cpython/Doc/library/profile.rst

624 lines
26 KiB
ReStructuredText
Raw Normal View History

2007-08-15 11:28:22 -03:00
.. _profile:
********************
The Python Profilers
********************
.. sectionauthor:: James Roskind
Merged revisions 75149,75260-75263,75265-75267,75292,75300,75376,75405,75429-75433,75437,75445,75501,75551,75572,75589-75591,75657,75742,75868,75952-75957,76057,76105,76139,76143,76162,76223 via svnmerge from svn+ssh://pythondev@svn.python.org/python/trunk ........ r75149 | gregory.p.smith | 2009-09-29 16:56:31 -0500 (Tue, 29 Sep 2009) | 3 lines Mention issue6972 in extractall docs about overwriting things outside of the supplied path. ........ r75260 | andrew.kuchling | 2009-10-05 16:24:20 -0500 (Mon, 05 Oct 2009) | 1 line Wording fix ........ r75261 | andrew.kuchling | 2009-10-05 16:24:35 -0500 (Mon, 05 Oct 2009) | 1 line Fix narkup ........ r75262 | andrew.kuchling | 2009-10-05 16:25:03 -0500 (Mon, 05 Oct 2009) | 1 line Document 'skip' parameter to constructor ........ r75263 | andrew.kuchling | 2009-10-05 16:25:35 -0500 (Mon, 05 Oct 2009) | 1 line Note side benefit of socket.create_connection() ........ r75265 | andrew.kuchling | 2009-10-05 17:31:11 -0500 (Mon, 05 Oct 2009) | 1 line Reword sentence ........ r75266 | andrew.kuchling | 2009-10-05 17:32:48 -0500 (Mon, 05 Oct 2009) | 1 line Use standard comma punctuation; reword some sentences in the docs ........ r75267 | andrew.kuchling | 2009-10-05 17:42:56 -0500 (Mon, 05 Oct 2009) | 1 line Backport r73983: Document the thousands separator. ........ r75292 | benjamin.peterson | 2009-10-08 22:11:36 -0500 (Thu, 08 Oct 2009) | 1 line death to old CVS keyword ........ r75300 | benjamin.peterson | 2009-10-09 16:48:14 -0500 (Fri, 09 Oct 2009) | 1 line fix some coding style ........ r75376 | benjamin.peterson | 2009-10-11 20:26:07 -0500 (Sun, 11 Oct 2009) | 1 line platform we don't care about ........ r75405 | neil.schemenauer | 2009-10-14 12:17:14 -0500 (Wed, 14 Oct 2009) | 4 lines Issue #1754094: Improve the stack depth calculation in the compiler. There should be no other effect than a small decrease in memory use. Patch by Christopher Tur Lesniewski-Laas. ........ r75429 | benjamin.peterson | 2009-10-14 20:47:28 -0500 (Wed, 14 Oct 2009) | 1 line pep8ify if blocks ........ r75430 | benjamin.peterson | 2009-10-14 20:49:37 -0500 (Wed, 14 Oct 2009) | 1 line use floor division and add a test that exercises the tabsize codepath ........ r75431 | benjamin.peterson | 2009-10-14 20:56:25 -0500 (Wed, 14 Oct 2009) | 1 line change test to what I intended ........ r75432 | benjamin.peterson | 2009-10-14 22:05:39 -0500 (Wed, 14 Oct 2009) | 1 line some cleanups ........ r75433 | benjamin.peterson | 2009-10-14 22:06:55 -0500 (Wed, 14 Oct 2009) | 1 line make inspect.isabstract() always return a boolean; add a test for it, too #7069 ........ r75437 | benjamin.peterson | 2009-10-15 10:44:46 -0500 (Thu, 15 Oct 2009) | 1 line only clear a module's __dict__ if the module is the only one with a reference to it #7140 ........ r75445 | vinay.sajip | 2009-10-16 09:06:44 -0500 (Fri, 16 Oct 2009) | 1 line Issue #7120: logging: Removed import of multiprocessing which is causing crash in GAE. ........ r75501 | antoine.pitrou | 2009-10-18 13:37:11 -0500 (Sun, 18 Oct 2009) | 3 lines Add a comment about unreachable code, and fix a typo ........ r75551 | benjamin.peterson | 2009-10-19 22:14:10 -0500 (Mon, 19 Oct 2009) | 1 line use property api ........ r75572 | benjamin.peterson | 2009-10-20 16:55:17 -0500 (Tue, 20 Oct 2009) | 1 line clarify buffer arg #7178 ........ r75589 | benjamin.peterson | 2009-10-21 21:26:47 -0500 (Wed, 21 Oct 2009) | 1 line whitespace ........ r75590 | benjamin.peterson | 2009-10-21 21:36:47 -0500 (Wed, 21 Oct 2009) | 1 line rewrite to be nice to other implementations ........ r75591 | benjamin.peterson | 2009-10-21 21:50:38 -0500 (Wed, 21 Oct 2009) | 4 lines rewrite for style, clarify, and comments Also, use the hasattr() like scheme of allowing BaseException exceptions through. ........ r75657 | antoine.pitrou | 2009-10-24 07:41:27 -0500 (Sat, 24 Oct 2009) | 3 lines Fix compilation error in debug mode. ........ r75742 | benjamin.peterson | 2009-10-26 17:51:16 -0500 (Mon, 26 Oct 2009) | 1 line use 'is' instead of id() ........ r75868 | benjamin.peterson | 2009-10-27 15:59:18 -0500 (Tue, 27 Oct 2009) | 1 line test expect base classes ........ r75952 | georg.brandl | 2009-10-29 15:38:32 -0500 (Thu, 29 Oct 2009) | 1 line Use the correct function name in docstring. ........ r75953 | georg.brandl | 2009-10-29 15:39:50 -0500 (Thu, 29 Oct 2009) | 1 line Remove mention of the old -X command line switch. ........ r75954 | georg.brandl | 2009-10-29 15:53:00 -0500 (Thu, 29 Oct 2009) | 1 line Use constants instead of magic integers for test result. Do not re-run with --verbose3 for environment changing tests. ........ r75955 | georg.brandl | 2009-10-29 15:54:03 -0500 (Thu, 29 Oct 2009) | 1 line Use a single style for all the docstrings in the math module. ........ r75956 | georg.brandl | 2009-10-29 16:16:34 -0500 (Thu, 29 Oct 2009) | 1 line I do not think the "railroad" program mentioned is still available. ........ r75957 | georg.brandl | 2009-10-29 16:44:56 -0500 (Thu, 29 Oct 2009) | 1 line Fix constant name. ........ r76057 | benjamin.peterson | 2009-11-02 09:06:45 -0600 (Mon, 02 Nov 2009) | 1 line prevent a rather unlikely segfault ........ r76105 | georg.brandl | 2009-11-04 01:38:12 -0600 (Wed, 04 Nov 2009) | 1 line #7259: show correct equivalent for operator.i* operations in docstring; fix minor issues in operator docs. ........ r76139 | benjamin.peterson | 2009-11-06 19:04:38 -0600 (Fri, 06 Nov 2009) | 1 line spelling ........ r76143 | georg.brandl | 2009-11-07 02:26:07 -0600 (Sat, 07 Nov 2009) | 1 line #7271: fix typo. ........ r76162 | benjamin.peterson | 2009-11-08 22:10:53 -0600 (Sun, 08 Nov 2009) | 1 line discuss how to use -p ........ r76223 | georg.brandl | 2009-11-12 02:29:46 -0600 (Thu, 12 Nov 2009) | 1 line Give the profile module a module directive. ........
2009-11-12 22:25:08 -04:00
.. module:: profile
:synopsis: Python source profiler.
2007-08-15 11:28:22 -03:00
2011-01-27 16:38:46 -04:00
**Source code:** :source:`Lib/profile.py` and :source:`Lib/pstats.py`
--------------
2007-08-15 11:28:22 -03:00
.. _profiler-introduction:
Introduction to the profilers
=============================
.. index::
single: deterministic profiling
single: profiling, deterministic
A :dfn:`profiler` is a program that describes the run time performance of a
program, providing a variety of statistics. This documentation describes the
profiler functionality provided in the modules :mod:`cProfile`, :mod:`profile`
and :mod:`pstats`. This profiler provides :dfn:`deterministic profiling` of
Python programs. It also provides a series of report generation tools to allow
users to rapidly examine the results of a profile operation.
2007-08-15 11:28:22 -03:00
2007-10-11 15:01:43 -03:00
The Python standard library provides two different profilers:
2007-08-15 11:28:22 -03:00
1. :mod:`cProfile` is recommended for most users; it's a C extension with
reasonable overhead that makes it suitable for profiling long-running
programs. Based on :mod:`lsprof`, contributed by Brett Rosen and Ted
Czotter.
2007-08-15 11:28:22 -03:00
2. :mod:`profile`, a pure Python module whose interface is imitated by
:mod:`cProfile`. Adds significant overhead to profiled programs. If you're
trying to extend the profiler in some way, the task might be easier with this
module.
2007-08-15 11:28:22 -03:00
The :mod:`profile` and :mod:`cProfile` modules export the same interface, so
Merged revisions 62350-62355,62358-62359,62364-62365,62370,62372-62375,62378-62379,62381 via svnmerge from svn+ssh://pythondev@svn.python.org/python/trunk ........ r62350 | nick.coghlan | 2008-04-15 12:25:31 +0200 (Tue, 15 Apr 2008) | 1 line Issue 2439: add pkgutils.get_data() as a convenience wrapper for the PEP 302 get_data() API (contributed by Paul Moore) ........ r62351 | nick.coghlan | 2008-04-15 12:28:14 +0200 (Tue, 15 Apr 2008) | 1 line Add test file missing from rev 62350 ........ r62352 | benjamin.peterson | 2008-04-15 13:58:46 +0200 (Tue, 15 Apr 2008) | 2 lines Add myself to Doc/ACKS.txt ........ r62353 | andrew.kuchling | 2008-04-15 15:10:07 +0200 (Tue, 15 Apr 2008) | 6 lines Add *,**,@ to index, as suggested by http://farmdev.com/thoughts/24/what-does-the-def-star-variable-or-def-asterisk-parameter-syntax-do-in-python-/ The right entry type to use isn't clear; operator seems wrong, because *,**,@ aren't being used in expressions here. I put them as 'statement'; 'syntax' might be better. ........ r62354 | andrew.kuchling | 2008-04-15 15:10:41 +0200 (Tue, 15 Apr 2008) | 1 line Typo fix ........ r62355 | mark.dickinson | 2008-04-15 22:51:18 +0200 (Tue, 15 Apr 2008) | 3 lines Fix for possible signed overflow: the behaviour of -LONG_MIN is undefined in ANSI C. ........ r62358 | jeroen.ruigrok | 2008-04-16 14:47:01 +0200 (Wed, 16 Apr 2008) | 2 lines Reformat to 80 columns prior to adding documentation. ........ r62359 | jeroen.ruigrok | 2008-04-16 14:57:43 +0200 (Wed, 16 Apr 2008) | 2 lines Add details about the return value for mmap.flush(). ........ r62364 | raymond.hettinger | 2008-04-17 12:48:31 +0200 (Thu, 17 Apr 2008) | 1 line Issue 2648: Add leading zero to money format recipe in the docs. ........ r62365 | jeroen.ruigrok | 2008-04-17 14:39:45 +0200 (Thu, 17 Apr 2008) | 2 lines Be consistent in the use of read-only. ........ r62370 | andrew.kuchling | 2008-04-17 22:44:06 +0200 (Thu, 17 Apr 2008) | 1 line Typo fixes ........ r62372 | andrew.kuchling | 2008-04-18 04:40:47 +0200 (Fri, 18 Apr 2008) | 1 line Use correct parameter name ........ r62373 | andrew.kuchling | 2008-04-18 18:53:09 +0200 (Fri, 18 Apr 2008) | 1 line #2654: fix typo ........ r62374 | andrew.kuchling | 2008-04-18 20:28:23 +0200 (Fri, 18 Apr 2008) | 4 lines Remove personal note from Jim Roskind; it no longer applies, and the e-mail address is for a previous employer. Can we move the big long copyright statement into a sidebar or something? ........ r62375 | andrew.kuchling | 2008-04-18 20:39:55 +0200 (Fri, 18 Apr 2008) | 1 line Rewrite introductory section, and remove old section. (It was already commented-out, but why keep it?) ........ r62378 | skip.montanaro | 2008-04-18 22:35:46 +0200 (Fri, 18 Apr 2008) | 1 line resolve issue 2014 ........ r62379 | benjamin.peterson | 2008-04-18 22:45:33 +0200 (Fri, 18 Apr 2008) | 2 lines Fix indentation in sysmodule.c ........ r62381 | amaury.forgeotdarc | 2008-04-19 01:31:33 +0200 (Sat, 19 Apr 2008) | 3 lines Some tests did not pass on repeated calls (regrtest -R::) Perform additional cleanup, mostly deleting from sys.modules, or clearing the warnings registry. ........
2008-04-18 21:55:37 -03:00
they are mostly interchangeable; :mod:`cProfile` has a much lower overhead but
is newer and might not be available on all systems. :mod:`cProfile` is really a
compatibility layer on top of the internal :mod:`_lsprof` module.
.. note::
The profiler modules are designed to provide an execution profile for a given
program, not for benchmarking purposes (for that, there is :mod:`timeit` for
2011-04-03 10:20:21 -03:00
reasonably accurate results). This particularly applies to benchmarking
Python code against C code: the profilers introduce overhead for Python code,
but not for C-level functions, and so the C code would seem faster than any
Python one.
2007-08-15 11:28:22 -03:00
.. _profile-instant:
Instant User's Manual
=====================
This section is provided for users that "don't want to read the manual." It
provides a very brief overview, and allows a user to rapidly perform profiling
on an existing application.
To profile an application with a main entry point of :func:`foo`, you would add
the following to your module::
import cProfile
cProfile.run('foo()')
(Use :mod:`profile` instead of :mod:`cProfile` if the latter is not available on
your system.)
The above action would cause :func:`foo` to be run, and a series of informative
lines (the profile) to be printed. The above approach is most useful when
working with the interpreter. If you would like to save the results of a
profile into a file for later examination, you can supply a file name as the
second argument to the :func:`run` function::
import cProfile
cProfile.run('foo()', 'fooprof')
The file :file:`cProfile.py` can also be invoked as a script to profile another
script. For example::
python -m cProfile myscript.py
:file:`cProfile.py` accepts two optional arguments on the command line::
cProfile.py [-o output_file] [-s sort_order]
Merged revisions 77484,77487,77561,77570,77593,77603,77608,77667,77702-77703,77739,77858,77887,77889 via svnmerge from svn+ssh://pythondev@svn.python.org/python/trunk ........ r77484 | skip.montanaro | 2010-01-13 19:12:34 -0600 (Wed, 13 Jan 2010) | 4 lines Update PyEval_EvalFrame to PyEval_EvalFrameEx. This looks to have been done partially before. Also add a comment describing how this might have to work with different versions of the interpreter. ........ r77487 | ezio.melotti | 2010-01-14 05:34:10 -0600 (Thu, 14 Jan 2010) | 1 line Fixed typo ........ r77561 | georg.brandl | 2010-01-17 02:42:30 -0600 (Sun, 17 Jan 2010) | 1 line #7699: improve datetime docs: straightforward linking to strftime/strptime section, mark classmethods as such. ........ r77570 | georg.brandl | 2010-01-17 06:14:42 -0600 (Sun, 17 Jan 2010) | 1 line Add note about usage of STRINGLIB_EMPTY. ........ r77593 | georg.brandl | 2010-01-17 17:33:53 -0600 (Sun, 17 Jan 2010) | 1 line Fix internal reference. ........ r77603 | benjamin.peterson | 2010-01-18 17:07:56 -0600 (Mon, 18 Jan 2010) | 8 lines data descriptors do not override the class dictionary if __get__ is not defined Adjust documentation and add a test to verify this behavior. See http://mail.python.org/pipermail/python-dev/2010-January/095637.html for discussion. ........ r77608 | gregory.p.smith | 2010-01-19 02:19:03 -0600 (Tue, 19 Jan 2010) | 6 lines Do not compile stubs for the sha2 series hashes in the openssl hashlib module when the openssl version is too old to support them. That leads both compiled code bloat and to unittests attempting to test implementations that don't exist for comparison purposes on such platforms. ........ r77667 | mark.dickinson | 2010-01-21 12:32:27 -0600 (Thu, 21 Jan 2010) | 1 line Add two more test_strtod test values. ........ r77702 | georg.brandl | 2010-01-23 02:43:31 -0600 (Sat, 23 Jan 2010) | 1 line #7762: fix refcount annotation of PyUnicode_Tailmatch(). ........ r77703 | georg.brandl | 2010-01-23 02:47:54 -0600 (Sat, 23 Jan 2010) | 1 line #7725: fix referencing issue. ........ r77739 | benjamin.peterson | 2010-01-24 21:52:52 -0600 (Sun, 24 Jan 2010) | 1 line mention from_float() in error message ........ r77858 | georg.brandl | 2010-01-30 11:57:48 -0600 (Sat, 30 Jan 2010) | 1 line #7802: fix invalid example (heh). ........ r77887 | georg.brandl | 2010-01-31 12:51:49 -0600 (Sun, 31 Jan 2010) | 5 lines Fix-up ftplib documentation: move exception descriptions to toplevel, not inside a class remove attribution in "versionadded" spell and grammar check docstring of FTP_TLS ........ r77889 | michael.foord | 2010-01-31 13:59:26 -0600 (Sun, 31 Jan 2010) | 1 line Minor modification to unittest documentation. ........
2010-02-02 22:35:45 -04:00
``-s`` only applies to standard output (``-o`` is not supplied).
2007-08-15 11:28:22 -03:00
Look in the :class:`Stats` documentation for valid sort values.
When you wish to review the profile, you should use the methods in the
:mod:`pstats` module. Typically you would load the statistics data as follows::
import pstats
p = pstats.Stats('fooprof')
The class :class:`Stats` (the above code just created an instance of this class)
has a variety of methods for manipulating and printing the data that was just
read into ``p``. When you ran :func:`cProfile.run` above, what was printed was
the result of three method calls::
p.strip_dirs().sort_stats(-1).print_stats()
The first method removed the extraneous path from all the module names. The
second method sorted all the entries according to the standard module/line/name
string that is printed. The third method printed out all the statistics. You
might try the following sort calls:
.. (this is to comply with the semantics of the old profiler).
2007-08-15 11:28:22 -03:00
::
p.sort_stats('name')
p.print_stats()
The first call will actually sort the list by function name, and the second call
will print out the statistics. The following are some interesting calls to
experiment with::
p.sort_stats('cumulative').print_stats(10)
This sorts the profile by cumulative time in a function, and then only prints
the ten most significant lines. If you want to understand what algorithms are
taking time, the above line is what you would use.
If you were looking to see what functions were looping a lot, and taking a lot
of time, you would do::
p.sort_stats('time').print_stats(10)
to sort according to time spent within each function, and then print the
statistics for the top ten functions.
You might also try::
p.sort_stats('file').print_stats('__init__')
This will sort all the statistics by file name, and then print out statistics
for only the class init methods (since they are spelled with ``__init__`` in
them). As one final example, you could try::
p.sort_stats('time', 'cum').print_stats(.5, 'init')
This line sorts statistics with a primary key of time, and a secondary key of
cumulative time, and then prints out some of the statistics. To be specific, the
list is first culled down to 50% (re: ``.5``) of its original size, then only
lines containing ``init`` are maintained, and that sub-sub-list is printed.
If you wondered what functions called the above functions, you could now (``p``
is still sorted according to the last criteria) do::
p.print_callers(.5, 'init')
and you would get a list of callers for each of the listed functions.
If you want more functionality, you're going to have to read the manual, or
guess what the following functions do::
p.print_callees()
p.add('fooprof')
Invoked as a script, the :mod:`pstats` module is a statistics browser for
reading and examining profile dumps. It has a simple line-oriented interface
(implemented using :mod:`cmd`) and interactive help.
.. _deterministic-profiling:
What Is Deterministic Profiling?
================================
:dfn:`Deterministic profiling` is meant to reflect the fact that all *function
call*, *function return*, and *exception* events are monitored, and precise
timings are made for the intervals between these events (during which time the
user's code is executing). In contrast, :dfn:`statistical profiling` (which is
not done by this module) randomly samples the effective instruction pointer, and
deduces where time is being spent. The latter technique traditionally involves
less overhead (as the code does not need to be instrumented), but provides only
relative indications of where time is being spent.
In Python, since there is an interpreter active during execution, the presence
of instrumented code is not required to do deterministic profiling. Python
automatically provides a :dfn:`hook` (optional callback) for each event. In
addition, the interpreted nature of Python tends to add so much overhead to
execution, that deterministic profiling tends to only add small processing
overhead in typical applications. The result is that deterministic profiling is
not that expensive, yet provides extensive run time statistics about the
execution of a Python program.
Call count statistics can be used to identify bugs in code (surprising counts),
and to identify possible inline-expansion points (high call counts). Internal
time statistics can be used to identify "hot loops" that should be carefully
optimized. Cumulative time statistics should be used to identify high level
errors in the selection of algorithms. Note that the unusual handling of
cumulative times in this profiler allows statistics for recursive
implementations of algorithms to be directly compared to iterative
implementations.
Reference Manual -- :mod:`profile` and :mod:`cProfile`
======================================================
.. module:: cProfile
:synopsis: Python profiler
The primary entry point for the profiler is the global function
:func:`profile.run` (resp. :func:`cProfile.run`). It is typically used to create
any profile information. The reports are formatted and printed using methods of
the class :class:`pstats.Stats`. The following is a description of all of these
standard entry points and functions. For a more in-depth view of some of the
code, consider reading the later section on Profiler Extensions, which includes
discussion of how to derive "better" profilers from the classes presented, or
reading the source code for these modules.
.. function:: run(command, filename=None, sort=-1)
2007-08-15 11:28:22 -03:00
This function takes a single argument that can be passed to the :func:`exec`
function, and an optional file name. In all cases this routine attempts to
:func:`exec` its first argument, and gather profiling statistics from the
execution. If no file name is present, then this function automatically
prints a simple profiling report, sorted by the standard name string
(file/line/function-name) that is presented in each line. The following is a
typical output from such a call::
2706 function calls (2004 primitive calls) in 4.504 CPU seconds
Ordered by: standard name
ncalls tottime percall cumtime percall filename:lineno(function)
2 0.006 0.003 0.953 0.477 pobject.py:75(save_objects)
43/3 0.533 0.012 0.749 0.250 pobject.py:99(evaluate)
...
The first line indicates that 2706 calls were monitored. Of those
calls, 2004 were :dfn:`primitive`. We define :dfn:`primitive` to
mean that the call was not induced via recursion. The next line:
``Ordered by: standard name``, indicates that the text string in
the far right column was used to sort the output. The column
headings include:
2007-08-15 11:28:22 -03:00
2009-01-03 17:18:54 -04:00
ncalls
2007-08-15 11:28:22 -03:00
for the number of calls,
2009-01-03 17:18:54 -04:00
tottime
for the total time spent in the given function (and excluding time made in
calls to sub-functions),
2007-08-15 11:28:22 -03:00
2009-01-03 17:18:54 -04:00
percall
2007-08-15 11:28:22 -03:00
is the quotient of ``tottime`` divided by ``ncalls``
2009-01-03 17:18:54 -04:00
cumtime
2007-08-15 11:28:22 -03:00
is the total time spent in this and all subfunctions (from invocation till
exit). This figure is accurate *even* for recursive functions.
2009-01-03 17:18:54 -04:00
percall
2007-08-15 11:28:22 -03:00
is the quotient of ``cumtime`` divided by primitive calls
2009-01-03 17:18:54 -04:00
filename:lineno(function)
2007-08-15 11:28:22 -03:00
provides the respective data of each function
When there are two numbers in the first column (for example,
``43/3``), then the latter is the number of primitive calls, and
the former is the actual number of calls. Note that when the
function does not recurse, these two values are the same, and only
the single figure is printed.
2007-08-15 11:28:22 -03:00
If *sort* is given, it can be one of values allowed for *key*
parameter from :meth:`pstats.Stats.sort_stats`.
2007-08-15 11:28:22 -03:00
.. function:: runctx(command, globals, locals, filename=None)
2007-08-15 11:28:22 -03:00
This function is similar to :func:`run`, with added arguments to supply the
globals and locals dictionaries for the *command* string.
Analysis of the profiler data is done using the :class:`pstats.Stats` class.
2007-08-15 11:28:22 -03:00
.. module:: pstats
:synopsis: Statistics object for use with the profiler.
.. class:: Stats(*filenames, stream=sys.stdout)
2007-08-15 11:28:22 -03:00
This class constructor creates an instance of a "statistics object"
from a *filename* (or set of filenames). :class:`Stats` objects
are manipulated by methods, in order to print useful reports. You
may specify an alternate output stream by giving the keyword
argument, ``stream``.
The file selected by the above constructor must have been created
by the corresponding version of :mod:`profile` or :mod:`cProfile`.
To be specific, there is *no* file compatibility guaranteed with
future versions of this profiler, and there is no compatibility
with files produced by other profilers. If several files are
provided, all the statistics for identical functions will be
coalesced, so that an overall view of several processes can be
considered in a single report. If additional files need to be
combined with data in an existing :class:`Stats` object, the
:meth:`add` method can be used.
2007-08-15 11:28:22 -03:00
.. (such as the old system profiler).
2007-08-15 11:28:22 -03:00
.. _profile-stats:
The :class:`Stats` Class
------------------------
:class:`Stats` objects have the following methods:
.. method:: Stats.strip_dirs()
This method for the :class:`Stats` class removes all leading path
information from file names. It is very useful in reducing the
size of the printout to fit within (close to) 80 columns. This
method modifies the object, and the stripped information is lost.
After performing a strip operation, the object is considered to
have its entries in a "random" order, as it was just after object
initialization and loading. If :meth:`strip_dirs` causes two
function names to be indistinguishable (they are on the same line
of the same filename, and have the same function name), then the
statistics for these two entries are accumulated into a single
entry.
2007-08-15 11:28:22 -03:00
.. method:: Stats.add(*filenames)
2007-08-15 11:28:22 -03:00
This method of the :class:`Stats` class accumulates additional profiling
information into the current profiling object. Its arguments should refer to
filenames created by the corresponding version of :func:`profile.run` or
:func:`cProfile.run`. Statistics for identically named (re: file, line, name)
functions are automatically accumulated into single function statistics.
.. method:: Stats.dump_stats(filename)
Save the data loaded into the :class:`Stats` object to a file named
*filename*. The file is created if it does not exist, and is
overwritten if it already exists. This is equivalent to the method
of the same name on the :class:`profile.Profile` and
:class:`cProfile.Profile` classes.
2007-08-15 11:28:22 -03:00
.. method:: Stats.sort_stats(*keys)
2007-08-15 11:28:22 -03:00
This method modifies the :class:`Stats` object by sorting it
according to the supplied criteria. The argument is typically a
string identifying the basis of a sort (example: ``'time'`` or
``'name'``).
2007-08-15 11:28:22 -03:00
When more than one key is provided, then additional keys are used
as secondary criteria when there is equality in all keys selected
before them. For example, ``sort_stats('name', 'file')`` will sort
all the entries according to their function name, and resolve all
ties (identical function names) by sorting by file name.
2007-08-15 11:28:22 -03:00
Abbreviations can be used for any key names, as long as the abbreviation is
unambiguous. The following are the keys currently defined:
+------------------+----------------------+
| Valid Arg | Meaning |
+==================+======================+
| ``'calls'`` | call count |
+------------------+----------------------+
| ``'cumulative'`` | cumulative time |
+------------------+----------------------+
| ``'cumtime'`` | cumulative time |
+------------------+----------------------+
2007-08-15 11:28:22 -03:00
| ``'file'`` | file name |
+------------------+----------------------+
| ``'filename'`` | file name |
+------------------+----------------------+
2007-08-15 11:28:22 -03:00
| ``'module'`` | file name |
+------------------+----------------------+
| ``'ncalls'`` | call count |
+------------------+----------------------+
2007-08-15 11:28:22 -03:00
| ``'pcalls'`` | primitive call count |
+------------------+----------------------+
| ``'line'`` | line number |
+------------------+----------------------+
| ``'name'`` | function name |
+------------------+----------------------+
| ``'nfl'`` | name/file/line |
+------------------+----------------------+
| ``'stdname'`` | standard name |
+------------------+----------------------+
| ``'time'`` | internal time |
+------------------+----------------------+
| ``'tottime'`` | internal time |
+------------------+----------------------+
2007-08-15 11:28:22 -03:00
Note that all sorts on statistics are in descending order (placing
most time consuming items first), where as name, file, and line
number searches are in ascending order (alphabetical). The subtle
distinction between ``'nfl'`` and ``'stdname'`` is that the
standard name is a sort of the name as printed, which means that
the embedded line numbers get compared in an odd way. For example,
lines 3, 20, and 40 would (if the file names were the same) appear
in the string order 20, 3 and 40. In contrast, ``'nfl'`` does a
numeric compare of the line numbers. In fact,
``sort_stats('nfl')`` is the same as ``sort_stats('name', 'file',
'line')``.
For backward-compatibility reasons, the numeric arguments ``-1``,
``0``, ``1``, and ``2`` are permitted. They are interpreted as
``'stdname'``, ``'calls'``, ``'time'``, and ``'cumulative'``
respectively. If this old style format (numeric) is used, only one
sort key (the numeric key) will be used, and additional arguments
will be silently ignored.
2007-08-15 11:28:22 -03:00
.. For compatibility with the old profiler,
2007-08-15 11:28:22 -03:00
.. method:: Stats.reverse_order()
This method for the :class:`Stats` class reverses the ordering of
the basic list within the object. Note that by default ascending
vs descending order is properly selected based on the sort key of
choice.
2007-08-15 11:28:22 -03:00
.. This method is provided primarily for compatibility with the old profiler.
2007-08-15 11:28:22 -03:00
.. method:: Stats.print_stats(*restrictions)
2007-08-15 11:28:22 -03:00
This method for the :class:`Stats` class prints out a report as
described in the :func:`profile.run` definition.
2007-08-15 11:28:22 -03:00
The order of the printing is based on the last :meth:`sort_stats`
operation done on the object (subject to caveats in :meth:`add` and
:meth:`strip_dirs`).
2007-08-15 11:28:22 -03:00
The arguments provided (if any) can be used to limit the list down
to the significant entries. Initially, the list is taken to be the
complete set of profiled functions. Each restriction is either an
integer (to select a count of lines), or a decimal fraction between
0.0 and 1.0 inclusive (to select a percentage of lines), or a
regular expression (to pattern match the standard name that is
printed; as of Python 1.5b1, this uses the Perl-style regular
expression syntax defined by the :mod:`re` module). If several
restrictions are provided, then they are applied sequentially. For
example::
2007-08-15 11:28:22 -03:00
print_stats(.1, 'foo:')
would first limit the printing to first 10% of list, and then only print
functions that were part of filename :file:`.\*foo:`. In contrast, the
command::
print_stats('foo:', .1)
would limit the list to all functions having file names :file:`.\*foo:`, and
then proceed to only print the first 10% of them.
.. method:: Stats.print_callers(*restrictions)
2007-08-15 11:28:22 -03:00
This method for the :class:`Stats` class prints a list of all functions that
called each function in the profiled database. The ordering is identical to
that provided by :meth:`print_stats`, and the definition of the restricting
argument is also identical. Each caller is reported on its own line. The
format differs slightly depending on the profiler that produced the stats:
* With :mod:`profile`, a number is shown in parentheses after each caller to
show how many times this specific call was made. For convenience, a second
non-parenthesized number repeats the cumulative time spent in the function
at the right.
* With :mod:`cProfile`, each caller is preceded by three numbers:
the number of times this specific call was made, and the total
and cumulative times spent in the current function while it was
invoked by this specific caller.
2007-08-15 11:28:22 -03:00
.. method:: Stats.print_callees(*restrictions)
2007-08-15 11:28:22 -03:00
This method for the :class:`Stats` class prints a list of all
function that were called by the indicated function. Aside from
this reversal of direction of calls (re: called vs was called by),
the arguments and ordering are identical to the
:meth:`print_callers` method.
2007-08-15 11:28:22 -03:00
.. _profile-limits:
Limitations
===========
One limitation has to do with accuracy of timing information. There is a
fundamental problem with deterministic profilers involving accuracy. The most
obvious restriction is that the underlying "clock" is only ticking at a rate
(typically) of about .001 seconds. Hence no measurements will be more accurate
than the underlying clock. If enough measurements are taken, then the "error"
will tend to average out. Unfortunately, removing this first error induces a
second source of error.
The second problem is that it "takes a while" from when an event is dispatched
until the profiler's call to get the time actually *gets* the state of the
clock. Similarly, there is a certain lag when exiting the profiler event
handler from the time that the clock's value was obtained (and then squirreled
away), until the user's code is once again executing. As a result, functions
that are called many times, or call many functions, will typically accumulate
this error. The error that accumulates in this fashion is typically less than
the accuracy of the clock (less than one clock tick), but it *can* accumulate
and become very significant.
The problem is more important with :mod:`profile` than with the lower-overhead
:mod:`cProfile`. For this reason, :mod:`profile` provides a means of
calibrating itself for a given platform so that this error can be
probabilistically (on the average) removed. After the profiler is calibrated, it
will be more accurate (in a least square sense), but it will sometimes produce
negative numbers (when call counts are exceptionally low, and the gods of
probability work against you :-). ) Do *not* be alarmed by negative numbers in
the profile. They should *only* appear if you have calibrated your profiler,
and the results are actually better than without calibration.
.. _profile-calibration:
Calibration
===========
The profiler of the :mod:`profile` module subtracts a constant from each event
handling time to compensate for the overhead of calling the time function, and
socking away the results. By default, the constant is 0. The following
procedure can be used to obtain a better constant for a given platform (see
discussion in section Limitations above). ::
import profile
pr = profile.Profile()
for i in range(5):
print(pr.calibrate(10000))
2007-08-15 11:28:22 -03:00
The method executes the number of Python calls given by the argument, directly
and again under the profiler, measuring the time for both. It then computes the
hidden overhead per profiler event, and returns that as a float. For example,
on an 800 MHz Pentium running Windows 2000, and using Python's time.clock() as
the timer, the magical number is about 12.5e-6.
The object of this exercise is to get a fairly consistent result. If your
computer is *very* fast, or your timer function has poor resolution, you might
have to pass 100000, or even 1000000, to get consistent results.
When you have a consistent answer, there are three ways you can use it::
2007-08-15 11:28:22 -03:00
import profile
# 1. Apply computed bias to all Profile instances created hereafter.
profile.Profile.bias = your_computed_bias
# 2. Apply computed bias to a specific Profile instance.
pr = profile.Profile()
pr.bias = your_computed_bias
# 3. Specify computed bias in instance constructor.
pr = profile.Profile(bias=your_computed_bias)
If you have a choice, you are better off choosing a smaller constant, and then
your results will "less often" show up as negative in profile statistics.
.. _profiler-extensions:
Extensions --- Deriving Better Profilers
========================================
The :class:`Profile` class of both modules, :mod:`profile` and :mod:`cProfile`,
were written so that derived classes could be developed to extend the profiler.
The details are not described here, as doing this successfully requires an
expert understanding of how the :class:`Profile` class works internally. Study
the source code of the module carefully if you want to pursue this.
If all you want to do is change how current time is determined (for example, to
force use of wall-clock time or elapsed process time), pass the timing function
you want to the :class:`Profile` class constructor::
pr = profile.Profile(your_time_func)
The resulting profiler will then call :func:`your_time_func`.
:class:`profile.Profile`
:func:`your_time_func` should return a single number, or a list of
numbers whose sum is the current time (like what :func:`os.times`
returns). If the function returns a single time number, or the
list of returned numbers has length 2, then you will get an
especially fast version of the dispatch routine.
Be warned that you should calibrate the profiler class for the
timer function that you choose. For most machines, a timer that
returns a lone integer value will provide the best results in terms
of low overhead during profiling. (:func:`os.times` is *pretty*
bad, as it returns a tuple of floating point values). If you want
to substitute a better timer in the cleanest fashion, derive a
class and hardwire a replacement dispatch method that best handles
your timer call, along with the appropriate calibration constant.
2007-08-15 11:28:22 -03:00
:class:`cProfile.Profile`
:func:`your_time_func` should return a single number. If it
returns integers, you can also invoke the class constructor with a
second argument specifying the real duration of one unit of time.
For example, if :func:`your_integer_time_func` returns times
measured in thousands of seconds, you would construct the
:class:`Profile` instance as follows::
2007-08-15 11:28:22 -03:00
pr = profile.Profile(your_integer_time_func, 0.001)
As the :mod:`cProfile.Profile` class cannot be calibrated, custom
timer functions should be used with care and should be as fast as
possible. For the best results with a custom timer, it might be
necessary to hard-code it in the C source of the internal
:mod:`_lsprof` module.