ref: 94916808dd9a53ac297ab537851791894b919828
dir: /sys/src/cmd/python/Tools/pybench/Unicode.py/
try: unicode except NameError: raise ImportError from pybench import Test from string import join class ConcatUnicode(Test): version = 2.0 operations = 10 * 5 rounds = 60000 def test(self): # Make sure the strings are *not* interned s = unicode(join(map(str,range(100)))) t = unicode(join(map(str,range(1,101)))) for i in xrange(self.rounds): t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s t + s def calibrate(self): s = unicode(join(map(str,range(100)))) t = unicode(join(map(str,range(1,101)))) for i in xrange(self.rounds): pass class CompareUnicode(Test): version = 2.0 operations = 10 * 5 rounds = 150000 def test(self): # Make sure the strings are *not* interned s = unicode(join(map(str,range(10)))) t = unicode(join(map(str,range(10))) + "abc") for i in xrange(self.rounds): t < s t > s t == s t > s t < s t < s t > s t == s t > s t < s t < s t > s t == s t > s t < s t < s t > s t == s t > s t < s t < s t > s t == s t > s t < s t < s t > s t == s t > s t < s t < s t > s t == s t > s t < s t < s t > s t == s t > s t < s t < s t > s t == s t > s t < s t < s t > s t == s t > s t < s def calibrate(self): s = unicode(join(map(str,range(10)))) t = unicode(join(map(str,range(10))) + "abc") for i in xrange(self.rounds): pass class CreateUnicodeWithConcat(Test): version = 2.0 operations = 10 * 5 rounds = 80000 def test(self): for i in xrange(self.rounds): s = u'om' s = s + u'xbx' s = s + u'xcx' s = s + u'xdx' s = s + u'xex' s = s + u'xax' s = s + u'xbx' s = s + u'xcx' s = s + u'xdx' s = s + u'xex' s = s + u'xax' s = s + u'xbx' s = s + u'xcx' s = s + u'xdx' s = s + u'xex' s = s + u'xax' s = s + u'xbx' s = s + u'xcx' s = s + u'xdx' s = s + u'xex' s = s + u'xax' s = s + u'xbx' s = s + u'xcx' s = s + u'xdx' s = s + u'xex' s = s + u'xax' s = s + u'xbx' s = s + u'xcx' s = s + u'xdx' s = s + u'xex' s = s + u'xax' s = s + u'xbx' s = s + u'xcx' s = s + u'xdx' s = s + u'xex' s = s + u'xax' s = s + u'xbx' s = s + u'xcx' s = s + u'xdx' s = s + u'xex' s = s + u'xax' s = s + u'xbx' s = s + u'xcx' s = s + u'xdx' s = s + u'xex' s = s + u'xax' s = s + u'xbx' s = s + u'xcx' s = s + u'xdx' s = s + u'xex' def calibrate(self): for i in xrange(self.rounds): pass class UnicodeSlicing(Test): version = 2.0 operations = 5 * 7 rounds = 140000 def test(self): s = unicode(join(map(str,range(100)))) for i in xrange(self.rounds): s[50:] s[:25] s[50:55] s[-1:] s[:1] s[2:] s[11:-11] s[50:] s[:25] s[50:55] s[-1:] s[:1] s[2:] s[11:-11] s[50:] s[:25] s[50:55] s[-1:] s[:1] s[2:] s[11:-11] s[50:] s[:25] s[50:55] s[-1:] s[:1] s[2:] s[11:-11] s[50:] s[:25] s[50:55] s[-1:] s[:1] s[2:] s[11:-11] def calibrate(self): s = unicode(join(map(str,range(100)))) for i in xrange(self.rounds): pass ### String methods class UnicodeMappings(Test): version = 2.0 operations = 3 * (5 + 4 + 2 + 1) rounds = 10000 def test(self): s = join(map(unichr,range(20)),'') t = join(map(unichr,range(100)),'') u = join(map(unichr,range(500)),'') v = join(map(unichr,range(1000)),'') for i in xrange(self.rounds): s.lower() s.lower() s.lower() s.lower() s.lower() s.upper() s.upper() s.upper() s.upper() s.upper() s.title() s.title() s.title() s.title() s.title() t.lower() t.lower() t.lower() t.lower() t.upper() t.upper() t.upper() t.upper() t.title() t.title() t.title() t.title() u.lower() u.lower() u.upper() u.upper() u.title() u.title() v.lower() v.upper() v.title() def calibrate(self): s = join(map(unichr,range(20)),'') t = join(map(unichr,range(100)),'') u = join(map(unichr,range(500)),'') v = join(map(unichr,range(1000)),'') for i in xrange(self.rounds): pass class UnicodePredicates(Test): version = 2.0 operations = 5 * 9 rounds = 120000 def test(self): data = (u'abc', u'123', u' ', u'\u1234\u2345\u3456', u'\uFFFF'*10) len_data = len(data) for i in xrange(self.rounds): s = data[i % len_data] s.isalnum() s.isalpha() s.isdecimal() s.isdigit() s.islower() s.isnumeric() s.isspace() s.istitle() s.isupper() s.isalnum() s.isalpha() s.isdecimal() s.isdigit() s.islower() s.isnumeric() s.isspace() s.istitle() s.isupper() s.isalnum() s.isalpha() s.isdecimal() s.isdigit() s.islower() s.isnumeric() s.isspace() s.istitle() s.isupper() s.isalnum() s.isalpha() s.isdecimal() s.isdigit() s.islower() s.isnumeric() s.isspace() s.istitle() s.isupper() s.isalnum() s.isalpha() s.isdecimal() s.isdigit() s.islower() s.isnumeric() s.isspace() s.istitle() s.isupper() def calibrate(self): data = (u'abc', u'123', u' ', u'\u1234\u2345\u3456', u'\uFFFF'*10) len_data = len(data) for i in xrange(self.rounds): s = data[i % len_data] try: import unicodedata except ImportError: pass else: class UnicodeProperties(Test): version = 2.0 operations = 5 * 8 rounds = 100000 def test(self): data = (u'a', u'1', u' ', u'\u1234', u'\uFFFF') len_data = len(data) digit = unicodedata.digit numeric = unicodedata.numeric decimal = unicodedata.decimal category = unicodedata.category bidirectional = unicodedata.bidirectional decomposition = unicodedata.decomposition mirrored = unicodedata.mirrored combining = unicodedata.combining for i in xrange(self.rounds): c = data[i % len_data] digit(c, None) numeric(c, None) decimal(c, None) category(c) bidirectional(c) decomposition(c) mirrored(c) combining(c) digit(c, None) numeric(c, None) decimal(c, None) category(c) bidirectional(c) decomposition(c) mirrored(c) combining(c) digit(c, None) numeric(c, None) decimal(c, None) category(c) bidirectional(c) decomposition(c) mirrored(c) combining(c) digit(c, None) numeric(c, None) decimal(c, None) category(c) bidirectional(c) decomposition(c) mirrored(c) combining(c) digit(c, None) numeric(c, None) decimal(c, None) category(c) bidirectional(c) decomposition(c) mirrored(c) combining(c) def calibrate(self): data = (u'a', u'1', u' ', u'\u1234', u'\uFFFF') len_data = len(data) digit = unicodedata.digit numeric = unicodedata.numeric decimal = unicodedata.decimal category = unicodedata.category bidirectional = unicodedata.bidirectional decomposition = unicodedata.decomposition mirrored = unicodedata.mirrored combining = unicodedata.combining for i in xrange(self.rounds): c = data[i % len_data]