Untitled diff
创建于 差异永不过期
import traceback
# pylint: disable=W0122
from cStringIO import StringIO
import cProfile
import cProfile
import pstats
try:
import pstats
except ImportError:
# pstats.py was not available in python 2.6.6 distributed on Debian squeeze
# systems and was included only starting from 2.6.7-2. That is why import
# from a local copy
import _pstats as pstats
import gc
import hashlib
import hashlib
from cStringIO import StringIO
import time
from utils import indent, magic_timeit, magic_memit, getTable
import traceback
import inspect
# from pandas.util.testing import set_trace
class Benchmark(object):
class Benchmark(object):
def __init__(self, code, setup, ncalls=None, repeat=3, cleanup=None,
def __init__(self, code, setup, ncalls=None, repeat=3, cleanup=None,
name=None, description=None, logy=False, db_path=None):
name=None, module_name=None, description=None, start_date=None,
logy=False):
self.code = code
self.code = code
self.setup = setup
self.setup = setup
self.cleanup = cleanup or ''
self.cleanup = cleanup or ''
self.ncalls = ncalls
self.ncalls = ncalls
self.repeat = repeat
self.repeat = repeat
if name is None:
try:
name = _get_assigned_name(inspect.currentframe().f_back)
except:
pass
self.name = name
self.name = name
self.module_name = module_name
self.description = description
self.description = description
self.start_date = start_date
self.logy = logy
self.logy = logy
self.db_path = db_path
def __repr__(self):
def __repr__(self):
return "Benchmark('%s')" % self.name
return "Benchmark('%s')" % self.name
def _setup(self):
def _setup(self):
ns = globals().copy()
ns = globals().copy()
exec self.setup in ns
exec self.setup in ns
return ns
return ns
def _cleanup(self, ns):
exec self.cleanup in ns
@property
@property
def checksum(self):
def checksum(self):
return hashlib.md5(self.setup + self.code + self.cleanup).hexdigest()
return hashlib.md5(self.setup + self.code + self.cleanup).hexdigest()
Text moved from lines 58-60
def _cleanup(self, ns):
def profile(self, ncalls):
exec self.cleanup in ns
prof = cProfile.Profile()
ns = self._setup()
def to_rst(self, results):
code = compile(self.code, '<f>', 'exec')
def f(*args, **kw):
for i in xrange(ncalls):
exec code in ns
prof.runcall(f)
self._cleanup(ns)
return pstats.Stats(prof).sort_stats('cumulative')
def get_results(self, db_path):
from vbench.db import BenchmarkDB
db = BenchmarkDB.get_instance(db_path)
return db.get_benchmark_results(self.checksum)
def run(self):
ns = None
try:
stage = 'setup'
ns = self._setup()
stage = 'benchmark'
result = magic_timeit(ns, self.code, ncalls=self.ncalls,
repeat=self.repeat, force_ms=True)
result['succeeded'] = True
Text moved with changes from lines 88-91 (93.3% similarity)
except:
buf = StringIO()
traceback.print_exc(file=buf)
result = {'succeeded': False,
'stage': stage,
'traceback': buf.getvalue()}
if ns:
self._cleanup(ns)
return result
def _run(self, ns, ncalls, disable_gc=False):
if ncalls is None:
ncalls = self.ncalls
code = self.code
if disable_gc:
gc.disable()
start = time.clock()
for _ in xrange(ncalls):
exec code in ns
elapsed = time.clock() - start
if disable_gc:
gc.enable()
return elapsed
def to_rst(self, image_path=None):
output = """**Benchmark setup**
output = """**Benchmark setup**
.. code-block:: python
.. code-block:: python
%s
%s
**Benchmark statement**
**Benchmark statement**
.. code-block:: python
.. code-block:: python
%s
%s
%s
""" % (indent(self.setup), indent(self.code))
""" % (indent(self.setup), indent(self.code),
getTable(results['runtime'], self.name,
if image_path is not None:
['name', 'repeat', 'timing', 'loops', 'units']))
output += ("**Performance graph**\n\n.. image:: %s"
"\n :width: 6in" % image_path)
return output
return output
Text moved to lines 62-64
def profile(self, ncalls):
def plot(self, db_path, label='time', ax=None, title=True):
prof = cProfile.Profile()
import matplotlib.pyplot as plt
ns = self._setup()
from matplotlib.dates import MonthLocator, DateFormatter
code = compile(self.code, '<f>', 'exec')
results = self.get_results(db_path)
def f(*args, **kwargs):
if ax is None:
for i in xrange(ncalls):
fig = plt.figure()
exec code in ns
ax = fig.add_subplot(111)
prof.runcall(f)
timing = results['timing']
if self.start_date is not None:
timing = timing.truncate(before=self.start_date)
return pstats.Stats(prof).sort_stats('cumulative')
timing.plot(ax=ax, style='b-', label=label)
ax.set_xlabel('Date')
ax.set_ylabel('milliseconds')
def run(self):
if self.logy:
results = {}
ax2 = ax.twinx()
results['memory'] = self.run_memit()
try:
results['runtime'] = self.run_timeit()
timing.plot(ax=ax2, label='%s (log scale)' % label,
style='r-',
logy=self.logy)
ax2.set_ylabel('milliseconds (log scale)')
ax.legend(loc='best')
ax2.legend(loc='best')
except ValueError:
pass
return results
ylo, yhi = ax.get_ylim()
def run_timeit(self):
if ylo < 1:
ns = self._setup()
ax.set_ylim([0, yhi])
try:
formatter = DateFormatter("%b %Y")
result = magic_timeit(ns, self.code, ncalls=self.ncalls,
ax.xaxis.set_major_locator(MonthLocator())
repeat=self.repeat, force_ms=True)
ax.xaxis.set_major_formatter(formatter)
ax.autoscale_view(scalex=True)
result['success'] = True
if title:
ax.set_title(self.name)
Text moved with changes to lines 92-95 (93.3% similarity)
except:
return ax
buf = StringIO()
traceback.print_exc(file=buf)
result = {'success': False, 'traceback': buf.getvalue()}
self._cleanup(ns)
return result
def run_memit(self):
def _get_assigned_name(frame):
ns = self._setup()
import ast
# hackjob to retrieve assigned name for Benchmark
info = inspect.getframeinfo(frame)
line = info.code_context[0]
path = info.filename
lineno = info.lineno - 1
def _has_assignment(line):
try:
try:
result = magic_memit(ns, self.code, ncalls=self.ncalls,
mod = ast.parse(line.strip())
repeat=self.repeat)
return isinstance(mod.body[0], ast.Assign)
except SyntaxError:
return False
result['success'] = True
if not _has_assignment(line):
while not 'Benchmark' in line:
prev = open(path).readlines()[lineno - 1]
line = prev + line
lineno -= 1
except:
if not _has_assignment(line):
buf = StringIO()
prev = open(path).readlines()[lineno - 1]
traceback.print_exc(file=buf)
line = prev + line
result = {'success': False, 'traceback': buf.getvalue()}
varname = line.split('=', 1)[0].strip()
return varname
self._cleanup(ns)
return result
def parse_stmt(frame):
import ast
info = inspect.getframeinfo(frame)
call = info[-2][0]
mod = ast.parse(call)
body = mod.body[0]
if isinstance(body, (ast.Assign, ast.Expr)):
call = body.value
elif isinstance(body, ast.Call):
call = body
return _parse_call(call)
def _parse_call(call):
import ast
func = _maybe_format_attribute(call.func)
str_args = []
for arg in call.args:
if isinstance(arg, ast.Name):
str_args.append(arg.id)
elif isinstance(arg, ast.Call):
formatted = _format_call(arg)
str_args.append(formatted)
return func, str_args, {}
def _format_call(call):
func, args, kwds = _parse_call(call)
content = ''
if args:
content += ', '.join(args)
if kwds:
fmt_kwds = ['%s=%s' % item for item in kwds.iteritems()]
joined_kwds = ', '.join(fmt_kwds)
if args:
content = content + ', ' + joined_kwds
else:
content += joined_kwds
return '%s(%s)' % (func, content)
def _maybe_format_attribute(name):
import ast
if isinstance(name, ast.Attribute):
return _format_attribute(name)
return name.id
def _format_attribute(attr):
import ast
obj = attr.value
if isinstance(attr.value, ast.Attribute):
obj = _format_attribute(attr.value)
else:
obj = obj.id
return '.'.join((obj, attr.attr))
def indent(string, spaces=4):
dent = ' ' * spaces
return '\n'.join([dent + x for x in string.split('\n')])
class BenchmarkSuite(list):
class BenchmarkSuite(list):
"""Basically a list, but the special type is needed for discovery"""
"""Basically a list, but the special type is needed for discovery"""
@property
@property
def benchmarks(self):
def benchmarks(self):
"""Discard non-benchmark elements of the list"""
"""Discard non-benchmark elements of the list"""
return filter(lambda elem: isinstance(elem, Benchmark), self)
return filter(lambda elem: isinstance(elem, Benchmark), self)
# Modified from IPython project, http://ipython.org
def magic_timeit(ns, stmt, ncalls=None, repeat=3, force_ms=False):
"""Time execution of a Python statement or expression
Usage:\\
%timeit [-n<N> -r<R> [-t|-c]] statement
Time execution of a Python statement or expression using the timeit
module.
Options:
-n<N>: execute the given statement <N> times in a loop. If this value
is not given, a fitting value is chosen.
-r<R>: repeat the loop iteration <R> times and take the best result.
Default: 3
-t: use time.time to measure the time, which is the default on Unix.
This function measures wall time.
-c: use time.clock to measure the time, which is the default on
Windows and measures wall time. On Unix, resource.getrusage is used
instead and returns the CPU user time.
-p<P>: use a precision of <P> digits to display the timing result.
Default: 3
Examples:
In [1]: %timeit pass
10000000 loops, best of 3: 53.3 ns per loop
In [2]: u = None
In [3]: %timeit u is None
10000000 loops, best of 3: 184 ns per loop
In [4]: %timeit -r 4 u == None
1000000 loops, best of 4: 242 ns per loop
In [5]: import time
In [6]: %timeit -n1 time.sleep(2)
1 loops, best of 3: 2 s per loop
The times reported by %timeit will be slightly higher than those
reported by the timeit.py script when variables are accessed. This is
due to the fact that %timeit executes the statement in the namespace
of the shell, compared with timeit.py, which uses a single setup
statement to import function or create variables. Generally, the bias
does not matter as long as results from timeit.py are not mixed with
those from %timeit."""
import timeit
import math
units = ["s", "ms", 'us', "ns"]
scaling = [1, 1e3, 1e6, 1e9]
timefunc = timeit.default_timer
timer = timeit.Timer(timer=timefunc)
# this code has tight coupling to the inner workings of timeit.Timer,
# but is there a better way to achieve that the code stmt has access
# to the shell namespace?
src = timeit.template % {'stmt': timeit.reindent(stmt, 8),
'setup': "pass"}
# Track compilation time so it can be reported if too long
# Minimum time above which compilation time will be reported
code = compile(src, "<magic-timeit>", "exec")
exec code in ns
timer.inner = ns["inner"]
if ncalls is None:
# determine number so that 0.2 <= total time < 2.0
number = 1
for _ in range(1, 10):
if timer.timeit(number) >= 0.1:
break
number *= 10
else:
number = ncalls
best = min(timer.repeat(repeat, number)) / number
if force_ms:
order = 1
else:
if best > 0.0 and best < 1000.0:
order = min(-int(math.floor(math.log10(best)) // 3), 3)
elif best >= 1000.0:
order = 0
else:
order = 3
return {'loops': number,
'repeat': repeat,
'timing': best * scaling[order],
'units': units[order]}
def gather_benchmarks(ns):
def gather_benchmarks(ns):
benchmarks = []
benchmarks = []
for v in ns.values():
for v in ns.values():
if isinstance(v, Benchmark):
if isinstance(v, Benchmark):
benchmarks.append(v)
benchmarks.append(v)
elif isinstance(v, BenchmarkSuite):
elif isinstance(v, BenchmarkSuite):
benchmarks.extend(v.benchmarks)
benchmarks.extend(v.benchmarks)
return benchmarks
return benchmarks