informational messages which are of use for the average user but not very important. Should be used very sparsely, everything which is of no interest to the average user should be a debug message.


Informational message which don’t indicate an error condition but are more important than a mere information. For example the calculated cross section or the output file name.

Deprecated since version release-01-00-00: use INFO messages instead


For messages which indicate something which is not correct but not fatal to the processing. This should not be used to make informational messages more prominent and they should not be ignored by the user but they are not critical.


For messages which indicate a clear error condition which needs to be recovered. If error messages are produced before event processing is started the processing will be aborted. During processing errors don’t lead to a stop of the processing but still indicate a problem.


For errors so severe that no recovery is possible. Emitting a fatal error will always stop the processing and the B2FATAL function is guaranteed to not return.


Sets the global log level which specifies up to which level the logging messages will be shown


level (basf2.LogLevel) – minimum severity of messages to be logged


Sets the global debug level which specifies up to which level the debug messages should be shown


level (int) – The debug level. The default value is 100


An instance of the LogPythonInterface class for fine grained control over all settings of the logging system.

4.2.1. Creating Log Messages

Log messages can be created in a very similar way in python and C++. You can call one of the logging functions like B2INFO and supply the message as string, for example

B2INFO("This is a log message of severity INFO")

In Python you can supply multiple arguments which will all be converted to string and concatenated to form the log message

for i in range(1,4):
    B2INFO("This is log message number ", i)

which will produce

[INFO] This is log message number 1
[INFO] This is log message number 2
[INFO] This is log message number 3

This works almost the same way in C++ except that you need the << operator to construct the log message from multiple parts

for(int i=1; i<4; ++i) {
  B2INFO("This is log message " << i << " in C++");

Log Variables

New in version release-03-00-00.

However, the log system has an additional feature to include variable parts in a fixed message to simplify grouping of similar log messages: If a log message only differs by a number or detector name it is very hard to filter repeating messages. So we have log message variables which can be used to specify varying parts while having a fixed message.

In Python these can just be given as keyword arguments to the logging functions

B2INFO("This is a log message", number=3.14, text="some text")

In C++ this again almost works the same way but we need to specify the variables a bit more explicitly.

B2INFO("This is a log message" << LogVar("number", 3.14) << LogVar("text", "some text"));

In both cases the names of the variables can be chosen feely and the output should be something like

[INFO] This is a log message
        number = 3.14
        text = some text

Logging functions

To emit log messages from within Python we have these functions:

basf2.B2DEBUG(debugLevel, message, *args, **kwargs)

Print a DEBUG message. The first argument is the debug_level. All additional positional arguments are converted to strings and concatenated to the log message. All keyword arguments are added to the function as Log Variables.

basf2.B2INFO(message, *args, **kwargs)

Print a INFO message. All additional positional arguments are converted to strings and concatenated to the log message. All keyword arguments are added to the function as Log Variables.

basf2.B2RESULT(message, *args, **kwargs)

Print a RESULT message. All additional positional arguments are converted to strings and concatenated to the log message. All keyword arguments are added to the function as Log Variables.

Deprecated since version release-01-00-00: use B2INFO() instead

basf2.B2WARNING(message, *args, **kwargs)

Print a WARNING message. All additional positional arguments are converted to strings and concatenated to the log message. All keyword arguments are added to the function as Log Variables.

basf2.B2ERROR(message, *args, **kwargs)

Print a ERROR message. All additional positional arguments are converted to strings and concatenated to the log message. All keyword arguments are added to the function as Log Variables.

basf2.B2FATAL(message, *args, **kwargs)

Print a FATAL message. All additional positional arguments are converted to strings and concatenated to the log message. All keyword arguments are added to the function as Log Variables.


This also exits the programm with an error and is guaranteed to not return.

The same functions are available in C++ as macros once you included <framework/logging/Logger.h>

4.2.2. The Logging Configuration Objects

The logging object provides a more fine grained control over the settings of the logging system and should be used if more than just a global log level should be changed

class basf2.LogPythonInterface

Logging configuration (for messages generated from C++ or Python), available as a global basf2.logging object in Python. See also basf2.set_log_level() and basf2.set_debug_level().

This class exposes a object called logging to the python interface. With this object it is possible to set all properties of the logging system directly in the steering file in a consistent manner This class also exposes the LogConfig class as well as the LogLevel and LogInfo enums to make setting of properties more transparent by using the names and not just the values. To set or get the log level, one can simply do:

>>> logging.log_level = LogLevel.FATAL
>>> print("Logging level set to", logging.log_level)

This module also allows to send log messages directly from python to ease consistent error reporting throughout the framework

>>> B2WARNING("This is a warning message")

See also

For all features, see b2logging.py

property abort_level

Attribute for setting/getting the log level at which to abort processing. Defaults to FATAL but can be set to a lower level in rare cases.

add_console([(bool)enable_color])None :

Write log output to console. (In addition to existing outputs). If enable_color is not specified color will be enabled if supported

add_file((str)filename[, (bool)append=False])None :

Write log output to given file. (In addition to existing outputs)nn”

  • filename (str) – Filename to to write log messages into

  • append (bool) – If set to True the file will be truncated before writing new messages.

add_json([(bool)complete_info=False])None :

Write log output to console, but format log messages as json objects for simplified parsing by other tools. Each log message will be printed as a one line JSON object.

New in version release-03-00-00.


complete_info (bool) – If this is set to True the complete log information is printed regardless of the LogInfo setting.

add_udp((str)hostname, (int)port)None :

Send the log output as a JSON object to the given hostname and port via UDP.

New in version release-04-00-00.

  • hostname (str) – The hostname to send the message to. If it can not be resolved, an exception will be thrown.

  • port (int) – The port on the host to send the message via UDP.

See also


property debug_level

Attribute for getting/setting the debug level. If debug messages are enabled, their level needs to be at least this high to be printed. Defaults to 100.

property enable_escape_newlines

Enable or disable escaping of newlines in log messages to the console. If this is set to true than any newline character in log messages printed to the console will be replaced by a “n” to ensure that every log messages fits exactly on one line.

New in version release-04-02-00.

property enable_python_logging

Enable or disable logging via python. If this is set to true than log messages will be sent via sys.stdout. This is probably slightly slower but is useful when running in jupyter notebooks or when trying to redirect stdout in python to a buffer. This setting affects all log connections to the console.

New in version release-03-00-00.

enable_summary((bool)on)None :

Enable or disable the error summary printed at the end of processing. Expects one argument whether or not the summary should be shown

get_info((LogLevel)log_level)int :

Get info to print for given log level.


log_level (basf2.LogLevel) – Log level for which to get the display info

property log_level

Attribute for setting/getting the current log level. Messages with a lower level are ignored.

property log_stats

Returns dictionary with message counters.

property max_repetitions

Set the maximum amount of times log messages with the same level and message text (excluding variables) will be repeated before it is suppressed. Suppressed messages will still be counted but not shown for the remainder of the processing.

This affects messages with the same text but different ref:Log Variables. If the same log message is repeated frequently with different variables all of these will be suppressed after the given amount of repetitions.

New in version release-05-00-00.

package((str)package)LogConfig :

Get the LogConfig for given package to set detailed logging pararameters for this package.

>>> logging.package('svd').debug_level = 10
>>> logging.package('svd').set_info(LogLevel.INFO, LogInfo.LEVEL | LogInfo.MESSAGE | LogInfo.FILE)
reset()None :

Remove all configured logging outputs. You can then configure your own via add_file() or add_console()

set_info((LogLevel)log_level, (int)log_info)None :

Set info to print for given log level. Should be an OR combination of basf2.LogInfo constants. As an example, to show only the level and text for all debug messages one could use

>>> basf2.logging.set_info(basf2.LogLevel.DEBUG, basf2.LogInfo.LEVEL | basf2.LogInfo.MESSAGE)
  • log_level (LogLevel) – log level for which to set the display info

  • log_info (int) – Bitmask of basf2.LogInfo constants.

set_package((str)package, (LogConfig)config)None :

Set basf2.LogConfig for given package, see also package().

static terminal_supports_colors()bool :

Returns true if the terminal supports colored output

zero_counters()None :

Reset the per-level message counters.

class basf2.LogConfig

Defines logging settings (log levels and items included in each message) for a certain context, e.g. a module or package.

property abort_level

set or get the severity which causes program abort

property debug_level

set or get the current debug level

get_info((LogLevel)log_level)int :

get the current bitmask of which parts of the log message will be printed for a given log level

property log_level

set or get the current log level

set_abort_level((LogLevel)abort_level)None :

Set the severity which causes program abort.

This can be set to a LogLevel which will cause the processing to be aborted if a message with the given level or higher is encountered. The default is FATAL. It cannot be set any higher but can be lowered.

set_debug_level((int)debug_level)None :

Set the maximum debug level to be shown. Any messages with log level DEBUG and a larger debug level will not be shown.

set_info((LogLevel)log_level, (int)log_info)None :

set the bitmask of LogInfo members to show when printing messages for a given log level

set_log_level((LogLevel)log_level)None :

Set the minimum log level to be shown. Messages with a log level below this value will not be shown at all.

class basf2.LogInfo

The different fields of a log message.

These fields can be used as a bitmask to configure the appearance of log messages.


The severity of the log message, one of basf2.LogLevel


The actual log message


The name of the module active when the message was emitted. Can be empty if no module was active (before/after processing or outside of the normal event loop)


The package the code that emitted the message belongs to. This is empty for messages emitted by python scripts


The function name that emitted the message


The filename containing the code emitting the message


The line number in the file emitting the message

4.3. Module Statistics

The basf2 software takes extensive statistics during event processing about the memory consumption and execution time of all modules. For most users a simple print of the statistics object will be enough and creates a text table of the execution times and memory conumption:

import basf2

However the statistics object provides full access to all the separate values directly in python if needed. See module_statistics.py for a full example.


The memory consumption is measured by looking into /proc/PID/statm between execution calls so for short running modules this might not be accurate but it should give a general idea.


Global instance of a ProcessStatistics object containing all the statistics

class basf2.ProcessStatistics

Interface for retrieving statistics about module execution at runtime or after basf2.process() returns. Should be accessed through a global instance basf2.statistics.

Statistics for event() calls are available as a string representation of the object:

>>> from basf2 import statistics
>>> print(statistics)
Name                  |      Calls | Memory(MB) |    Time(s) |     Time(ms)/Call
RootInput             |        101 |          0 |       0.01 |    0.05 +-   0.02
RootOutput            |        100 |          0 |       0.02 |    0.20 +-   0.87
ProgressBar           |        100 |          0 |       0.00 |    0.00 +-   0.00
Total                 |        101 |          0 |       0.03 |    0.26 +-   0.86

This provides information on the number of calls, elapsed time, and the average difference in resident memory before and after the event() call.


The module responsible for reading (or generating) events usually has one additional event() call which is used to determine whether event processing should stop.


Memory consumption is reporting the difference in memory usage as reported by the kernel before and after the call. This is not the maximum memory the module has consumed. Negative values indicate that this module has freed momemory which was allocated in other modules or function calls.

Information on other calls like initialize(), terminate(), etc. are also available through the different counters defined in StatisticCounters:

>>> print(statistics(statistics.INIT))
>>> print(statistics(statistics.BEGIN_RUN))
>>> print(statistics(statistics.END_RUN))
>>> print(statistics(statistics.TERM))
class ModuleStatistics

Execution statistics for a single module. All member functions take exactly one argument to select which counter to query which defaults to StatisticCounters.TOTAL if omitted.


Helper for pickle.


Return the total number of calls


Return the mean of the memory usage


Return the standard deviation of the memory usage


Return the sum of the total memory usage

property name

property to get the name of the module to be displayed in the statistics


Return the mean of all execution times


Return the correlaction factor between time and memory consumption


Return the standard deviation of all execution times


Return the sum of all execution times

class StatisticCounters

Available types of statistic counters (corresponds to Module functions)


Time spent or memory used in the initialize() function


Time spent or memory used in the beginRun() function


Time spent or memory used in the event() function


Time spent or memory used in the endRun() function


Time spent or memory used in the terminate() function


Time spent or memory used in any module function. This is the sum of all of the above.

__call__(counter=StatisticCounters.EVENT, modules=None)

Calling the statistics object directly like a function will return a string with the execution statistics in human readable form.

  • counter (StatisticCounters) – Which counter to use

  • modules (list[Module]) – A list of modules to include in the returned string. If omitted the statistics for all modules will be included.

  • print the beginRun() statistics for all modules:

    >>> print(statistics(statistics.BEGIN_RUN))
  • print the total execution times and memory consumption but only for the modules module1 and module2

    >>> print(statistics(statistics.TOTAL, [module1, module2]))
  • print the event statistics (default) for only two modules

    >>> print(statistics(modules=[module1, module2]))

Return the event statistics as a string in a human readable form

clear()None :

Clear collected statistics but keep names of modules

get((Module)module)ModuleStatistics :

Get ModuleStatistics for given Module.

get_global()ModuleStatistics :

Get global ModuleStatistics containing total elapsed time etc.

property modules

List of all ModuleStatistics objects.

4.4. Conditions Database

The conditions database is the place where we store additional data needed to interpret and analyse the data that can change over time, for example the detector configuration or calibration constants.

In many cases it should not be necessary to change the configuration but except for maybe adding an extra globaltag to the list via conditions.globaltags

4.5. Additional Functions

basf2.find_file((str)filename[, (str)data_type=''[, (bool)silent=False]])str :

Try to find a file and return its full path

If data_type is empty this function will try to find the file

  1. in $BELLE2_LOCAL_DIR,


  3. relative to the current working directory.

Other known data_type values are


Example data for examples and tutorials. Will try to find the file


  2. relative to the current working directory


Data for Validation purposes. Will try to find the file in


  2. relative to the current working directory

New in version release-03-00-00.

  • filename (str) – relative filename to look for, either in a central place or in the current working directory

  • data_type (str) – case insensitive data type to find. Either empty string or one of "examples" or "validation"

  • silent (bool) – If True don’t print any errors and just return an empty string if the file cannot be found

basf2.get_file_metadata((str)arg1)object :

Return the FileMetaData object for the given output file.

basf2.get_random_seed()str :

Return the current random seed


Adds the standard output stream to the list of logging destinations. The shell logging destination is added to the list by the framework by default.

basf2.log_to_file(filename, append=False)[source]

Adds a text file to the list of logging destinations.

  • filename – The path and filename of the text file

  • append – Should the logging system append the messages to the end of the file (True) or create a new file for each event processing session (False). Default is False.


Resets the logging by removing all logging destinations

basf2.set_nprocesses((int)arg1)None :

Sets number of worker processes for parallel processing.

Can be overridden using the -p argument to basf2.


Setting this to 1 will have one parallel worker job which is almost always slower than just running without parallel processing but is still provided to allow debugging of parallel execution.


nproc (int) – number of worker processes. 0 to disable parallel processing.

basf2.set_random_seed((object)seed)None :

Set the random seed. The argument can be any object and will be converted to a string using the builtin str() function and will be used to initialize the random generator.

basf2.set_streamobjs((list)arg1)None :

Set the names of all DataStore objects which should be sent between the parallel processes. This can be used to improve parallel processing performance by removing objects not required.

4.6. Other Modules

There more tools available in the software framework which might not be of general interest to all users and are separated into different python modules: