Merge pull request #553 from miloyip/issue158_parsestdstring

Issue158 parsestdstring

include/rapidjson/document.h

@@ -20,6 +20,8 @@
 #include "reader.h"
 #include "internal/meta.h"
 #include "internal/strfunc.h"
+#include "memorystream.h"
+#include "encodedstream.h"
 #include <new>      // placement new
 
 #ifdef _MSC_VER
@@ -2224,6 +2226,42 @@ public:
     GenericDocument& Parse(const Ch* str) {
         return Parse<kParseDefaultFlags>(str);
     }
+
+    template <unsigned parseFlags, typename SourceEncoding>
+    GenericDocument& Parse(const typename SourceEncoding::Ch* str, size_t length) {
+        RAPIDJSON_ASSERT(!(parseFlags & kParseInsituFlag));
+        MemoryStream ms(static_cast<const char*>(str), length * sizeof(typename SourceEncoding::Ch));
+        EncodedInputStream<SourceEncoding, MemoryStream> is(ms);
+        ParseStream<parseFlags, SourceEncoding>(is);
+        return *this;
+    }
+
+    template <unsigned parseFlags>
+    GenericDocument& Parse(const Ch* str, size_t length) {
+        return Parse<parseFlags, Encoding>(str, length);
+    }
+    
+    GenericDocument& Parse(const Ch* str, size_t length) {
+        return Parse<kParseDefaultFlags>(str, length);
+    }
+
+#if RAPIDJSON_HAS_STDSTRING
+    template <unsigned parseFlags, typename SourceEncoding>
+    GenericDocument& Parse(const std::basic_string<typename SourceEncoding::Ch>& str) {
+        // c_str() is constant complexity according to standard. Should be faster than Parse(const char*, size_t)
+        return Parse<parseFlags, SourceEncoding>(str.c_str());
+    }
+
+    template <unsigned parseFlags>
+    GenericDocument& Parse(const std::basic_string<Ch>& str) {
+        return Parse<parseFlags, Encoding>(str);
+    }
+
+    GenericDocument& Parse(const std::basic_string<Ch>& str) {
+        return Parse<kParseDefaultFlags>(str);
+    }
+#endif // RAPIDJSON_HAS_STDSTRING    
+
     //!@}
 
     //!@name Handling parse errors

include/rapidjson/encodedstream.h

@@ -16,6 +16,7 @@
 #define RAPIDJSON_ENCODEDSTREAM_H_
 
 #include "stream.h"
+#include "memorystream.h"
 
 #ifdef __GNUC__
 RAPIDJSON_DIAG_PUSH
@@ -62,6 +63,30 @@ private:
     Ch current_;
 };
 
+//! Specialized for UTF8 MemoryStream.
+template <>
+class EncodedInputStream<UTF8<>, MemoryStream> {
+public:
+    typedef UTF8<>::Ch Ch;
+
+    EncodedInputStream(MemoryStream& is) : is_(is) {
+        if (static_cast<unsigned char>(is_.Peek()) == 0xEFu) is_.Take();
+        if (static_cast<unsigned char>(is_.Peek()) == 0xBBu) is_.Take();
+        if (static_cast<unsigned char>(is_.Peek()) == 0xBFu) is_.Take();
+    }
+    Ch Peek() const { return is_.Peek(); }
+    Ch Take() { return is_.Take(); }
+    size_t Tell() const { return is_.Tell(); }
+
+    // Not implemented
+    void Put(Ch) {}
+    void Flush() {} 
+    Ch* PutBegin() { return 0; }
+    size_t PutEnd(Ch*) { return 0; }
+
+    MemoryStream& is_;
+};
+
 //! Output byte stream wrapper with statically bound encoding.
 /*!
     \tparam Encoding The interpretation of encoding of the stream. Either UTF8, UTF16LE, UTF16BE, UTF32LE, UTF32BE.

include/rapidjson/memorystream.h

@@ -42,8 +42,8 @@ struct MemoryStream {
 
     MemoryStream(const Ch *src, size_t size) : src_(src), begin_(src), end_(src + size), size_(size) {}
 
-    Ch Peek() const { return (src_ == end_) ? '\0' : *src_; }
-    Ch Take() { return (src_ == end_) ? '\0' : *src_++; }
+    Ch Peek() const { return RAPIDJSON_UNLIKELY(src_ == end_) ? '\0' : *src_; }
+    Ch Take() { return RAPIDJSON_UNLIKELY(src_ == end_) ? '\0' : *src_++; }
     size_t Tell() const { return static_cast<size_t>(src_ - begin_); }
 
     Ch* PutBegin() { RAPIDJSON_ASSERT(false); return 0; }

include/rapidjson/reader.h

@@ -19,6 +19,7 @@
 
 #include "allocators.h"
 #include "stream.h"
+#include "encodedstream.h"
 #include "internal/meta.h"
 #include "internal/stack.h"
 #include "internal/strtod.h"
@@ -259,6 +260,12 @@ void SkipWhitespace(InputStream& is) {
         s.Take();
 }
 
+inline const char* SkipWhitespace(const char* p, const char* end) {
+    while (p != end && (*p == ' ' || *p == '\n' || *p == '\r' || *p == '\t'))
+        ++p;
+    return p;
+}
+
 #ifdef RAPIDJSON_SSE42
 //! Skip whitespace with SSE 4.2 pcmpistrm instruction, testing 16 8-byte characters at once.
 inline const char *SkipWhitespace_SIMD(const char* p) {
@@ -295,6 +302,34 @@ inline const char *SkipWhitespace_SIMD(const char* p) {
     }
 }
 
+inline const char *SkipWhitespace_SIMD(const char* p, const char* end) {
+    // Fast return for single non-whitespace
+    if (p != end && (*p == ' ' || *p == '\n' || *p == '\r' || *p == '\t'))
+        ++p;
+    else
+        return p;
+
+    // The middle of string using SIMD
+    static const char whitespace[16] = " \n\r\t";
+    const __m128i w = _mm_loadu_si128(reinterpret_cast<const __m128i *>(&whitespace[0]));
+
+    for (; p <= end - 16; p += 16) {
+        const __m128i s = _mm_loadu_si128(reinterpret_cast<const __m128i *>(p));
+        const int r = _mm_cvtsi128_si32(_mm_cmpistrm(w, s, _SIDD_UBYTE_OPS | _SIDD_CMP_EQUAL_ANY | _SIDD_BIT_MASK | _SIDD_NEGATIVE_POLARITY));
+        if (r != 0) {   // some of characters is non-whitespace
+#ifdef _MSC_VER         // Find the index of first non-whitespace
+            unsigned long offset;
+            _BitScanForward(&offset, r);
+            return p + offset;
+#else
+            return p + __builtin_ffs(r) - 1;
+#endif
+        }
+    }
+
+    return SkipWhitespace(p, end);
+}
+
 #elif defined(RAPIDJSON_SSE2)
 
 //! Skip whitespace with SSE2 instructions, testing 16 8-byte characters at once.
@@ -342,6 +377,44 @@ inline const char *SkipWhitespace_SIMD(const char* p) {
     }
 }
 
+inline const char *SkipWhitespace_SIMD(const char* p, const char* end) {
+    // Fast return for single non-whitespace
+    if (p != end && (*p == ' ' || *p == '\n' || *p == '\r' || *p == '\t'))
+        ++p;
+    else
+        return p;
+
+    // The rest of string
+    #define C16(c) { c, c, c, c, c, c, c, c, c, c, c, c, c, c, c, c }
+    static const char whitespaces[4][16] = { C16(' '), C16('\n'), C16('\r'), C16('\t') };
+    #undef C16
+
+    const __m128i w0 = _mm_loadu_si128(reinterpret_cast<const __m128i *>(&whitespaces[0][0]));
+    const __m128i w1 = _mm_loadu_si128(reinterpret_cast<const __m128i *>(&whitespaces[1][0]));
+    const __m128i w2 = _mm_loadu_si128(reinterpret_cast<const __m128i *>(&whitespaces[2][0]));
+    const __m128i w3 = _mm_loadu_si128(reinterpret_cast<const __m128i *>(&whitespaces[3][0]));
+
+    for (; p <= end - 16; p += 16) {
+        const __m128i s = _mm_loadu_si128(reinterpret_cast<const __m128i *>(p));
+        __m128i x = _mm_cmpeq_epi8(s, w0);
+        x = _mm_or_si128(x, _mm_cmpeq_epi8(s, w1));
+        x = _mm_or_si128(x, _mm_cmpeq_epi8(s, w2));
+        x = _mm_or_si128(x, _mm_cmpeq_epi8(s, w3));
+        unsigned short r = static_cast<unsigned short>(~_mm_movemask_epi8(x));
+        if (r != 0) {   // some of characters may be non-whitespace
+#ifdef _MSC_VER         // Find the index of first non-whitespace
+            unsigned long offset;
+            _BitScanForward(&offset, r);
+            return p + offset;
+#else
+            return p + __builtin_ffs(r) - 1;
+#endif
+        }
+    }
+
+    return SkipWhitespace(p, end);
+}
+
 #endif // RAPIDJSON_SSE2
 
 #ifdef RAPIDJSON_SIMD
@@ -354,6 +427,10 @@ template<> inline void SkipWhitespace(InsituStringStream& is) {
 template<> inline void SkipWhitespace(StringStream& is) {
     is.src_ = SkipWhitespace_SIMD(is.src_);
 }
+
+template<> inline void SkipWhitespace(EncodedInputStream<UTF8<>, MemoryStream>& is) {
+    is.is_.src_ = SkipWhitespace_SIMD(is.is_.src_, is.is_.end_);
+}
 #endif // RAPIDJSON_SIMD
 
 ///////////////////////////////////////////////////////////////////////////////

test/perftest/perftest.h

@@ -30,6 +30,8 @@
 #  define RAPIDJSON_SSE2
 #endif
 
+#define RAPIDJSON_HAS_STDSTRING 1
+
 ////////////////////////////////////////////////////////////////////////////////
 // Google Test
 

test/perftest/rapidjsontest.cpp

@@ -187,6 +187,25 @@ TEST_F(RapidJson, SIMD_SUFFIX(DocumentParse_MemoryPoolAllocator)) {
     }
 }
 
+TEST_F(RapidJson, SIMD_SUFFIX(DocumentParseLength_MemoryPoolAllocator)) {
+    for (size_t i = 0; i < kTrialCount; i++) {
+        Document doc;
+        doc.Parse(json_, length_);
+        ASSERT_TRUE(doc.IsObject());
+    }
+}
+
+#if RAPIDJSON_HAS_STDSTRING
+TEST_F(RapidJson, SIMD_SUFFIX(DocumentParseStdString_MemoryPoolAllocator)) {
+    const std::string s(json_, length_);
+    for (size_t i = 0; i < kTrialCount; i++) {
+        Document doc;
+        doc.Parse(s);
+        ASSERT_TRUE(doc.IsObject());
+    }
+}
+#endif
+
 TEST_F(RapidJson, SIMD_SUFFIX(DocumentParseIterative_MemoryPoolAllocator)) {
     for (size_t i = 0; i < kTrialCount; i++) {
         Document doc;

test/unittest/documenttest.cpp

@@ -34,6 +34,8 @@ void ParseCheck(DocumentType& doc) {
     typedef typename DocumentType::ValueType ValueType;
 
     EXPECT_FALSE(doc.HasParseError());
+    if (doc.HasParseError())
+        printf("Error: %d at %zu\n", static_cast<int>(doc.GetParseError()), doc.GetErrorOffset());
     EXPECT_TRUE(static_cast<ParseResult>(doc));
 
     EXPECT_TRUE(doc.IsObject());
@@ -93,6 +95,26 @@ void ParseTest() {
     doc.ParseInsitu(buffer);
     ParseCheck(doc);
     free(buffer);
+
+    // Parse(const Ch*, size_t)
+    size_t length = strlen(json);
+    buffer = reinterpret_cast<char*>(malloc(length * 2));
+    memcpy(buffer, json, length);
+    memset(buffer + length, 'X', length);
+#if RAPIDJSON_HAS_STDSTRING
+    std::string s2(buffer, length); // backup buffer
+#endif
+    doc.SetNull();
+    doc.Parse(buffer, length);
+    free(buffer);
+    ParseCheck(doc);
+
+#if RAPIDJSON_HAS_STDSTRING
+    // Parse(std::string)
+    doc.SetNull();
+    doc.Parse(s2);
+    ParseCheck(doc);
+#endif
 }
 
 TEST(Document, Parse) {
@@ -140,6 +162,42 @@ static FILE* OpenEncodedFile(const char* filename) {
     return 0;
 }
 
+TEST(Document, Parse_Encoding) {
+    const char* json = " { \"hello\" : \"world\", \"t\" : true , \"f\" : false, \"n\": null, \"i\":123, \"pi\": 3.1416, \"a\":[1, 2, 3, 4] } ";
+
+    typedef GenericDocument<UTF16<> > DocumentType;
+    DocumentType doc;
+    
+    // Parse<unsigned, SourceEncoding>(const SourceEncoding::Ch*)
+    // doc.Parse<kParseDefaultFlags, UTF8<> >(json);
+    // EXPECT_FALSE(doc.HasParseError());
+    // EXPECT_EQ(0, StrCmp(doc[L"hello"].GetString(), L"world"));
+
+    // Parse<unsigned, SourceEncoding>(const SourceEncoding::Ch*, size_t)
+    size_t length = strlen(json);
+    char* buffer = reinterpret_cast<char*>(malloc(length * 2));
+    memcpy(buffer, json, length);
+    memset(buffer + length, 'X', length);
+#if RAPIDJSON_HAS_STDSTRING
+    std::string s2(buffer, length); // backup buffer
+#endif
+    doc.SetNull();
+    doc.Parse<kParseDefaultFlags, UTF8<> >(buffer, length);
+    free(buffer);
+    EXPECT_FALSE(doc.HasParseError());
+    if (doc.HasParseError())
+        printf("Error: %d at %zu\n", static_cast<int>(doc.GetParseError()), doc.GetErrorOffset());
+    EXPECT_EQ(0, StrCmp(doc[L"hello"].GetString(), L"world"));
+
+#if RAPIDJSON_HAS_STDSTRING
+    // Parse<unsigned, SourceEncoding>(std::string)
+    doc.SetNull();
+    doc.Parse<kParseDefaultFlags, UTF8<> >(s2);
+    EXPECT_FALSE(doc.HasParseError());
+    EXPECT_EQ(0, StrCmp(doc[L"hello"].GetString(), L"world"));
+#endif
+}
+
 TEST(Document, ParseStream_EncodedInputStream) {
     // UTF8 -> UTF16
     FILE* fp = OpenEncodedFile("utf8.json");

test/unittest/simdtest.cpp

@@ -73,6 +73,28 @@ TEST(SIMD, SIMD_SUFFIX(SkipWhitespace)) {
     TestSkipWhitespace<InsituStringStream>();
 }
 
+TEST(SIMD, SIMD_SUFFIX(SkipWhitespace_EncodedMemoryStream)) {
+    for (size_t step = 1; step < 32; step++) {
+        char buffer[1024];
+        for (size_t i = 0; i < 1024; i++)
+            buffer[i] = " \t\r\n"[i % 4];
+        for (size_t i = 0; i < 1024; i += step)
+            buffer[i] = 'X';
+
+        MemoryStream ms(buffer, 1024);
+        EncodedInputStream<UTF8<>, MemoryStream> s(ms);
+        size_t i = 0;
+        for (;;) {
+            SkipWhitespace(s);
+            if (s.Peek() == '\0')
+                break;
+            //EXPECT_EQ(i, s.Tell());
+            EXPECT_EQ('X', s.Take());
+            i += step;
+        }
+    }
+}
+
 struct ScanCopyUnescapedStringHandler : BaseReaderHandler<UTF8<>, ScanCopyUnescapedStringHandler> {
     bool String(const char* str, size_t length, bool) {
         memcpy(buffer, str, length + 1);