You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
445 lines
15 KiB
445 lines
15 KiB
/*
|
|
**************************************************************************
|
|
* © 2016 and later: Unicode, Inc. and others.
|
|
* License & terms of use: http://www.unicode.org/copyright.html
|
|
**************************************************************************
|
|
**************************************************************************
|
|
* Copyright (C) 2014, International Business Machines
|
|
* Corporation and others. All Rights Reserved.
|
|
**************************************************************************
|
|
* file name: unisetperf.cpp
|
|
* encoding: UTF-8
|
|
* tab size: 8 (not used)
|
|
* indentation:4
|
|
*
|
|
* created on: 2007jan31
|
|
* created by: Markus Scherer
|
|
*/
|
|
|
|
#include <stdio.h>
|
|
#include <stdlib.h>
|
|
#include <string.h>
|
|
#include "unicode/uperf.h"
|
|
#include "unicode/uniset.h"
|
|
#include "unicode/unistr.h"
|
|
#include "uoptions.h"
|
|
#include "cmemory.h" // for UPRV_LENGTHOF
|
|
|
|
// Command-line options specific to unisetperf.
|
|
// Options do not have abbreviations: Force readable command lines.
|
|
// (Using U+0001 for abbreviation characters.)
|
|
enum {
|
|
SET_PATTERN,
|
|
FAST_TYPE,
|
|
UNISETPERF_OPTIONS_COUNT
|
|
};
|
|
|
|
static UOption options[UNISETPERF_OPTIONS_COUNT]={
|
|
UOPTION_DEF("pattern", '\x01', UOPT_REQUIRES_ARG),
|
|
UOPTION_DEF("type", '\x01', UOPT_REQUIRES_ARG)
|
|
};
|
|
|
|
static const char *const unisetperf_usage =
|
|
"\t--pattern UnicodeSet pattern for instantiation.\n"
|
|
"\t Default: [:ID_Continue:]\n"
|
|
"\t--type Type of UnicodeSet: slow fast\n"
|
|
"\t Default: slow\n";
|
|
|
|
// Test object with setup data.
|
|
class UnicodeSetPerformanceTest : public UPerfTest {
|
|
public:
|
|
UnicodeSetPerformanceTest(int32_t argc, const char *argv[], UErrorCode &status)
|
|
: UPerfTest(argc, argv, options, UPRV_LENGTHOF(options), unisetperf_usage, status),
|
|
utf8(NULL), utf8Length(0), countInputCodePoints(0), spanCount(0) {
|
|
if (U_SUCCESS(status)) {
|
|
UnicodeString pattern=UnicodeString(options[SET_PATTERN].value, -1, US_INV).unescape();
|
|
set.applyPattern(pattern, status);
|
|
prefrozen=set;
|
|
if(0==strcmp(options[FAST_TYPE].value, "fast")) {
|
|
set.freeze();
|
|
}
|
|
|
|
int32_t inputLength;
|
|
UPerfTest::getBuffer(inputLength, status);
|
|
if(U_SUCCESS(status) && inputLength>0) {
|
|
countInputCodePoints = u_countChar32(buffer, bufferLen);
|
|
|
|
countSpans();
|
|
|
|
// Preflight the UTF-8 length and allocate utf8.
|
|
u_strToUTF8(NULL, 0, &utf8Length, buffer, bufferLen, &status);
|
|
if(status==U_BUFFER_OVERFLOW_ERROR) {
|
|
utf8=(char *)malloc(utf8Length);
|
|
if(utf8!=NULL) {
|
|
status=U_ZERO_ERROR;
|
|
u_strToUTF8(utf8, utf8Length, NULL, buffer, bufferLen, &status);
|
|
} else {
|
|
status=U_MEMORY_ALLOCATION_ERROR;
|
|
}
|
|
}
|
|
|
|
if(verbose) {
|
|
printf("code points:%ld len16:%ld len8:%ld spans:%ld "
|
|
"cp/span:%.3g UChar/span:%.3g B/span:%.3g B/cp:%.3g\n",
|
|
(long)countInputCodePoints, (long)bufferLen, (long)utf8Length, (long)spanCount,
|
|
(double)countInputCodePoints/spanCount, (double)bufferLen/spanCount, (double)utf8Length/spanCount,
|
|
(double)utf8Length/countInputCodePoints);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
virtual UPerfFunction* runIndexedTest(int32_t index, UBool exec, const char* &name, char* par = NULL);
|
|
|
|
// Count spans of characters that are in the set,
|
|
// and spans of characters that are not in the set.
|
|
// If the very first character is in the set, then one additional
|
|
// not-span is counted.
|
|
void countSpans() {
|
|
const UChar *s=getBuffer();
|
|
int32_t length=getBufferLen();
|
|
int32_t i=0;
|
|
UBool tf=FALSE;
|
|
while(i<length) {
|
|
i=span(s, length, i, tf);
|
|
tf=(UBool)(!tf);
|
|
++spanCount;
|
|
}
|
|
}
|
|
int32_t span(const UChar *s, int32_t length, int32_t start, UBool tf) const {
|
|
UChar32 c;
|
|
int32_t prev;
|
|
while((prev=start)<length) {
|
|
U16_NEXT(s, start, length, c);
|
|
if(tf!=set.contains(c)) {
|
|
break;
|
|
}
|
|
}
|
|
return prev;
|
|
}
|
|
|
|
const UChar *getBuffer() const { return buffer; }
|
|
int32_t getBufferLen() const { return bufferLen; }
|
|
|
|
char *utf8;
|
|
int32_t utf8Length;
|
|
|
|
// Number of code points in the input text.
|
|
int32_t countInputCodePoints;
|
|
int32_t spanCount;
|
|
|
|
UnicodeSet set;
|
|
UnicodeSet prefrozen;
|
|
};
|
|
|
|
// Performance test function object.
|
|
class Command : public UPerfFunction {
|
|
protected:
|
|
Command(const UnicodeSetPerformanceTest &testcase) : testcase(testcase) {}
|
|
|
|
public:
|
|
virtual ~Command() {}
|
|
|
|
// virtual void call(UErrorCode* pErrorCode) { ... }
|
|
|
|
virtual long getOperationsPerIteration() {
|
|
// Number of code points tested:
|
|
// Input code points, plus one for the end of each span except the last span.
|
|
return testcase.countInputCodePoints+testcase.spanCount-1;
|
|
}
|
|
|
|
virtual long getEventsPerIteration() {
|
|
return testcase.spanCount;
|
|
}
|
|
|
|
const UnicodeSetPerformanceTest &testcase;
|
|
};
|
|
|
|
class Contains : public Command {
|
|
protected:
|
|
Contains(const UnicodeSetPerformanceTest &testcase) : Command(testcase) {
|
|
// Verify that the frozen set is equal to the unfrozen one.
|
|
UnicodeSet set;
|
|
UChar32 c;
|
|
|
|
for(c=0; c<=0x10ffff; ++c) {
|
|
if(testcase.set.contains(c)) {
|
|
set.add(c);
|
|
}
|
|
}
|
|
if(set!=testcase.set) {
|
|
fprintf(stderr, "error: frozen set != original!\n");
|
|
}
|
|
}
|
|
public:
|
|
static UPerfFunction* get(const UnicodeSetPerformanceTest &testcase) {
|
|
return new Contains(testcase);
|
|
}
|
|
virtual void call(UErrorCode* pErrorCode) {
|
|
const UnicodeSet &set=testcase.set;
|
|
const UChar *s=testcase.getBuffer();
|
|
int32_t length=testcase.getBufferLen();
|
|
int32_t count=0;
|
|
int32_t i=0;
|
|
UBool tf=FALSE;
|
|
while(i<length) {
|
|
i+=span(set, s+i, length-i, tf);
|
|
tf=(UBool)(!tf);
|
|
++count;
|
|
}
|
|
if(count!=testcase.spanCount) {
|
|
fprintf(stderr, "error: Contains() count=%ld != %ld=UnicodeSetPerformanceTest.spanCount\n",
|
|
(long)count, (long)testcase.spanCount);
|
|
}
|
|
}
|
|
static int32_t span(const UnicodeSet &set, const UChar *s, int32_t length, UBool tf) {
|
|
UChar32 c;
|
|
int32_t start=0, prev;
|
|
while((prev=start)<length) {
|
|
U16_NEXT(s, start, length, c);
|
|
if(tf!=set.contains(c)) {
|
|
break;
|
|
}
|
|
}
|
|
return prev;
|
|
}
|
|
};
|
|
|
|
class SpanUTF16 : public Command {
|
|
protected:
|
|
SpanUTF16(const UnicodeSetPerformanceTest &testcase) : Command(testcase) {
|
|
// Verify that the frozen set is equal to the unfrozen one.
|
|
UnicodeSet set;
|
|
UChar utf16[2];
|
|
UChar32 c, c2;
|
|
|
|
for(c=0; c<=0xffff; ++c) {
|
|
utf16[0]=(UChar)c;
|
|
if(testcase.set.span(utf16, 1, USET_SPAN_CONTAINED)>0) {
|
|
set.add(c);
|
|
}
|
|
}
|
|
for(c=0xd800; c<=0xdbff; ++c) {
|
|
utf16[0]=(UChar)c;
|
|
for(c2=0xdc00; c2<=0xdfff; ++c2) {
|
|
utf16[1]=(UChar)c2;
|
|
if(testcase.set.span(utf16, 2, USET_SPAN_CONTAINED)>0) {
|
|
set.add(U16_GET_SUPPLEMENTARY(c, c2));
|
|
}
|
|
}
|
|
}
|
|
|
|
if(set!=testcase.set) {
|
|
fprintf(stderr, "error: frozen set != original!\n");
|
|
}
|
|
}
|
|
public:
|
|
static UPerfFunction* get(const UnicodeSetPerformanceTest &testcase) {
|
|
return new SpanUTF16(testcase);
|
|
}
|
|
virtual void call(UErrorCode* pErrorCode) {
|
|
const UnicodeSet &set=testcase.set;
|
|
const UChar *s=testcase.getBuffer();
|
|
int32_t length=testcase.getBufferLen();
|
|
int32_t count=0;
|
|
int32_t i=0;
|
|
UBool tf=FALSE;
|
|
while(i<length) {
|
|
i+=set.span(s+i, length-i, (USetSpanCondition)tf);
|
|
tf=(UBool)(!tf);
|
|
++count;
|
|
}
|
|
if(count!=testcase.spanCount) {
|
|
fprintf(stderr, "error: SpanUTF16() count=%ld != %ld=UnicodeSetPerformanceTest.spanCount\n",
|
|
(long)count, (long)testcase.spanCount);
|
|
}
|
|
}
|
|
};
|
|
|
|
class SpanBackUTF16 : public Command {
|
|
protected:
|
|
SpanBackUTF16(const UnicodeSetPerformanceTest &testcase) : Command(testcase) {
|
|
// Verify that the frozen set is equal to the unfrozen one.
|
|
UnicodeSet set;
|
|
UChar utf16[2];
|
|
UChar32 c, c2;
|
|
|
|
for(c=0; c<=0xffff; ++c) {
|
|
utf16[0]=(UChar)c;
|
|
if(testcase.set.spanBack(utf16, 1, USET_SPAN_CONTAINED)==0) {
|
|
set.add(c);
|
|
}
|
|
}
|
|
for(c=0xd800; c<=0xdbff; ++c) {
|
|
utf16[0]=(UChar)c;
|
|
for(c2=0xdc00; c2<=0xdfff; ++c2) {
|
|
utf16[1]=(UChar)c2;
|
|
if(testcase.set.spanBack(utf16, 2, USET_SPAN_CONTAINED)==0) {
|
|
set.add(U16_GET_SUPPLEMENTARY(c, c2));
|
|
}
|
|
}
|
|
}
|
|
|
|
if(set!=testcase.set) {
|
|
fprintf(stderr, "error: frozen set != original!\n");
|
|
}
|
|
}
|
|
public:
|
|
static UPerfFunction* get(const UnicodeSetPerformanceTest &testcase) {
|
|
return new SpanBackUTF16(testcase);
|
|
}
|
|
virtual void call(UErrorCode* pErrorCode) {
|
|
const UnicodeSet &set=testcase.set;
|
|
const UChar *s=testcase.getBuffer();
|
|
int32_t length=testcase.getBufferLen();
|
|
int32_t count=0;
|
|
/*
|
|
* Get the same spans as with span() where we always start with a not-contained span.
|
|
* If testcase.spanCount is an odd number, then the last span() was not-contained.
|
|
* The last spanBack() must be not-contained to match the first span().
|
|
*/
|
|
UBool tf=(UBool)((testcase.spanCount&1)==0);
|
|
while(length>0 || !tf) {
|
|
length=set.spanBack(s, length, (USetSpanCondition)tf);
|
|
tf=(UBool)(!tf);
|
|
++count;
|
|
}
|
|
if(count!=testcase.spanCount) {
|
|
fprintf(stderr, "error: SpanBackUTF16() count=%ld != %ld=UnicodeSetPerformanceTest.spanCount\n",
|
|
(long)count, (long)testcase.spanCount);
|
|
}
|
|
}
|
|
};
|
|
|
|
class SpanUTF8 : public Command {
|
|
protected:
|
|
SpanUTF8(const UnicodeSetPerformanceTest &testcase) : Command(testcase) {
|
|
// Verify that the frozen set is equal to the unfrozen one.
|
|
UnicodeSet set;
|
|
char utf8[4];
|
|
UChar32 c;
|
|
int32_t length;
|
|
|
|
for(c=0; c<=0x10ffff; ++c) {
|
|
if(c==0xd800) {
|
|
c=0xe000;
|
|
}
|
|
length=0;
|
|
U8_APPEND_UNSAFE(utf8, length, c);
|
|
if(testcase.set.spanUTF8(utf8, length, USET_SPAN_CONTAINED)>0) {
|
|
set.add(c);
|
|
}
|
|
}
|
|
if(set!=testcase.set) {
|
|
fprintf(stderr, "error: frozen set != original!\n");
|
|
}
|
|
}
|
|
public:
|
|
static UPerfFunction* get(const UnicodeSetPerformanceTest &testcase) {
|
|
return new SpanUTF8(testcase);
|
|
}
|
|
virtual void call(UErrorCode* pErrorCode) {
|
|
const UnicodeSet &set=testcase.set;
|
|
const char *s=testcase.utf8;
|
|
int32_t length=testcase.utf8Length;
|
|
int32_t count=0;
|
|
int32_t i=0;
|
|
UBool tf=FALSE;
|
|
while(i<length) {
|
|
i+=set.spanUTF8(s+i, length-i, (USetSpanCondition)tf);
|
|
tf=(UBool)(!tf);
|
|
++count;
|
|
}
|
|
if(count!=testcase.spanCount) {
|
|
fprintf(stderr, "error: SpanUTF8() count=%ld != %ld=UnicodeSetPerformanceTest.spanCount\n",
|
|
(long)count, (long)testcase.spanCount);
|
|
}
|
|
}
|
|
};
|
|
|
|
class SpanBackUTF8 : public Command {
|
|
protected:
|
|
SpanBackUTF8(const UnicodeSetPerformanceTest &testcase) : Command(testcase) {
|
|
// Verify that the frozen set is equal to the unfrozen one.
|
|
UnicodeSet set;
|
|
char utf8[4];
|
|
UChar32 c;
|
|
int32_t length;
|
|
|
|
for(c=0; c<=0x10ffff; ++c) {
|
|
if(c==0xd800) {
|
|
c=0xe000;
|
|
}
|
|
length=0;
|
|
U8_APPEND_UNSAFE(utf8, length, c);
|
|
if(testcase.set.spanBackUTF8(utf8, length, USET_SPAN_CONTAINED)==0) {
|
|
set.add(c);
|
|
}
|
|
}
|
|
if(set!=testcase.set) {
|
|
fprintf(stderr, "error: frozen set != original!\n");
|
|
}
|
|
}
|
|
public:
|
|
static UPerfFunction* get(const UnicodeSetPerformanceTest &testcase) {
|
|
return new SpanBackUTF8(testcase);
|
|
}
|
|
virtual void call(UErrorCode* pErrorCode) {
|
|
const UnicodeSet &set=testcase.set;
|
|
const char *s=testcase.utf8;
|
|
int32_t length=testcase.utf8Length;
|
|
int32_t count=0;
|
|
/*
|
|
* Get the same spans as with span() where we always start with a not-contained span.
|
|
* If testcase.spanCount is an odd number, then the last span() was not-contained.
|
|
* The last spanBack() must be not-contained to match the first span().
|
|
*/
|
|
UBool tf=(UBool)((testcase.spanCount&1)==0);
|
|
while(length>0 || !tf) {
|
|
length=set.spanBackUTF8(s, length, (USetSpanCondition)tf);
|
|
tf=(UBool)(!tf);
|
|
++count;
|
|
}
|
|
if(count!=testcase.spanCount) {
|
|
fprintf(stderr, "error: SpanBackUTF8() count=%ld != %ld=UnicodeSetPerformanceTest.spanCount\n",
|
|
(long)count, (long)testcase.spanCount);
|
|
}
|
|
}
|
|
};
|
|
|
|
UPerfFunction* UnicodeSetPerformanceTest::runIndexedTest(int32_t index, UBool exec, const char* &name, char* par) {
|
|
switch (index) {
|
|
case 0: name = "Contains"; if (exec) return Contains::get(*this); break;
|
|
case 1: name = "SpanUTF16"; if (exec) return SpanUTF16::get(*this); break;
|
|
case 2: name = "SpanBackUTF16";if (exec) return SpanBackUTF16::get(*this); break;
|
|
case 3: name = "SpanUTF8"; if (exec) return SpanUTF8::get(*this); break;
|
|
case 4: name = "SpanBackUTF8"; if (exec) return SpanBackUTF8::get(*this); break;
|
|
default: name = ""; break;
|
|
}
|
|
return NULL;
|
|
}
|
|
|
|
int main(int argc, const char *argv[])
|
|
{
|
|
// Default values for command-line options.
|
|
options[SET_PATTERN].value = "[:ID_Continue:]";
|
|
options[FAST_TYPE].value = "slow";
|
|
|
|
UErrorCode status = U_ZERO_ERROR;
|
|
UnicodeSetPerformanceTest test(argc, argv, status);
|
|
|
|
if (U_FAILURE(status)){
|
|
printf("The error is %s\n", u_errorName(status));
|
|
test.usage();
|
|
return status;
|
|
}
|
|
|
|
if (test.run() == FALSE){
|
|
fprintf(stderr, "FAILED: Tests could not be run, please check the "
|
|
"arguments.\n");
|
|
return 1;
|
|
}
|
|
|
|
return 0;
|
|
}
|