Tracing and Profiling Python Script Execution |
If you are writing a complicated Python script, you can link to code that will trace your program execution and display the inputs to and outputs from any functions that you want to track.
In addition, when a function is selected for tracing, the amount of running time used by this function is tracked. This tracking operation is known as function profiling.
 | ||
|
Note that only user-defined functions can be traced and profiled in this way. System functions cannot be traced or profiled. | ||
 |
To implement tracing and profiling in your Python script, use the tracing.py module, which is included in the lib/ifware subdirectory of the directory in which Chameleon is installed.
The tracing library consists of the following functions:
| Function name | Description |
|---|---|
| trace(func1, func2, ...) | Specifies a list of functions that you want to trace. func1, func2 and so on are the names of the functions to trace. |
| no_trace(func1, func2, ...) | Specifies a list of functions that you no longer want to trace. func1, func2 and so on are the names of the functions that you no longer want to trace. |
| start_tracing(quiet) | Start tracing the functions whose names have been specified by calls to trace(). If quiet is set to 1, all normal tracing output, including function calls and returns, is suppressed. If quiet is 0 (the default value), tracing output is displayed. Function profiling is always performed whenever tracing is performed. |
| trace_print(text) | If tracing is being performed, print the text specified by text as part of the trace output. |
| stop_tracing() | Stop tracing. Profiling is also stopped. |
| trace_all(module) | Traces all functions inside the module named module. If module is not specified, trace_all() selects all global functions for tracing and profiling. Note that functions defined inside other functions cannot be traced in this way; for these functions, you must call trace() directly. |
| trace_none() | Specifies that you no longer want to trace any functions. Equivalent to calling no_trace() for all functions for which tracing has been specified. |
| trace_history(count) | Displays profile information for the most frequently used functions that are currently being traced. count specifies the number of functions to display. If count is 0, profiling information is displayed for all traced functions. |
| clear_trace_history() | Clears the existing profiling information. Subsequent calls to trace_history() only display profiling information from the time at which clear_trace_history() was called. |
The output from trace_history() looks something like this:
== Function == | Time w/Children | Time Alone | #Calls echo (module.py) | 1.00s 25.0% | 1.00s 25.0% | 1 charlie (module.py) | 2.00s 50.0% | 2.00s 50.0% | 1 alpha | 4.00s 100.0% | 1.00s 25.0% | 1 |
In this output, Time w/Children tracks the amount of time used by both the specified function and any functions that it calls. Time Alone measures only the time actually spent inside this function. #Calls is the number of times this function was called.
Here is a very simple example of a Python script that uses the tracing.py module:
from ifware.tracing import *
def test2(x):
trace_print('Hello, world!')
return x * 2
def test1(x):
y = test2(x*3)
return x / (y - 42)
def test():
try:
print test1(7)
print 'Failed.'
except:
print 'Okay.'
trace(test, test1, test2)
start_tracing()
test()
|
When tracing has been enabled by start_tracing(), the call to test() produces tracing information that looks something like this:
tracing.py:100 | test() tracing.py:93 | test1(x=7) tracing.py:89 | test2(x=21) tracing.py:86 | Hello, world! tracing.py:89 | test2 => 42 tracing.py:93 | test1 FAILED 'integer division or modulo by zero' tracing.py:100 | test FAILED 'integer division or modulo by zero' Okay. |
 | ||
|
New Feature in Iguana 4.5 The ability to trace and profile Python scripts was added to Iguana 4.5. | ||
 |