]> git.saurik.com Git - apple/icu.git/blobdiff - icuSources/test/perf/unisetperf/unisetperf.cpp
ICU-400.37.tar.gz
[apple/icu.git] / icuSources / test / perf / unisetperf / unisetperf.cpp
diff --git a/icuSources/test/perf/unisetperf/unisetperf.cpp b/icuSources/test/perf/unisetperf/unisetperf.cpp
new file mode 100644 (file)
index 0000000..36d61c1
--- /dev/null
@@ -0,0 +1,441 @@
+/*  
+**********************************************************************
+*   Copyright (C) 2007, International Business Machines
+*   Corporation and others.  All Rights Reserved.
+**********************************************************************
+*   file name:  unisetperf.cpp
+*   encoding:   US-ASCII
+*   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"
+
+#define LENGTHOF(array) (int32_t)(sizeof(array)/sizeof((array)[0]))
+
+// 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, 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;
+}