diff options
author | edisonn@google.com <edisonn@google.com@2bbb7eff-a529-9590-31e7-b0007b416f81> | 2013-07-23 17:43:18 +0000 |
---|---|---|
committer | edisonn@google.com <edisonn@google.com@2bbb7eff-a529-9590-31e7-b0007b416f81> | 2013-07-23 17:43:18 +0000 |
commit | 2ccc3afa474f9485c39c2e863252ddaa3f35724b (patch) | |
tree | a0656156a32a57c109e6dc0ed8063b41b24ddcdc /experimental | |
parent | 0ab790bef3f4cc018015fce58b12265b5511afad (diff) |
pdfviewer: improve memory usage, improve parse time (by 30-50%) and don't allocate extra buffers (more to do, but low priority now), and put the page specific memory in an allocator.
Review URL: https://codereview.chromium.org/19793011
git-svn-id: http://skia.googlecode.com/svn/trunk@10282 2bbb7eff-a529-9590-31e7-b0007b416f81
Diffstat (limited to 'experimental')
-rw-r--r-- | experimental/PdfViewer/SkPdfBasics.cpp | 12 | ||||
-rw-r--r-- | experimental/PdfViewer/SkPdfBasics.h | 13 | ||||
-rw-r--r-- | experimental/PdfViewer/SkPdfConfig.h | 21 | ||||
-rw-r--r-- | experimental/PdfViewer/SkPdfFont.cpp | 8 | ||||
-rw-r--r-- | experimental/PdfViewer/SkPdfRenderer.cpp | 274 | ||||
-rw-r--r-- | experimental/PdfViewer/pdf_viewer_main.cpp | 1 | ||||
-rw-r--r-- | experimental/PdfViewer/pdfparser/native/SkNativeParsedPDF.cpp | 82 | ||||
-rw-r--r-- | experimental/PdfViewer/pdfparser/native/SkNativeParsedPDF.h | 17 | ||||
-rw-r--r-- | experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.cpp | 312 | ||||
-rw-r--r-- | experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.h | 16 | ||||
-rw-r--r-- | experimental/PdfViewer/pdfparser/native/SkPdfObject.cpp | 39 | ||||
-rw-r--r-- | experimental/PdfViewer/pdfparser/native/SkPdfObject.h | 99 |
12 files changed, 558 insertions, 336 deletions
diff --git a/experimental/PdfViewer/SkPdfBasics.cpp b/experimental/PdfViewer/SkPdfBasics.cpp index e69de29bb2..e345f48a45 100644 --- a/experimental/PdfViewer/SkPdfBasics.cpp +++ b/experimental/PdfViewer/SkPdfBasics.cpp @@ -0,0 +1,12 @@ +#include "SkPdfBasics.h" +#include "SkPdfNativeTokenizer.h" + +PdfContext::PdfContext(SkNativeParsedPDF* doc) + : fPdfDoc(doc) + , fTmpPageAllocator(new SkPdfAllocator()) { +} + +PdfContext::~PdfContext() { + delete fTmpPageAllocator; +} + diff --git a/experimental/PdfViewer/SkPdfBasics.h b/experimental/PdfViewer/SkPdfBasics.h index 747c32a3d4..31fc2cd071 100644 --- a/experimental/PdfViewer/SkPdfBasics.h +++ b/experimental/PdfViewer/SkPdfBasics.h @@ -16,11 +16,12 @@ class SkPdfObject; class SkPdfResourceDictionary; class SkNativeParsedPDF; +class SkPdfAllocator; // TODO(edisonn): better class design. struct SkPdfColorOperator { // does not own the char* - const char* fColorSpace; // TODO(edisonn): use SkString, or even char* + NotOwnedString fColorSpace; SkColor fColor; double fOpacity; // ca or CA // TODO(edisonn): add here other color space options. @@ -30,7 +31,9 @@ struct SkPdfColorOperator { fColor = color; } // TODO(edisonn): double check the default values for all fields. - SkPdfColorOperator() : fColorSpace(NULL), fColor(SK_ColorBLACK), fOpacity(1) {} + SkPdfColorOperator() : fColor(SK_ColorBLACK), fOpacity(1) { + NotOwnedString::init(&fColorSpace); + } void applyGraphicsState(SkPaint* paint) { paint->setColor(SkColorSetA(fColor, fOpacity * 255)); @@ -117,12 +120,14 @@ struct PdfContext { std::stack<SkPdfGraphicsState> fStateStack; SkPdfGraphicsState fGraphicsState; SkNativeParsedPDF* fPdfDoc; + // TODO(edisonn): the allocator, could be freed after the page is done drawing. + SkPdfAllocator* fTmpPageAllocator; SkMatrix fOriginalMatrix; SkPdfInlineImage fInlineImage; - PdfContext(SkNativeParsedPDF* doc) : fPdfDoc(doc) {} - + PdfContext(SkNativeParsedPDF* doc); + ~PdfContext(); }; // TODO(edisonn): temporary code, to report how much of the PDF we actually think we rendered. diff --git a/experimental/PdfViewer/SkPdfConfig.h b/experimental/PdfViewer/SkPdfConfig.h index 1092de17e2..3d24a86ff7 100644 --- a/experimental/PdfViewer/SkPdfConfig.h +++ b/experimental/PdfViewer/SkPdfConfig.h @@ -2,6 +2,7 @@ #define __DEFINED__SkPdfConfig //#define PDF_TRACE +//#define PDF_TRACE_READ_TOKEN //#define PDF_TRACE_DIFF_IN_PNG //#define PDF_DEBUG_NO_CLIPING //#define PDF_DEBUG_NO_PAGE_CLIPING @@ -16,4 +17,24 @@ void SkTraceRect(const SkRect& rect, const char* sz); #define SkTraceRect(a,b) #endif +struct NotOwnedString { + const unsigned char* fBuffer; + size_t fBytes; + + static void init(NotOwnedString* str) { + str->fBuffer = NULL; + str->fBytes = 0; + } + + static void init(NotOwnedString* str, const char* sz) { + str->fBuffer = (const unsigned char*)sz; + str->fBytes = strlen(sz); + } + + bool equals(const char* sz) { + return strncmp((const char*)fBuffer, sz, fBytes) == 0 && fBytes == strlen(sz); + + } +}; + #endif // __DEFINED__SkPdfConfig diff --git a/experimental/PdfViewer/SkPdfFont.cpp b/experimental/PdfViewer/SkPdfFont.cpp index 2574ce4165..0fec460bf5 100644 --- a/experimental/PdfViewer/SkPdfFont.cpp +++ b/experimental/PdfViewer/SkPdfFont.cpp @@ -164,12 +164,12 @@ SkPdfFont* SkPdfFont::fontFromFontDescriptor(SkNativeParsedPDF* doc, SkPdfFontDe } } - unsigned char* uncompressedStream = NULL; + const unsigned char* uncompressedStream = NULL; size_t uncompressedStreamLength = 0; // TODO(edisonn): report warning to be used in testing. if (!pdfStream || - !pdfStream->GetFilteredStreamRef(&uncompressedStream, &uncompressedStreamLength, doc->allocator()) || + !pdfStream->GetFilteredStreamRef(&uncompressedStream, &uncompressedStreamLength) || !uncompressedStream || !uncompressedStreamLength) { return NULL; @@ -321,7 +321,9 @@ SkPdfToUnicode::SkPdfToUnicode(SkNativeParsedPDF* parsed, SkPdfStream* stream) : fCMapEncodingFlag = NULL; if (stream) { - SkPdfNativeTokenizer* tokenizer = fParsed->tokenizerOfStream(stream); + // Since font will be cached, the font has to sit in the per doc allocator, not to be + // freed after the page is done drawing. + SkPdfNativeTokenizer* tokenizer = fParsed->tokenizerOfStream(stream, parsed->allocator()); PdfToken token; fCMapEncoding = new unsigned short[256 * 256]; diff --git a/experimental/PdfViewer/SkPdfRenderer.cpp b/experimental/PdfViewer/SkPdfRenderer.cpp index 1ace7ca5ae..6e86fcb6a5 100644 --- a/experimental/PdfViewer/SkPdfRenderer.cpp +++ b/experimental/PdfViewer/SkPdfRenderer.cpp @@ -16,11 +16,10 @@ #include "SkStream.h" #include "SkTypeface.h" #include "SkTArray.h" +#include "SkTDict.h" #include "SkPdfBasics.h" #include "SkPdfNativeTokenizer.h" - -#include <iostream> #include <cstdio> #include <stack> #include <set> @@ -78,7 +77,18 @@ __SK_FORCE_IMAGE_DECODER_LINKING; using namespace std; +NotOwnedString strings_DeviceRGB; +NotOwnedString strings_DeviceCMYK; +class StringsInit { +public: + StringsInit() { + NotOwnedString::init(&strings_DeviceRGB, "DeviceRGB"); + NotOwnedString::init(&strings_DeviceCMYK, "DeviceCMYK"); + } +}; + +StringsInit gStringsInit; // TODO(edisonn): Document PdfTokenLooper and subclasses. class PdfTokenLooper { @@ -150,16 +160,16 @@ static void setup_bitmap(SkBitmap* bitmap, int width, int height, SkColor color } // TODO(edisonn): synonyms? DeviceRGB and RGB ... -static int GetColorSpaceComponents(const std::string& colorSpace) { - if (colorSpace == "DeviceCMYK") { +static int GetColorSpaceComponents(NotOwnedString& colorSpace) { + if (colorSpace.equals("DeviceCMYK")) { return 4; - } else if (colorSpace == "DeviceGray" || - colorSpace == "CalGray" || - colorSpace == "Indexed") { + } else if (colorSpace.equals("DeviceGray") || + colorSpace.equals("CalGray") || + colorSpace.equals("Indexed")) { return 1; - } else if (colorSpace == "DeviceRGB" || - colorSpace == "CalRGB" || - colorSpace == "Lab") { + } else if (colorSpace.equals("DeviceRGB") || + colorSpace.equals("CalRGB") || + colorSpace.equals("Lab")) { return 3; } else { return 0; @@ -317,9 +327,15 @@ static bool readToken(SkPdfNativeTokenizer* fTokenizer, PdfToken* token) { typedef PdfResult (*PdfOperatorRenderer)(PdfContext*, SkCanvas*, PdfTokenLooper**); -map<std::string, PdfOperatorRenderer> gPdfOps; +SkTDict<PdfOperatorRenderer> gPdfOps(100); -map<std::string, int> gRenderStats[kCount_PdfResult]; + +template <typename T> class SkTDictWithDefaultConstructor : public SkTDict<T> { +public: + SkTDictWithDefaultConstructor() : SkTDict<T>(10) {} +}; + +SkTDictWithDefaultConstructor<int> gRenderStats[kCount_PdfResult]; const char* gRenderStatsNames[kCount_PdfResult] = { "Success", @@ -419,7 +435,7 @@ static SkColorTable* getGrayColortable() { return grayColortable; } -static SkBitmap transferImageStreamToBitmap(unsigned char* uncompressedStream, size_t uncompressedStreamLength, +static SkBitmap transferImageStreamToBitmap(const unsigned char* uncompressedStream, size_t uncompressedStreamLength, int width, int height, int bytesPerLine, int bpc, const std::string& colorSpace, bool transparencyMask) { @@ -508,12 +524,12 @@ static SkBitmap getImageFromObject(PdfContext* pdfContext, SkPdfImageDictionary* } */ - unsigned char* uncompressedStream = NULL; + const unsigned char* uncompressedStream = NULL; size_t uncompressedStreamLength = 0; SkPdfStream* stream = (SkPdfStream*)image; - if (!stream || !stream->GetFilteredStreamRef(&uncompressedStream, &uncompressedStreamLength, pdfContext->fPdfDoc->allocator()) || + if (!stream || !stream->GetFilteredStreamRef(&uncompressedStream, &uncompressedStreamLength) || uncompressedStream == NULL || uncompressedStreamLength == 0) { // TODO(edisonn): report warning to be used in testing. return SkBitmap(); @@ -654,7 +670,8 @@ static PdfResult doXObject_Form(PdfContext* pdfContext, SkCanvas* canvas, SkPdfT SkPdfStream* stream = (SkPdfStream*)skobj; - SkPdfNativeTokenizer* tokenizer = pdfContext->fPdfDoc->tokenizerOfStream(stream); + SkPdfNativeTokenizer* tokenizer = + pdfContext->fPdfDoc->tokenizerOfStream(stream, pdfContext->fTmpPageAllocator); if (tokenizer != NULL) { PdfMainLooper looper(NULL, tokenizer, pdfContext, canvas); looper.loop(); @@ -702,7 +719,8 @@ PdfResult doType3Char(PdfContext* pdfContext, SkCanvas* canvas, const SkPdfObjec SkPdfStream* stream = (SkPdfStream*)skobj; - SkPdfNativeTokenizer* tokenizer = pdfContext->fPdfDoc->tokenizerOfStream(stream); + SkPdfNativeTokenizer* tokenizer = + pdfContext->fPdfDoc->tokenizerOfStream(stream, pdfContext->fTmpPageAllocator); if (tokenizer != NULL) { PdfMainLooper looper(NULL, tokenizer, pdfContext, canvas); looper.loop(); @@ -1151,10 +1169,10 @@ static PdfResult PdfOp_ET(PdfContext* pdfContext, SkCanvas* canvas, PdfTokenLoop //size; they must be specified explicitly using Tf before any text is shown. static PdfResult PdfOp_Tf(PdfContext* pdfContext, SkCanvas* canvas, PdfTokenLooper** looper) { pdfContext->fGraphicsState.fCurFontSize = pdfContext->fObjectStack.top()->numberValue(); pdfContext->fObjectStack.pop(); - const char* fontName = pdfContext->fObjectStack.top()->nameValue(); pdfContext->fObjectStack.pop(); + SkPdfObject* fontName = pdfContext->fObjectStack.top(); pdfContext->fObjectStack.pop(); #ifdef PDF_TRACE - printf("font name: %s\n", fontName); + printf("font name: %s\n", fontName->nameValue2().c_str()); #endif if (pdfContext->fGraphicsState.fResources->Font(pdfContext->fPdfDoc)) { @@ -1263,7 +1281,7 @@ static PdfResult PdfOp_TJ(PdfContext* pdfContext, SkCanvas* canvas, PdfTokenLoop } static PdfResult PdfOp_CS_cs(PdfContext* pdfContext, SkCanvas* canvas, SkPdfColorOperator* colorOperator) { - colorOperator->fColorSpace = pdfContext->fObjectStack.top()->nameValue(); pdfContext->fObjectStack.pop(); + colorOperator->fColorSpace = pdfContext->fObjectStack.top()->strRef(); pdfContext->fObjectStack.pop(); return kOK_PdfResult; } @@ -1282,12 +1300,12 @@ static PdfResult PdfOp_SC_sc(PdfContext* pdfContext, SkCanvas* canvas, SkPdfColo int n = GetColorSpaceComponents(colorOperator->fColorSpace); bool doubles = true; - if (strcmp(colorOperator->fColorSpace, "Indexed") == 0) { + if (colorOperator->fColorSpace.equals("Indexed")) { doubles = false; } #ifdef PDF_TRACE - printf("color space = %s, N = %i\n", colorOperator->fColorSpace, n); + printf("color space = %s, N = %i\n", colorOperator->fColorSpace.fBuffer, n); #endif for (int i = n - 1; i >= 0 ; i--) { @@ -1301,7 +1319,7 @@ static PdfResult PdfOp_SC_sc(PdfContext* pdfContext, SkCanvas* canvas, SkPdfColo // TODO(edisonn): Now, set that color. Only DeviceRGB supported. // TODO(edisonn): do possible field values to enum at parsing time! // TODO(edisonn): support also abreviations /DeviceRGB == /RGB - if (strcmp(colorOperator->fColorSpace, "DeviceRGB") == 0 || strcmp(colorOperator->fColorSpace, "RGB") == 0) { + if (colorOperator->fColorSpace.equals("DeviceRGB") || colorOperator->fColorSpace.equals("RGB")) { colorOperator->setRGBColor(SkColorSetRGB(255*c[0], 255*c[1], 255*c[2])); } return kPartial_PdfResult; @@ -1354,7 +1372,7 @@ static PdfResult PdfOp_RG_rg(PdfContext* pdfContext, SkCanvas* canvas, SkPdfColo double g = pdfContext->fObjectStack.top()->numberValue(); pdfContext->fObjectStack.pop(); double r = pdfContext->fObjectStack.top()->numberValue(); pdfContext->fObjectStack.pop(); - colorOperator->fColorSpace = "DeviceRGB"; + colorOperator->fColorSpace = strings_DeviceRGB; colorOperator->setRGBColor(SkColorSetRGB(255*r, 255*g, 255*b)); return kOK_PdfResult; } @@ -1374,7 +1392,7 @@ static PdfResult PdfOp_K_k(PdfContext* pdfContext, SkCanvas* canvas, SkPdfColorO /*double m = */pdfContext->fObjectStack.top()->numberValue(); pdfContext->fObjectStack.pop(); /*double c = */pdfContext->fObjectStack.top()->numberValue(); pdfContext->fObjectStack.pop(); - colorOperator->fColorSpace = "DeviceCMYK"; + colorOperator->fColorSpace = strings_DeviceCMYK; // TODO(edisonn): Set color. return kNYI_PdfResult; } @@ -1504,7 +1522,7 @@ static PdfResult PdfOp_i(PdfContext* pdfContext, SkCanvas* canvas, PdfTokenLoope //dictName gs (PDF 1.2) Set the specified parameters in the graphics state. dictName is //the name of a graphics state parameter dictionary in the ExtGState subdictionary of the current resource dictionary (see the next section). static PdfResult PdfOp_gs(PdfContext* pdfContext, SkCanvas* canvas, PdfTokenLooper** looper) { - const char* name = pdfContext->fObjectStack.top()->nameValue(); pdfContext->fObjectStack.pop(); + SkPdfObject* name = pdfContext->fObjectStack.top(); pdfContext->fObjectStack.pop(); #ifdef PDF_TRACE std::string str; @@ -1622,7 +1640,7 @@ static PdfResult PdfOp_sh(PdfContext* pdfContext, SkCanvas* canvas, PdfTokenLoop //name Do static PdfResult PdfOp_Do(PdfContext* pdfContext, SkCanvas* canvas, PdfTokenLooper** looper) { - const char* name = pdfContext->fObjectStack.top()->nameValue(); pdfContext->fObjectStack.pop(); + SkPdfObject* name = pdfContext->fObjectStack.top(); pdfContext->fObjectStack.pop(); SkPdfDictionary* xObject = pdfContext->fGraphicsState.fResources->XObject(pdfContext->fPdfDoc); @@ -1693,94 +1711,94 @@ static void initPdfOperatorRenderes() { return; } - gPdfOps["q"] = PdfOp_q; - gPdfOps["Q"] = PdfOp_Q; - gPdfOps["cm"] = PdfOp_cm; - - gPdfOps["TD"] = PdfOp_TD; - gPdfOps["Td"] = PdfOp_Td; - gPdfOps["Tm"] = PdfOp_Tm; - gPdfOps["T*"] = PdfOp_T_star; - - gPdfOps["m"] = PdfOp_m; - gPdfOps["l"] = PdfOp_l; - gPdfOps["c"] = PdfOp_c; - gPdfOps["v"] = PdfOp_v; - gPdfOps["y"] = PdfOp_y; - gPdfOps["h"] = PdfOp_h; - gPdfOps["re"] = PdfOp_re; - - gPdfOps["S"] = PdfOp_S; - gPdfOps["s"] = PdfOp_s; - gPdfOps["f"] = PdfOp_f; - gPdfOps["F"] = PdfOp_F; - gPdfOps["f*"] = PdfOp_f_star; - gPdfOps["B"] = PdfOp_B; - gPdfOps["B*"] = PdfOp_B_star; - gPdfOps["b"] = PdfOp_b; - gPdfOps["b*"] = PdfOp_b_star; - gPdfOps["n"] = PdfOp_n; - - gPdfOps["BT"] = PdfOp_BT; - gPdfOps["ET"] = PdfOp_ET; - - gPdfOps["Tj"] = PdfOp_Tj; - gPdfOps["'"] = PdfOp_quote; - gPdfOps["\""] = PdfOp_doublequote; - gPdfOps["TJ"] = PdfOp_TJ; - - gPdfOps["CS"] = PdfOp_CS; - gPdfOps["cs"] = PdfOp_cs; - gPdfOps["SC"] = PdfOp_SC; - gPdfOps["SCN"] = PdfOp_SCN; - gPdfOps["sc"] = PdfOp_sc; - gPdfOps["scn"] = PdfOp_scn; - gPdfOps["G"] = PdfOp_G; - gPdfOps["g"] = PdfOp_g; - gPdfOps["RG"] = PdfOp_RG; - gPdfOps["rg"] = PdfOp_rg; - gPdfOps["K"] = PdfOp_K; - gPdfOps["k"] = PdfOp_k; - - gPdfOps["W"] = PdfOp_W; - gPdfOps["W*"] = PdfOp_W_star; - - gPdfOps["BX"] = PdfOp_BX; - gPdfOps["EX"] = PdfOp_EX; - - gPdfOps["BI"] = PdfOp_BI; - gPdfOps["ID"] = PdfOp_ID; - gPdfOps["EI"] = PdfOp_EI; - - gPdfOps["w"] = PdfOp_w; - gPdfOps["J"] = PdfOp_J; - gPdfOps["j"] = PdfOp_j; - gPdfOps["M"] = PdfOp_M; - gPdfOps["d"] = PdfOp_d; - gPdfOps["ri"] = PdfOp_ri; - gPdfOps["i"] = PdfOp_i; - gPdfOps["gs"] = PdfOp_gs; - - gPdfOps["Tc"] = PdfOp_Tc; - gPdfOps["Tw"] = PdfOp_Tw; - gPdfOps["Tz"] = PdfOp_Tz; - gPdfOps["TL"] = PdfOp_TL; - gPdfOps["Tf"] = PdfOp_Tf; - gPdfOps["Tr"] = PdfOp_Tr; - gPdfOps["Ts"] = PdfOp_Ts; - - gPdfOps["d0"] = PdfOp_d0; - gPdfOps["d1"] = PdfOp_d1; - - gPdfOps["sh"] = PdfOp_sh; - - gPdfOps["Do"] = PdfOp_Do; - - gPdfOps["MP"] = PdfOp_MP; - gPdfOps["DP"] = PdfOp_DP; - gPdfOps["BMC"] = PdfOp_BMC; - gPdfOps["BDC"] = PdfOp_BDC; - gPdfOps["EMC"] = PdfOp_EMC; + gPdfOps.set("q", PdfOp_q); + gPdfOps.set("Q", PdfOp_Q); + gPdfOps.set("cm", PdfOp_cm); + + gPdfOps.set("TD", PdfOp_TD); + gPdfOps.set("Td", PdfOp_Td); + gPdfOps.set("Tm", PdfOp_Tm); + gPdfOps.set("T*", PdfOp_T_star); + + gPdfOps.set("m", PdfOp_m); + gPdfOps.set("l", PdfOp_l); + gPdfOps.set("c", PdfOp_c); + gPdfOps.set("v", PdfOp_v); + gPdfOps.set("y", PdfOp_y); + gPdfOps.set("h", PdfOp_h); + gPdfOps.set("re", PdfOp_re); + + gPdfOps.set("S", PdfOp_S); + gPdfOps.set("s", PdfOp_s); + gPdfOps.set("f", PdfOp_f); + gPdfOps.set("F", PdfOp_F); + gPdfOps.set("f*", PdfOp_f_star); + gPdfOps.set("B", PdfOp_B); + gPdfOps.set("B*", PdfOp_B_star); + gPdfOps.set("b", PdfOp_b); + gPdfOps.set("b*", PdfOp_b_star); + gPdfOps.set("n", PdfOp_n); + + gPdfOps.set("BT", PdfOp_BT); + gPdfOps.set("ET", PdfOp_ET); + + gPdfOps.set("Tj", PdfOp_Tj); + gPdfOps.set("'", PdfOp_quote); + gPdfOps.set("\"", PdfOp_doublequote); + gPdfOps.set("TJ", PdfOp_TJ); + + gPdfOps.set("CS", PdfOp_CS); + gPdfOps.set("cs", PdfOp_cs); + gPdfOps.set("SC", PdfOp_SC); + gPdfOps.set("SCN", PdfOp_SCN); + gPdfOps.set("sc", PdfOp_sc); + gPdfOps.set("scn", PdfOp_scn); + gPdfOps.set("G", PdfOp_G); + gPdfOps.set("g", PdfOp_g); + gPdfOps.set("RG", PdfOp_RG); + gPdfOps.set("rg", PdfOp_rg); + gPdfOps.set("K", PdfOp_K); + gPdfOps.set("k", PdfOp_k); + + gPdfOps.set("W", PdfOp_W); + gPdfOps.set("W*", PdfOp_W_star); + + gPdfOps.set("BX", PdfOp_BX); + gPdfOps.set("EX", PdfOp_EX); + + gPdfOps.set("BI", PdfOp_BI); + gPdfOps.set("ID", PdfOp_ID); + gPdfOps.set("EI", PdfOp_EI); + + gPdfOps.set("w", PdfOp_w); + gPdfOps.set("J", PdfOp_J); + gPdfOps.set("j", PdfOp_j); + gPdfOps.set("M", PdfOp_M); + gPdfOps.set("d", PdfOp_d); + gPdfOps.set("ri", PdfOp_ri); + gPdfOps.set("i", PdfOp_i); + gPdfOps.set("gs", PdfOp_gs); + + gPdfOps.set("Tc", PdfOp_Tc); + gPdfOps.set("Tw", PdfOp_Tw); + gPdfOps.set("Tz", PdfOp_Tz); + gPdfOps.set("TL", PdfOp_TL); + gPdfOps.set("Tf", PdfOp_Tf); + gPdfOps.set("Tr", PdfOp_Tr); + gPdfOps.set("Ts", PdfOp_Ts); + + gPdfOps.set("d0", PdfOp_d0); + gPdfOps.set("d1", PdfOp_d1); + + gPdfOps.set("sh", PdfOp_sh); + + gPdfOps.set("Do", PdfOp_Do); + + gPdfOps.set("MP", PdfOp_MP); + gPdfOps.set("DP", PdfOp_DP); + gPdfOps.set("BMC", PdfOp_BMC); + gPdfOps.set("BDC", PdfOp_BDC); + gPdfOps.set("EMC", PdfOp_EMC); gInitialized = true; } @@ -1798,8 +1816,11 @@ void reportPdfRenderStats() { std::map<std::string, int>::iterator iter; for (int i = 0 ; i < kCount_PdfResult; i++) { - for (iter = gRenderStats[i].begin(); iter != gRenderStats[i].end(); ++iter) { - printf("%s: %s -> count %i\n", gRenderStatsNames[i], iter->first.c_str(), iter->second); + SkTDict<int>::Iter iter(gRenderStats[i]); + const char* key; + int value = 0; + while ((key = iter.next(&value)) != NULL) { + printf("%s: %s -> count %i\n", gRenderStatsNames[i], key, value); } } } @@ -1812,11 +1833,15 @@ PdfResult PdfMainLooper::consumeToken(PdfToken& token) { strncpy(keyword, token.fKeyword, token.fKeywordLength); keyword[token.fKeywordLength] = '\0'; // TODO(edisonn): log trace flag (verbose, error, info, warning, ...) - PdfOperatorRenderer pdfOperatorRenderer = gPdfOps[keyword]; - if (pdfOperatorRenderer) { + PdfOperatorRenderer pdfOperatorRenderer = NULL; + if (gPdfOps.find(keyword, &pdfOperatorRenderer) && pdfOperatorRenderer) { // caller, main work is done by pdfOperatorRenderer(...) PdfTokenLooper* childLooper = NULL; - gRenderStats[pdfOperatorRenderer(fPdfContext, fCanvas, &childLooper)][keyword]++; + PdfResult result = pdfOperatorRenderer(fPdfContext, fCanvas, &childLooper); + + int cnt = 0; + gRenderStats[result].find(keyword, &cnt); + gRenderStats[result].set(keyword, cnt + 1); if (childLooper) { childLooper->setUp(this); @@ -1824,7 +1849,9 @@ PdfResult PdfMainLooper::consumeToken(PdfToken& token) { delete childLooper; } } else { - gRenderStats[kUnsupported_PdfResult][keyword]++; + int cnt = 0; + gRenderStats[kUnsupported_PdfResult].find(keyword, &cnt); + gRenderStats[kUnsupported_PdfResult].set(keyword, cnt + 1); } } else if (token.fType == kObject_TokenType) @@ -1919,9 +1946,10 @@ bool SkPdfRenderer::renderPage(int page, SkCanvas* canvas, const SkRect& dst) co return false; } - SkPdfNativeTokenizer* tokenizer = fPdfDoc->tokenizerOfPage(page); - PdfContext pdfContext(fPdfDoc); + + SkPdfNativeTokenizer* tokenizer = fPdfDoc->tokenizerOfPage(page, pdfContext.fTmpPageAllocator); + pdfContext.fOriginalMatrix = SkMatrix::I(); pdfContext.fGraphicsState.fResources = fPdfDoc->pageResources(page); diff --git a/experimental/PdfViewer/pdf_viewer_main.cpp b/experimental/PdfViewer/pdf_viewer_main.cpp index f08fa510fd..7541cdfae6 100644 --- a/experimental/PdfViewer/pdf_viewer_main.cpp +++ b/experimental/PdfViewer/pdf_viewer_main.cpp @@ -199,6 +199,7 @@ static bool process_pdf(const SkString& inputPath, const SkString& outputDir, SkDebugf("ERROR: Empty PDF Document %s\n", inputPath.c_str()); return false; } else { + printf("renders = %i\n", FLAGS_benchRender); for (int i = 0; i < FLAGS_benchRender + 1; i++) { // TODO(edisonn) if (i == 1) start timer if (strcmp(FLAGS_pages[0], "all") == 0) { diff --git a/experimental/PdfViewer/pdfparser/native/SkNativeParsedPDF.cpp b/experimental/PdfViewer/pdfparser/native/SkNativeParsedPDF.cpp index fe5d0ee830..ac543040b0 100644 --- a/experimental/PdfViewer/pdfparser/native/SkNativeParsedPDF.cpp +++ b/experimental/PdfViewer/pdfparser/native/SkNativeParsedPDF.cpp @@ -23,14 +23,14 @@ static long getFileSize(const char* filename) return rc == 0 ? (long)stat_buf.st_size : -1; } -static unsigned char* lineHome(unsigned char* start, unsigned char* current) { +static const unsigned char* lineHome(const unsigned char* start, const unsigned char* current) { while (current > start && !isPdfEOL(*(current - 1))) { current--; } return current; } -static unsigned char* previousLineHome(unsigned char* start, unsigned char* current) { +static const unsigned char* previousLineHome(const unsigned char* start, const unsigned char* current) { if (current > start && isPdfEOL(*(current - 1))) { current--; } @@ -47,7 +47,7 @@ static unsigned char* previousLineHome(unsigned char* start, unsigned char* curr return current; } -static unsigned char* ignoreLine(unsigned char* current, unsigned char* end) { +static const unsigned char* ignoreLine(const unsigned char* current, const unsigned char* end) { while (current < end && !isPdfEOL(*current)) { current++; } @@ -74,9 +74,10 @@ SkNativeParsedPDF::SkNativeParsedPDF(const char* path) gDoc = this; FILE* file = fopen(path, "r"); fContentLength = getFileSize(path); - fFileContent = new unsigned char[fContentLength + 1]; - bool ok = (0 != fread(fFileContent, fContentLength, 1, file)); - fFileContent[fContentLength] = '\0'; + unsigned char* content = new unsigned char[fContentLength + 1]; + bool ok = (0 != fread(content, fContentLength, 1, file)); + content[fContentLength] = '\0'; + fFileContent = content; fclose(file); file = NULL; @@ -85,9 +86,9 @@ SkNativeParsedPDF::SkNativeParsedPDF(const char* path) return; // Doc will have 0 pages } - unsigned char* eofLine = lineHome(fFileContent, fFileContent + fContentLength - 1); - unsigned char* xrefByteOffsetLine = previousLineHome(fFileContent, eofLine); - unsigned char* xrefstartKeywordLine = previousLineHome(fFileContent, xrefByteOffsetLine); + const unsigned char* eofLine = lineHome(fFileContent, fFileContent + fContentLength - 1); + const unsigned char* xrefByteOffsetLine = previousLineHome(fFileContent, eofLine); + const unsigned char* xrefstartKeywordLine = previousLineHome(fFileContent, xrefByteOffsetLine); if (strcmp((char*)xrefstartKeywordLine, "startxref") != 0) { // TODO(edisonn): report/issue @@ -97,7 +98,7 @@ SkNativeParsedPDF::SkNativeParsedPDF(const char* path) bool storeCatalog = true; while (xrefByteOffset >= 0) { - unsigned char* trailerStart = readCrossReferenceSection(fFileContent + xrefByteOffset, xrefstartKeywordLine); + const unsigned char* trailerStart = readCrossReferenceSection(fFileContent + xrefByteOffset, xrefstartKeywordLine); xrefByteOffset = readTrailer(trailerStart, xrefstartKeywordLine, storeCatalog); storeCatalog = false; } @@ -129,21 +130,21 @@ SkNativeParsedPDF::~SkNativeParsedPDF() { delete fAllocator; } -unsigned char* SkNativeParsedPDF::readCrossReferenceSection(unsigned char* xrefStart, unsigned char* trailerEnd) { - unsigned char* current = ignoreLine(xrefStart, trailerEnd); // TODO(edisonn): verify next keyord is "xref", use nextObject here +const unsigned char* SkNativeParsedPDF::readCrossReferenceSection(const unsigned char* xrefStart, const unsigned char* trailerEnd) { + const unsigned char* current = ignoreLine(xrefStart, trailerEnd); // TODO(edisonn): verify next keyord is "xref", use nextObject here SkPdfObject token; while (current < trailerEnd) { token.reset(); - unsigned char* previous = current; - current = nextObject(current, trailerEnd, &token, NULL, NULL); + const unsigned char* previous = current; + current = nextObject(0, current, trailerEnd, &token, NULL, NULL); if (!token.isInteger()) { return previous; } int startId = (int)token.intValue(); token.reset(); - current = nextObject(current, trailerEnd, &token, NULL, NULL); + current = nextObject(0, current, trailerEnd, &token, NULL, NULL); if (!token.isInteger()) { // TODO(edisonn): report/warning @@ -154,7 +155,7 @@ unsigned char* SkNativeParsedPDF::readCrossReferenceSection(unsigned char* xrefS for (int i = 0; i < entries; i++) { token.reset(); - current = nextObject(current, trailerEnd, &token, NULL, NULL); + current = nextObject(0, current, trailerEnd, &token, NULL, NULL); if (!token.isInteger()) { // TODO(edisonn): report/warning return current; @@ -162,7 +163,7 @@ unsigned char* SkNativeParsedPDF::readCrossReferenceSection(unsigned char* xrefS int offset = (int)token.intValue(); token.reset(); - current = nextObject(current, trailerEnd, &token, NULL, NULL); + current = nextObject(0, current, trailerEnd, &token, NULL, NULL); if (!token.isInteger()) { // TODO(edisonn): report/warning return current; @@ -170,7 +171,7 @@ unsigned char* SkNativeParsedPDF::readCrossReferenceSection(unsigned char* xrefS int generation = (int)token.intValue(); token.reset(); - current = nextObject(current, trailerEnd, &token, NULL, NULL); + current = nextObject(0, current, trailerEnd, &token, NULL, NULL); if (!token.isKeyword() || token.len() != 1 || (*token.c_str() != 'f' && *token.c_str() != 'n')) { // TODO(edisonn): report/warning return current; @@ -183,11 +184,21 @@ unsigned char* SkNativeParsedPDF::readCrossReferenceSection(unsigned char* xrefS return current; } -long SkNativeParsedPDF::readTrailer(unsigned char* trailerStart, unsigned char* trailerEnd, bool storeCatalog) { - unsigned char* current = ignoreLine(trailerStart, trailerEnd); // TODO(edisonn): verify next keyord is "trailer" use nextObject here +long SkNativeParsedPDF::readTrailer(const unsigned char* trailerStart, const unsigned char* trailerEnd, bool storeCatalog) { + SkPdfObject trailerKeyword; + // TODO(edisonn): use null allocator, and let it just fail if memory + // needs allocated (but no crash)! + const unsigned char* current = + nextObject(0, trailerStart, trailerEnd, &trailerKeyword, fAllocator, NULL); + + if (strlen("trailer") != trailerKeyword.len() && + strncmp(trailerKeyword.c_str(), "trailer", strlen("trailer")) != 0) { + // TODO(edisonn): report warning, rebuild trailer from objects. + return -1; + } SkPdfObject token; - current = nextObject(current, trailerEnd, &token, fAllocator, NULL); + current = nextObject(0, current, trailerEnd, &token, fAllocator, NULL); if (!token.isDictionary()) { return -1; } @@ -230,8 +241,8 @@ SkPdfObject* SkNativeParsedPDF::readObject(int id/*, int expectedGeneration*/) { // to decrease memory usage, we wither need to be smart and know where objects end, and we will // alocate only the chancks needed, or the tokenizer will not make copies, but then it needs to // cache the results so it does not go twice on the same buffer - unsigned char* current = fFileContent + startOffset; - unsigned char* end = fFileContent + fContentLength; + const unsigned char* current = fFileContent + startOffset; + const unsigned char* end = fFileContent + fContentLength; SkPdfNativeTokenizer tokenizer(current, end - current, fMapper, fAllocator, this); @@ -240,19 +251,19 @@ SkPdfObject* SkNativeParsedPDF::readObject(int id/*, int expectedGeneration*/) { SkPdfObject objKeyword; SkPdfObject* dict = fAllocator->allocObject(); - current = nextObject(current, end, &idObj, NULL, NULL); + current = nextObject(0, current, end, &idObj, NULL, NULL); if (current >= end) { // TODO(edisonn): report warning/error return NULL; } - current = nextObject(current, end, &generationObj, NULL, NULL); + current = nextObject(0, current, end, &generationObj, NULL, NULL); if (current >= end) { // TODO(edisonn): report warning/error return NULL; } - current = nextObject(current, end, &objKeyword, NULL, NULL); + current = nextObject(0, current, end, &objKeyword, NULL, NULL); if (current >= end) { // TODO(edisonn): report warning/error return NULL; @@ -266,7 +277,7 @@ SkPdfObject* SkNativeParsedPDF::readObject(int id/*, int expectedGeneration*/) { // TODO(edisonn): report warning/error } - current = nextObject(current, end, dict, fAllocator, this); + current = nextObject(1, current, end, dict, fAllocator, this); // TODO(edisonn): report warning/error - verify last token is endobj @@ -313,9 +324,10 @@ SkRect SkNativeParsedPDF::MediaBox(int page) { } // TODO(edisonn): stream or array ... ? for now only array -SkPdfNativeTokenizer* SkNativeParsedPDF::tokenizerOfPage(int page) { +SkPdfNativeTokenizer* SkNativeParsedPDF::tokenizerOfPage(int page, + SkPdfAllocator* allocator) { if (fPages[page]->isContentsAStream(this)) { - return tokenizerOfStream(fPages[page]->getContentsAsStream(this)); + return tokenizerOfStream(fPages[page]->getContentsAsStream(this), allocator); } else { // TODO(edisonn): NYI, we need to concatenate all streams in the array or make the tokenizer smart // so we don't allocate new memory @@ -323,19 +335,21 @@ SkPdfNativeTokenizer* SkNativeParsedPDF::tokenizerOfPage(int page) { } } -SkPdfNativeTokenizer* SkNativeParsedPDF::tokenizerOfStream(SkPdfObject* stream) { +SkPdfNativeTokenizer* SkNativeParsedPDF::tokenizerOfStream(SkPdfObject* stream, + SkPdfAllocator* allocator) { if (stream == NULL) { return NULL; } - return new SkPdfNativeTokenizer(stream, fMapper, fAllocator, this); + return new SkPdfNativeTokenizer(stream, fMapper, allocator, this); } // TODO(edisonn): NYI -SkPdfNativeTokenizer* SkNativeParsedPDF::tokenizerOfBuffer(unsigned char* buffer, size_t len) { +SkPdfNativeTokenizer* SkNativeParsedPDF::tokenizerOfBuffer(const unsigned char* buffer, size_t len, + SkPdfAllocator* allocator) { // warning does not track two calls in the same buffer! the buffer is updated! // make a clean copy if needed! - return new SkPdfNativeTokenizer(buffer, len, fMapper, fAllocator, this); + return new SkPdfNativeTokenizer(buffer, len, fMapper, allocator, this); } size_t SkNativeParsedPDF::objects() const { @@ -374,7 +388,7 @@ SkPdfInteger* SkNativeParsedPDF::createInteger(int value) const { return (SkPdfInteger*)obj; } -SkPdfString* SkNativeParsedPDF::createString(unsigned char* sz, size_t len) const { +SkPdfString* SkNativeParsedPDF::createString(const unsigned char* sz, size_t len) const { SkPdfObject* obj = fAllocator->allocObject(); SkPdfObject::makeString(sz, len, obj); return (SkPdfString*)obj; diff --git a/experimental/PdfViewer/pdfparser/native/SkNativeParsedPDF.h b/experimental/PdfViewer/pdfparser/native/SkNativeParsedPDF.h index b91d574f26..c520b0394a 100644 --- a/experimental/PdfViewer/pdfparser/native/SkNativeParsedPDF.h +++ b/experimental/PdfViewer/pdfparser/native/SkNativeParsedPDF.h @@ -41,10 +41,11 @@ public: int pages() const; SkPdfResourceDictionary* pageResources(int page); SkRect MediaBox(int page); - SkPdfNativeTokenizer* tokenizerOfPage(int n); + SkPdfNativeTokenizer* tokenizerOfPage(int n, SkPdfAllocator* allocator); - SkPdfNativeTokenizer* tokenizerOfStream(SkPdfObject* stream); - SkPdfNativeTokenizer* tokenizerOfBuffer(unsigned char* buffer, size_t len); + SkPdfNativeTokenizer* tokenizerOfStream(SkPdfObject* stream, SkPdfAllocator* allocator); + SkPdfNativeTokenizer* tokenizerOfBuffer(const unsigned char* buffer, size_t len, + SkPdfAllocator* allocator); size_t objects() const; SkPdfObject* object(int i); @@ -55,9 +56,7 @@ public: SkPdfReal* createReal(double value) const; SkPdfInteger* createInteger(int value) const; // the string does not own the char* - SkPdfString* createString(unsigned char* sz, size_t len) const; - - void drawPage(int page, SkCanvas* canvas); + SkPdfString* createString(const unsigned char* sz, size_t len) const; SkPdfObject* resolveReference(const SkPdfObject* ref); @@ -66,8 +65,8 @@ public: private: - unsigned char* readCrossReferenceSection(unsigned char* xrefStart, unsigned char* trailerEnd); - long readTrailer(unsigned char* trailerStart, unsigned char* trailerEnd, bool storeCatalog); + const unsigned char* readCrossReferenceSection(const unsigned char* xrefStart, const unsigned char* trailerEnd); + long readTrailer(const unsigned char* trailerStart, const unsigned char* trailerEnd, bool storeCatalog); // TODO(edisonn): updates not supported right now, generation ignored void addCrossSectionInfo(int id, int generation, int offset, bool isFreed); @@ -84,7 +83,7 @@ private: // private fields SkPdfAllocator* fAllocator; SkPdfMapper* fMapper; - unsigned char* fFileContent; + const unsigned char* fFileContent; size_t fContentLength; const SkPdfObject* fRootCatalogRef; SkPdfCatalogDictionary* fRootCatalog; diff --git a/experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.cpp b/experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.cpp index babfedfb05..4798031808 100644 --- a/experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.cpp +++ b/experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.cpp @@ -28,17 +28,74 @@ static char* strrstrk(char* hayStart, char* hayEnd, const char* needle) { return NULL; } +#ifdef PDF_TRACE +static void TRACE_INDENT(int level, const char* type) { + static int id = 0; + id++; + if (478613 == id) { + printf("break;\n"); + } + // all types should have 2 letters, so the text is alligned nicely + printf("\n%10i %15s: ", id, type); + for (int i = 0 ; i < level; i++) { + printf(" "); + } +} + +static void TRACE_COMMENT(char ch) { + printf("%c", ch); +} + +static void TRACE_TK(char ch) { + printf("%c", ch); +} + +static void TRACE_NAME(const unsigned char* start, const unsigned char* end) { + while (start < end) { + printf("%c", *start); + start++; + } + printf("\n"); +} -static unsigned char* skipPdfWhiteSpaces(unsigned char* start, unsigned char* end) { +static void TRACE_STRING(const unsigned char* start, const unsigned char* end) { + while (start < end) { + printf("%c", *start); + start++; + } + printf("\n"); +} + +static void TRACE_HEXSTRING(const unsigned char* start, const unsigned char* end) { + while (start < end) { + printf("%c", *start); + start++; + } + printf("\n"); +} + +#else +#define TRACE_INDENT(level,type) +#define TRACE_COMMENT(ch) +#define TRACE_TK(ch) +#define TRACE_NAME(start,end) +#define TRACE_STRING(start,end) +#define TRACE_HEXSTRING(start,end) +#endif + +static const unsigned char* skipPdfWhiteSpaces(int level, const unsigned char* start, const unsigned char* end) { + TRACE_INDENT(level, "White Space"); while (start < end && isPdfWhiteSpace(*start)) { + TRACE_COMMENT(*start); if (*start == kComment_PdfDelimiter) { // skip the comment until end of line while (start < end && !isPdfEOL(*start)) { - *start = '\0'; + //*start = '\0'; start++; + TRACE_COMMENT(*start); } } else { - *start = '\0'; + //*start = '\0'; start++; } } @@ -46,30 +103,34 @@ static unsigned char* skipPdfWhiteSpaces(unsigned char* start, unsigned char* en } // TODO(edisonn) '(' can be used, will it break the string a delimiter or space inside () ? -static unsigned char* endOfPdfToken(unsigned char* start, unsigned char* end) { +static const unsigned char* endOfPdfToken(int level, const unsigned char* start, const unsigned char* end) { //int opened brackets //TODO(edisonn): what out for special chars, like \n, \032 + TRACE_INDENT(level, "Token"); SkASSERT(!isPdfWhiteSpace(*start)); if (start < end && isPdfDelimiter(*start)) { + TRACE_TK(*start); start++; return start; } while (start < end && !isPdfWhiteSpaceOrPdfDelimiter(*start)) { + TRACE_TK(*start); start++; } return start; } // last elem has to be ] -static unsigned char* readArray(unsigned char* start, unsigned char* end, SkPdfObject* array, SkPdfAllocator* allocator, SkNativeParsedPDF* doc) { +static const unsigned char* readArray(int level, const unsigned char* start, const unsigned char* end, SkPdfObject* array, SkPdfAllocator* allocator, SkNativeParsedPDF* doc) { + TRACE_INDENT(level, "Array"); while (start < end) { // skip white spaces - start = skipPdfWhiteSpaces(start, end); + start = skipPdfWhiteSpaces(level + 1, start, end); - unsigned char* endOfToken = endOfPdfToken(start, end); + const unsigned char* endOfToken = endOfPdfToken(level + 1, start, end); if (endOfToken == start) { // TODO(edisonn): report error in pdf file (end of stream with ] for end of aray @@ -81,7 +142,7 @@ static unsigned char* readArray(unsigned char* start, unsigned char* end, SkPdfO } SkPdfObject* newObj = allocator->allocObject(); - start = nextObject(start, end, newObj, allocator, doc); + start = nextObject(level + 1, start, end, newObj, allocator, doc); // TODO(edisonn): perf/memory: put the variables on the stack, and flush them on the array only when // we are sure they are not references! if (newObj->isKeywordReference() && array->size() >= 2 && array->objAtAIndex(array->size() - 1)->isInteger() && array->objAtAIndex(array->size() - 2)->isInteger()) { @@ -101,38 +162,47 @@ static unsigned char* readArray(unsigned char* start, unsigned char* end, SkPdfO // When we read strings we will rewrite the string so we will reuse the memory // when we start to read the string, we already consumed the opened bracket -static unsigned char* readString(unsigned char* start, unsigned char* end, SkPdfObject* str) { - unsigned char* out = start; - unsigned char* in = start; - int openRoundBrackets = 0; - while (in < end && (*in != kClosedRoundBracket_PdfDelimiter || openRoundBrackets > 0)) { +// TODO(edisonn): space: add paramater, taht would report if we need to allocate new buffer, or we can reuse the one we have + +static const unsigned char* readString(int level, const unsigned char* start, const unsigned char* end, unsigned char* out) { + TRACE_INDENT(level, "String"); + const unsigned char* in = start; + bool hasOut = (out != NULL); + + int openRoundBrackets = 1; + while (in < end) { openRoundBrackets += ((*in) == kOpenedRoundBracket_PdfDelimiter); openRoundBrackets -= ((*in) == kClosedRoundBracket_PdfDelimiter); + if (openRoundBrackets == 0) { + in++; // consumed ) + break; + } + if (*in == kEscape_PdfSpecial) { if (in + 1 < end) { switch (in[1]) { case 'n': - *out = kLF_PdfWhiteSpace; + if (hasOut) { *out = kLF_PdfWhiteSpace; } out++; in += 2; break; case 'r': - *out = kCR_PdfWhiteSpace; + if (hasOut) { *out = kCR_PdfWhiteSpace; } out++; in += 2; break; case 't': - *out = kHT_PdfWhiteSpace; + if (hasOut) { *out = kHT_PdfWhiteSpace; } out++; in += 2; break; case 'b': // TODO(edisonn): any special meaning to backspace? - *out = kBackspace_PdfSpecial; + if (hasOut) { *out = kBackspace_PdfSpecial; } out++; in += 2; break; @@ -144,19 +214,19 @@ static unsigned char* readString(unsigned char* start, unsigned char* end, SkPdf break; case kOpenedRoundBracket_PdfDelimiter: - *out = kOpenedRoundBracket_PdfDelimiter; + if (hasOut) { *out = kOpenedRoundBracket_PdfDelimiter; } out++; in += 2; break; case kClosedRoundBracket_PdfDelimiter: - *out = kClosedRoundBracket_PdfDelimiter; + if (hasOut) { *out = kClosedRoundBracket_PdfDelimiter; } out++; in += 2; break; case kEscape_PdfSpecial: - *out = kEscape_PdfSpecial; + if (hasOut) { *out = kEscape_PdfSpecial; } out++; in += 2; break; @@ -179,13 +249,13 @@ static unsigned char* readString(unsigned char* start, unsigned char* end, SkPdf i++; in++; if (i == 3) { - *out = code & 0xff; + if (hasOut) { *out = code & 0xff; } out++; i = 0; } } if (i > 0) { - *out = code & 0xff; + if (hasOut) { *out = code & 0xff; } out++; } } @@ -203,20 +273,37 @@ static unsigned char* readString(unsigned char* start, unsigned char* end, SkPdf // TODO(edisonn): perf, avoid copy into itself, maybe first do a simple scan until found backslash ? // we could have one look that first just inc current, and when we find the backslash // we go to this loop - *in = *out; + if (hasOut) { *out = *in; } in++; out++; } } + if (hasOut) { + return in; // consumed already ) at the end of the string + } else { + return start + (out - (const unsigned char*)NULL); // return where the string would end if we reuse the string + } +} + +static int readStringLength(int level, const unsigned char* start, const unsigned char* end) { + return readString(level, start, end, NULL) - start; +} - SkPdfObject::makeString(start, out, str); - return in + 1; // consume ) at the end of the string +static const unsigned char* readString(int level, const unsigned char* start, const unsigned char* end, SkPdfObject* str, SkPdfAllocator* allocator) { + int outLength = readStringLength(level, start, end); + // TODO(edisonn): optimize the allocation, don't allocate new string, but put it in a preallocated buffer + unsigned char* out = (unsigned char*)allocator->alloc(outLength); + start = readString(level, start, end, out); + SkPdfObject::makeString(out, out + outLength, str); + TRACE_STRING(out, out + outLength); + return start; // consumed already ) at the end of the string } -static unsigned char* readHexString(unsigned char* start, unsigned char* end, SkPdfObject* str) { - unsigned char* out = start; - unsigned char* in = start; +static const unsigned char* readHexString(int level, const unsigned char* start, const unsigned char* end, unsigned char* out) { + TRACE_INDENT(level, "HexString"); + bool hasOut = (out != NULL); + const unsigned char* in = start; unsigned char code = 0; @@ -226,8 +313,8 @@ static unsigned char* readHexString(unsigned char* start, unsigned char* end, Sk } if (*in == kClosedInequityBracket_PdfDelimiter) { - *in = '\0'; - in++; + //*in = '\0'; + in++; // consume > // normal exit break; } @@ -282,13 +369,13 @@ static unsigned char* readHexString(unsigned char* start, unsigned char* end, Sk // TODO(edisonn): report error if (in >= end) { - *out = code; + if (hasOut) { *out = code; } out++; break; } if (*in == kClosedInequityBracket_PdfDelimiter) { - *out = code; + if (hasOut) { *out = code; } out++; break; } @@ -330,23 +417,37 @@ static unsigned char* readHexString(unsigned char* start, unsigned char* end, Sk break; } - *out = code; + if (hasOut) { *out = code; } out++; in++; } - if (out < in) { - *out = '\0'; + if (hasOut) { + return in; // consumed already > at the end of the string + } else { + return start + (out - (const unsigned char*)NULL); // return where the string would end if we reuse the string } +} - SkPdfObject::makeHexString(start, out, str); - return in; // consume > at the end of the string +static int readHexStringLength(int level, const unsigned char* start, const unsigned char* end) { + return readHexString(level, start, end, NULL) - start; +} + +static const unsigned char* readHexString(int level, const unsigned char* start, const unsigned char* end, SkPdfObject* str, SkPdfAllocator* allocator) { + int outLength = readHexStringLength(level, start, end); + // TODO(edisonn): optimize the allocation, don't allocate new string, but put it in a preallocated buffer + unsigned char* out = (unsigned char*)allocator->alloc(outLength); + start = readHexString(level, start, end, out); + SkPdfObject::makeHexString(out, out + outLength, str); + TRACE_HEXSTRING(out, out + outLength); + return start; // consumed already > at the end of the string } // TODO(edisonn): before PDF 1.2 name could not have special characters, add version parameter -static unsigned char* readName(unsigned char* start, unsigned char* end, SkPdfObject* name) { - unsigned char* out = start; - unsigned char* in = start; +static const unsigned char* readName(int level, const unsigned char* start, const unsigned char* end, unsigned char* out) { + TRACE_INDENT(level, "Name"); + bool hasOut = (out != NULL); + const unsigned char* in = start; unsigned char code = 0; @@ -433,18 +534,35 @@ static unsigned char* readName(unsigned char* start, unsigned char* end, SkPdfOb break; } - *out = code; + if (hasOut) { *out = code; } out++; in++; } else { - *out = *in; + if (hasOut) { *out = *in; } out++; in++; } } - SkPdfObject::makeName(start, out, name); - return in; + if (hasOut) { + return in; + } else { + return start + (out - (const unsigned char*)NULL); // return where the string would end if we reuse the string + } +} + +static int readNameLength(int level, const unsigned char* start, const unsigned char* end) { + return readName(level, start, end, NULL) - start; +} + +static const unsigned char* readName(int level, const unsigned char* start, const unsigned char* end, SkPdfObject* name, SkPdfAllocator* allocator) { + int outLength = readNameLength(level, start, end); + // TODO(edisonn): optimize the allocation, don't allocate new string, but put it in a preallocated buffer + unsigned char* out = (unsigned char*)allocator->alloc(outLength); + start = readName(level, start, end, out); + SkPdfObject::makeName(out, out + outLength, name); + TRACE_NAME(out, out + outLength); + return start; } // TODO(edisonn): pdf spec let Length to be an indirect object define after the stream @@ -471,8 +589,9 @@ and it could get worse, with multiple object like this // right now implement the silly algorithm that assumes endstream is finishing the stream -static unsigned char* readStream(unsigned char* start, unsigned char* end, SkPdfObject* dict, SkNativeParsedPDF* doc) { - start = skipPdfWhiteSpaces(start, end); +static const unsigned char* readStream(int level, const unsigned char* start, const unsigned char* end, SkPdfObject* dict, SkNativeParsedPDF* doc) { + TRACE_INDENT(level, "Stream"); + start = skipPdfWhiteSpaces(level, start, end); if (!(start[0] == 's' && start[1] == 't' && start[2] == 'r' && start[3] == 'e' && start[4] == 'a' && start[5] == 'm')) { // no stream. return. return start; @@ -505,7 +624,7 @@ static unsigned char* readStream(unsigned char* start, unsigned char* end, SkPdf if (length < 0) { // scan the buffer, until we find first endstream // TODO(edisonn): all buffers must have a 0 at the end now, - unsigned char* endstream = (unsigned char*)strrstrk((char*)start, (char*)end, "endstream"); + const unsigned char* endstream = (const unsigned char*)strrstrk((char*)start, (char*)end, "endstream"); if (endstream) { length = endstream - start; @@ -514,7 +633,7 @@ static unsigned char* readStream(unsigned char* start, unsigned char* end, SkPdf } } if (length >= 0) { - unsigned char* endstream = start + length; + const unsigned char* endstream = start + length; if (endstream[0] == kCR_PdfWhiteSpace && endstream[1] == kLF_PdfWhiteSpace) { endstream += 2; @@ -532,7 +651,8 @@ static unsigned char* readStream(unsigned char* start, unsigned char* end, SkPdf return start; } -static unsigned char* readInlineImageStream(unsigned char* start, unsigned char* end, SkPdfImageDictionary* inlineImage, SkNativeParsedPDF* doc) { +static const unsigned char* readInlineImageStream(int level, const unsigned char* start, const unsigned char* end, SkPdfImageDictionary* inlineImage, SkNativeParsedPDF* doc) { + TRACE_INDENT(level, "Inline Image"); // We already processed ID keyword, and we should be positioned immediately after it // TODO(edisonn): security: read after end check, or make buffers with extra 2 bytes @@ -547,8 +667,8 @@ static unsigned char* readInlineImageStream(unsigned char* start, unsigned char* // TODO(edisonn): warning? } - unsigned char* endstream = (unsigned char*)strrstrk((char*)start, (char*)end, "EI"); - unsigned char* endEI = endstream ? endstream + 2 : NULL; // 2 == strlen("EI") + const unsigned char* endstream = (const unsigned char*)strrstrk((char*)start, (char*)end, "EI"); + const unsigned char* endEI = endstream ? endstream + 2 : NULL; // 2 == strlen("EI") if (endstream) { int length = endstream - start; @@ -563,32 +683,34 @@ static unsigned char* readInlineImageStream(unsigned char* start, unsigned char* return endEI; } -static unsigned char* readDictionary(unsigned char* start, unsigned char* end, SkPdfObject* dict, SkPdfAllocator* allocator, SkNativeParsedPDF* doc) { +static const unsigned char* readDictionary(int level, const unsigned char* start, const unsigned char* end, SkPdfObject* dict, SkPdfAllocator* allocator, SkNativeParsedPDF* doc) { + TRACE_INDENT(level, "Dictionary"); SkPdfObject::makeEmptyDictionary(dict); - start = skipPdfWhiteSpaces(start, end); + start = skipPdfWhiteSpaces(level, start, end); + SkPdfAllocator tmpStorage; // keys will be stored in dict, we can free them immediately after set. while (start < end && *start == kNamed_PdfDelimiter) { SkPdfObject key; - *start = '\0'; + //*start = '\0'; start++; - start = readName(start, end, &key); - start = skipPdfWhiteSpaces(start, end); + start = readName(level + 1, start, end, &key, &tmpStorage); + start = skipPdfWhiteSpaces(level + 1, start, end); if (start < end) { SkPdfObject* value = allocator->allocObject(); - start = nextObject(start, end, value, allocator, doc); + start = nextObject(level + 1, start, end, value, allocator, doc); - start = skipPdfWhiteSpaces(start, end); + start = skipPdfWhiteSpaces(level + 1, start, end); if (start < end) { // seems we have an indirect reference if (isPdfDigit(*start)) { SkPdfObject generation; - start = nextObject(start, end, &generation, allocator, doc); + start = nextObject(level + 1, start, end, &generation, allocator, doc); SkPdfObject keywordR; - start = nextObject(start, end, &keywordR, allocator, doc); + start = nextObject(level + 1, start, end, &keywordR, allocator, doc); if (value->isInteger() && generation.isInteger() && keywordR.isKeywordReference()) { int64_t id = value->intValue(); @@ -608,7 +730,7 @@ static unsigned char* readDictionary(unsigned char* start, unsigned char* end, S dict->set(&key, value); return end; } - start = skipPdfWhiteSpaces(start, end); + start = skipPdfWhiteSpaces(level + 1, start, end); } else { dict->set(&key, &SkPdfObject::kNull); return end; @@ -618,30 +740,30 @@ static unsigned char* readDictionary(unsigned char* start, unsigned char* end, S // TODO(edisonn): options to ignore these errors // now we should expect >> - start = skipPdfWhiteSpaces(start, end); + start = skipPdfWhiteSpaces(level, start, end); if (*start != kClosedInequityBracket_PdfDelimiter) { // TODO(edisonn): report/warning } - *start = '\0'; + //*start = '\0'; start++; // skip > if (*start != kClosedInequityBracket_PdfDelimiter) { // TODO(edisonn): report/warning } - *start = '\0'; + //*start = '\0'; start++; // skip > - start = readStream(start, end, dict, doc); + start = readStream(level, start, end, dict, doc); return start; } -unsigned char* nextObject(unsigned char* start, unsigned char* end, SkPdfObject* token, SkPdfAllocator* allocator, SkNativeParsedPDF* doc) { - unsigned char* current; +const unsigned char* nextObject(int level, const unsigned char* start, const unsigned char* end, SkPdfObject* token, SkPdfAllocator* allocator, SkNativeParsedPDF* doc) { + const unsigned char* current; // skip white spaces - start = skipPdfWhiteSpaces(start, end); + start = skipPdfWhiteSpaces(level, start, end); - current = endOfPdfToken(start, end); + current = endOfPdfToken(level, start, end); // no token, len would be 0 if (current == start) { @@ -654,27 +776,27 @@ unsigned char* nextObject(unsigned char* start, unsigned char* end, SkPdfObject* // start array switch (*start) { case kOpenedSquareBracket_PdfDelimiter: - *start = '\0'; + //*start = '\0'; SkPdfObject::makeEmptyArray(token); - return readArray(current, end, token, allocator, doc); + return readArray(level + 1, current, end, token, allocator, doc); case kOpenedRoundBracket_PdfDelimiter: - *start = '\0'; - return readString(start, end, token); + //*start = '\0'; + return readString(level, start + 1, end, token, allocator); case kOpenedInequityBracket_PdfDelimiter: - *start = '\0'; + //*start = '\0'; if (end > start + 1 && start[1] == kOpenedInequityBracket_PdfDelimiter) { - start[1] = '\0'; // optional + //start[1] = '\0'; // optional // TODO(edisonn): pass here the length somehow? - return readDictionary(start + 2, end, token, allocator, doc); // skip << + return readDictionary(level + 1, start + 2, end, token, allocator, doc); // skip << } else { - return readHexString(start + 1, end, token); // skip < + return readHexString(level, start + 1, end, token, allocator); // skip < } case kNamed_PdfDelimiter: - *start = '\0'; - return readName(start + 1, end, token); + //*start = '\0'; + return readName(level, start + 1, end, token, allocator); // TODO(edisonn): what to do curly brackets? read spec! case kOpenedCurlyBracket_PdfDelimiter: @@ -746,28 +868,26 @@ SkPdfObject* SkPdfAllocator::allocObject() { // TODO(edisonn): perf: do no copy the buffers, but use them, and mark cache the result, so there is no need of a second pass SkPdfNativeTokenizer::SkPdfNativeTokenizer(SkPdfObject* objWithStream, const SkPdfMapper* mapper, SkPdfAllocator* allocator, SkNativeParsedPDF* doc) : fDoc(doc), fMapper(mapper), fAllocator(allocator), fUncompressedStream(NULL), fUncompressedStreamEnd(NULL), fEmpty(false), fHasPutBack(false) { - unsigned char* buffer = NULL; + const unsigned char* buffer = NULL; size_t len = 0; - objWithStream->GetFilteredStreamRef(&buffer, &len, fAllocator); + objWithStream->GetFilteredStreamRef(&buffer, &len); // TODO(edisonn): hack, find end of object char* endobj = strrstrk((char*)buffer, (char*)buffer + len, "endobj"); if (endobj) { len = endobj - (char*)buffer + strlen("endobj"); } - fUncompressedStreamStart = fUncompressedStream = (unsigned char*)fAllocator->alloc(len); + fUncompressedStreamStart = fUncompressedStream = buffer; fUncompressedStreamEnd = fUncompressedStream + len; - memcpy(fUncompressedStream, buffer, len); } -SkPdfNativeTokenizer::SkPdfNativeTokenizer(unsigned char* buffer, int len, const SkPdfMapper* mapper, SkPdfAllocator* allocator, SkNativeParsedPDF* doc) : fDoc(doc), fMapper(mapper), fAllocator(allocator), fEmpty(false), fHasPutBack(false) { +SkPdfNativeTokenizer::SkPdfNativeTokenizer(const unsigned char* buffer, int len, const SkPdfMapper* mapper, SkPdfAllocator* allocator, SkNativeParsedPDF* doc) : fDoc(doc), fMapper(mapper), fAllocator(allocator), fEmpty(false), fHasPutBack(false) { // TODO(edisonn): hack, find end of object char* endobj = strrstrk((char*)buffer, (char*)buffer + len, "endobj"); if (endobj) { len = endobj - (char*)buffer + strlen("endobj"); } - fUncompressedStreamStart = fUncompressedStream = (unsigned char*)fAllocator->alloc(len); + fUncompressedStreamStart = fUncompressedStream = buffer; fUncompressedStreamEnd = fUncompressedStream + len; - memcpy(fUncompressedStream, buffer, len); } SkPdfNativeTokenizer::~SkPdfNativeTokenizer() { @@ -777,13 +897,13 @@ bool SkPdfNativeTokenizer::readTokenCore(PdfToken* token) { token->fKeyword = NULL; token->fObject = NULL; - fUncompressedStream = skipPdfWhiteSpaces(fUncompressedStream, fUncompressedStreamEnd); + fUncompressedStream = skipPdfWhiteSpaces(0, fUncompressedStream, fUncompressedStreamEnd); if (fUncompressedStream >= fUncompressedStreamEnd) { return false; } SkPdfObject obj; - fUncompressedStream = nextObject(fUncompressedStream, fUncompressedStreamEnd, &obj, fAllocator, fDoc); + fUncompressedStream = nextObject(0, fUncompressedStream, fUncompressedStreamEnd, &obj, fAllocator, fDoc); // If it is a keyword, we will only get the pointer of the string if (obj.type() == SkPdfObject::kKeyword_PdfObjectType) { @@ -797,7 +917,7 @@ bool SkPdfNativeTokenizer::readTokenCore(PdfToken* token) { token->fType = kObject_TokenType; } -#ifdef PDF_TRACE +#ifdef PDF_TRACE_READ_TOKEN static int read_op = 0; read_op++; if (548 == read_op) { @@ -813,7 +933,7 @@ void SkPdfNativeTokenizer::PutBack(PdfToken token) { SkASSERT(!fHasPutBack); fHasPutBack = true; fPutBack = token; -#ifdef PDF_TRACE +#ifdef PDF_TRACE_READ_TOKEN printf("PUT_BACK %s %s\n", token.fType == kKeyword_TokenType ? "Keyword" : "Object", token.fKeyword ? std::string(token.fKeyword, token.fKeywordLength).c_str(): token.fObject->toString().c_str()); #endif } @@ -822,14 +942,14 @@ bool SkPdfNativeTokenizer::readToken(PdfToken* token) { if (fHasPutBack) { *token = fPutBack; fHasPutBack = false; -#ifdef PDF_TRACE +#ifdef PDF_TRACE_READ_TOKEN printf("READ_BACK %s %s\n", token->fType == kKeyword_TokenType ? "Keyword" : "Object", token->fKeyword ? std::string(token->fKeyword, token->fKeywordLength).c_str() : token->fObject->toString().c_str()); #endif return true; } if (fEmpty) { -#ifdef PDF_TRACE +#ifdef PDF_TRACE_READ_TOKEN printf("EMPTY TOKENIZER\n"); #endif return false; @@ -911,7 +1031,7 @@ static SkPdfObject* inlineImageValueAbbreviationExpand(SkPdfObject* value) { SkPdfImageDictionary* SkPdfNativeTokenizer::readInlineImage() { // BI already processed - fUncompressedStream = skipPdfWhiteSpaces(fUncompressedStream, fUncompressedStreamEnd); + fUncompressedStream = skipPdfWhiteSpaces(0, fUncompressedStream, fUncompressedStreamEnd); if (fUncompressedStream >= fUncompressedStreamEnd) { return NULL; } @@ -921,14 +1041,14 @@ SkPdfImageDictionary* SkPdfNativeTokenizer::readInlineImage() { while (fUncompressedStream < fUncompressedStreamEnd) { SkPdfObject* key = fAllocator->allocObject(); - fUncompressedStream = nextObject(fUncompressedStream, fUncompressedStreamEnd, key, fAllocator, fDoc); + fUncompressedStream = nextObject(0, fUncompressedStream, fUncompressedStreamEnd, key, fAllocator, fDoc); if (key->isKeyword() && key->len() == 2 && key->c_str()[0] == 'I' && key->c_str()[1] == 'D') { // ID - fUncompressedStream = readInlineImageStream(fUncompressedStream, fUncompressedStreamEnd, inlineImage, fDoc); + fUncompressedStream = readInlineImageStream(0, fUncompressedStream, fUncompressedStreamEnd, inlineImage, fDoc); return inlineImage; } else { SkPdfObject* obj = fAllocator->allocObject(); - fUncompressedStream = nextObject(fUncompressedStream, fUncompressedStreamEnd, obj, fAllocator, fDoc); + fUncompressedStream = nextObject(0, fUncompressedStream, fUncompressedStreamEnd, obj, fAllocator, fDoc); // TODO(edisonn): perf maybe we should not expand abreviation like this inlineImage->set(inlineImageKeyAbbreviationExpand(key), inlineImageValueAbbreviationExpand(obj)); diff --git a/experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.h b/experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.h index ad425d8e02..2884937aaf 100644 --- a/experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.h +++ b/experimental/PdfViewer/pdfparser/native/SkPdfNativeTokenizer.h @@ -62,9 +62,9 @@ class SkPdfImageDictionary; #define isPdfDigit(ch) ((ch)>='0'&&(ch)<='9') #define isPdfNumeric(ch) (isPdfDigit(ch)||(ch)=='+'||(ch)=='-') -unsigned char* skipPdfWhiteSpaces(unsigned char* buffer, size_t len); -unsigned char* endOfPdfToken(unsigned char* start, size_t len); -unsigned char* skipPdfComment(unsigned char* start, size_t len); +const unsigned char* skipPdfWhiteSpaces(int level, const unsigned char* buffer, size_t len); +const unsigned char* endOfPdfToken(int level, const unsigned char* start, size_t len); +const unsigned char* skipPdfComment(int level, const unsigned char* start, size_t len); // TODO(edisonn): typedef read and integer tyepes? make less readable... //typedef double SkPdfReal; @@ -111,7 +111,7 @@ public: }; class SkNativeParsedPDF; -unsigned char* nextObject(unsigned char* start, unsigned char* end, SkPdfObject* token, SkPdfAllocator* allocator, SkNativeParsedPDF* doc); +const unsigned char* nextObject(int level, const unsigned char* start, const unsigned char* end, SkPdfObject* token, SkPdfAllocator* allocator, SkNativeParsedPDF* doc); enum SkPdfTokenType { kKeyword_TokenType, @@ -130,7 +130,7 @@ struct PdfToken { class SkPdfNativeTokenizer { public: SkPdfNativeTokenizer(SkPdfObject* objWithStream, const SkPdfMapper* mapper, SkPdfAllocator* allocator, SkNativeParsedPDF* doc); - SkPdfNativeTokenizer(unsigned char* buffer, int len, const SkPdfMapper* mapper, SkPdfAllocator* allocator, SkNativeParsedPDF* doc); + SkPdfNativeTokenizer(const unsigned char* buffer, int len, const SkPdfMapper* mapper, SkPdfAllocator* allocator, SkNativeParsedPDF* doc); virtual ~SkPdfNativeTokenizer(); @@ -144,9 +144,9 @@ private: const SkPdfMapper* fMapper; SkPdfAllocator* fAllocator; - unsigned char* fUncompressedStreamStart; - unsigned char* fUncompressedStream; - unsigned char* fUncompressedStreamEnd; + const unsigned char* fUncompressedStreamStart; + const unsigned char* fUncompressedStream; + const unsigned char* fUncompressedStreamEnd; bool fEmpty; bool fHasPutBack; diff --git a/experimental/PdfViewer/pdfparser/native/SkPdfObject.cpp b/experimental/PdfViewer/pdfparser/native/SkPdfObject.cpp index 57e9abb368..eb342af730 100644 --- a/experimental/PdfViewer/pdfparser/native/SkPdfObject.cpp +++ b/experimental/PdfViewer/pdfparser/native/SkPdfObject.cpp @@ -8,19 +8,27 @@ SkPdfObject SkPdfObject::kNull = SkPdfObject::makeNull(); -bool SkPdfObject::applyFlateDecodeFilter(SkPdfAllocator* allocator) { +bool SkPdfObject::applyFlateDecodeFilter() { if (!SkFlate::HaveFlate()) { // TODO(edisonn): warn, make callers handle it return false; } - SkMemoryStream skstream(fStr.fBuffer, fStr.fBytes >> 1, false); + const unsigned char* old = fStr.fBuffer; + bool deleteOld = isStreamOwned(); + + SkMemoryStream skstream(fStr.fBuffer, fStr.fBytes >> 2, false); SkDynamicMemoryWStream uncompressedData; if (SkFlate::Inflate(&skstream, &uncompressedData)) { - fStr.fBytes = (uncompressedData.bytesWritten() << 1) + kUnfilteredStreamBit; - fStr.fBuffer = (unsigned char*)allocator->alloc(uncompressedData.bytesWritten()); - uncompressedData.copyTo(fStr.fBuffer); + fStr.fBytes = (uncompressedData.bytesWritten() << 2) + kOwnedStreamBit + kUnfilteredStreamBit; + fStr.fBuffer = (const unsigned char*)new unsigned char[uncompressedData.bytesWritten()]; + uncompressedData.copyTo((void*)fStr.fBuffer); + + if (deleteOld) { + delete[] old; + } + return true; } else { // TODO(edisonn): warn, make callers handle it @@ -28,24 +36,24 @@ bool SkPdfObject::applyFlateDecodeFilter(SkPdfAllocator* allocator) { } } -bool SkPdfObject::applyDCTDecodeFilter(SkPdfAllocator* allocator) { +bool SkPdfObject::applyDCTDecodeFilter() { // this would fail, and it won't allow any more filters. // technically, it would be possible, but not a real world scenario // TODO(edisonn): or get the image here and store it for fast retrieval? return false; } -bool SkPdfObject::applyFilter(const char* name, SkPdfAllocator* allocator) { +bool SkPdfObject::applyFilter(const char* name) { if (strcmp(name, "FlateDecode") == 0) { - return applyFlateDecodeFilter(allocator); + return applyFlateDecodeFilter(); } else if (strcmp(name, "DCTDecode") == 0) { - return applyDCTDecodeFilter(allocator); + return applyDCTDecodeFilter(); } // TODO(edisonn): allert, not supported, but should be implemented asap return false; } -bool SkPdfObject::filterStream(SkPdfAllocator* allocator) { +bool SkPdfObject::filterStream() { if (!hasStream()) { return false; } @@ -58,19 +66,16 @@ bool SkPdfObject::filterStream(SkPdfAllocator* allocator) { if (!stream->has_Filter()) { fStr.fBytes = ((fStr.fBytes >> 1) << 1) + kFilteredStreamBit; - return true; - } - - if (stream->isFilterAName(NULL)) { + } else if (stream->isFilterAName(NULL)) { std::string filterName = stream->getFilterAsName(NULL); - applyFilter(filterName.c_str(), allocator); + applyFilter(filterName.c_str()); } else if (stream->isFilterAArray(NULL)) { const SkPdfArray* filters = stream->getFilterAsArray(NULL); int cnt = filters->size(); for (int i = cnt - 1; i >= 0; i--) { const SkPdfObject* filterName = filters->objAtAIndex(i); if (filterName != NULL && filterName->isName()) { - if (!applyFilter(filterName->nameValue(), allocator)) { + if (!applyFilter(filterName->nameValue())) { break; } } else { @@ -79,7 +84,5 @@ bool SkPdfObject::filterStream(SkPdfAllocator* allocator) { } } - fStr.fBytes = ((fStr.fBytes >> 1) << 1) + kFilteredStreamBit; - return true; } diff --git a/experimental/PdfViewer/pdfparser/native/SkPdfObject.h b/experimental/PdfViewer/pdfparser/native/SkPdfObject.h index d496670bc1..f7171191c8 100644 --- a/experimental/PdfViewer/pdfparser/native/SkPdfObject.h +++ b/experimental/PdfViewer/pdfparser/native/SkPdfObject.h @@ -23,7 +23,7 @@ SkMatrix SkMatrixFromPdfMatrix(double array[6]); #define kFilteredStreamBit 0 #define kUnfilteredStreamBit 1 - +#define kOwnedStreamBit 2 class SkPdfObject { public: @@ -50,11 +50,6 @@ class SkPdfObject { }; private: - struct NotOwnedString { - unsigned char* fBuffer; - size_t fBytes; - }; - struct Reference { unsigned int fId; unsigned int fGen; @@ -108,6 +103,11 @@ public: case kDictionary_PdfObjectType: delete fMap; + if (isStreamOwned()) { + delete[] fStr.fBuffer; + fStr.fBuffer = NULL; + fStr.fBytes = 0; + } break; default: @@ -212,13 +212,13 @@ public: static SkPdfObject kNull; - static void makeNumeric(unsigned char* start, unsigned char* end, SkPdfObject* obj) { + static void makeNumeric(const unsigned char* start, const unsigned char* end, SkPdfObject* obj) { SkASSERT(obj->fObjectType == kInvalid_PdfObjectType); // TODO(edisonn): NYI properly // if has dot (impl), or exceeds max int, is real, otherwise is int bool isInt = true; - for (unsigned char* current = start; current < end; current++) { + for (const unsigned char* current = start; current < end; current++) { if (*current == '.') { isInt = false; break; @@ -241,54 +241,54 @@ public: } - static void makeString(unsigned char* start, SkPdfObject* obj) { + static void makeString(const unsigned char* start, SkPdfObject* obj) { makeStringCore(start, strlen((const char*)start), obj, kString_PdfObjectType); } - static void makeString(unsigned char* start, unsigned char* end, SkPdfObject* obj) { + static void makeString(const unsigned char* start, const unsigned char* end, SkPdfObject* obj) { makeStringCore(start, end - start, obj, kString_PdfObjectType); } - static void makeString(unsigned char* start, size_t bytes, SkPdfObject* obj) { + static void makeString(const unsigned char* start, size_t bytes, SkPdfObject* obj) { makeStringCore(start, bytes, obj, kString_PdfObjectType); } - static void makeHexString(unsigned char* start, SkPdfObject* obj) { + static void makeHexString(const unsigned char* start, SkPdfObject* obj) { makeStringCore(start, strlen((const char*)start), obj, kHexString_PdfObjectType); } - static void makeHexString(unsigned char* start, unsigned char* end, SkPdfObject* obj) { + static void makeHexString(const unsigned char* start, const unsigned char* end, SkPdfObject* obj) { makeStringCore(start, end - start, obj, kHexString_PdfObjectType); } - static void makeHexString(unsigned char* start, size_t bytes, SkPdfObject* obj) { + static void makeHexString(const unsigned char* start, size_t bytes, SkPdfObject* obj) { makeStringCore(start, bytes, obj, kHexString_PdfObjectType); } - static void makeName(unsigned char* start, SkPdfObject* obj) { + static void makeName(const unsigned char* start, SkPdfObject* obj) { makeStringCore(start, strlen((const char*)start), obj, kName_PdfObjectType); } - static void makeName(unsigned char* start, unsigned char* end, SkPdfObject* obj) { + static void makeName(const unsigned char* start, const unsigned char* end, SkPdfObject* obj) { makeStringCore(start, end - start, obj, kName_PdfObjectType); } - static void makeName(unsigned char* start, size_t bytes, SkPdfObject* obj) { + static void makeName(const unsigned char* start, size_t bytes, SkPdfObject* obj) { makeStringCore(start, bytes, obj, kName_PdfObjectType); } - static void makeKeyword(unsigned char* start, SkPdfObject* obj) { + static void makeKeyword(const unsigned char* start, SkPdfObject* obj) { makeStringCore(start, strlen((const char*)start), obj, kKeyword_PdfObjectType); } - static void makeKeyword(unsigned char* start, unsigned char* end, SkPdfObject* obj) { + static void makeKeyword(const unsigned char* start, const unsigned char* end, SkPdfObject* obj) { makeStringCore(start, end - start, obj, kKeyword_PdfObjectType); } - static void makeKeyword(unsigned char* start, size_t bytes, SkPdfObject* obj) { + static void makeKeyword(const unsigned char* start, size_t bytes, SkPdfObject* obj) { makeStringCore(start, bytes, obj, kKeyword_PdfObjectType); } @@ -381,8 +381,8 @@ public: return false; } - // we rewrite all delimiters and white spaces with '\0', so we expect the end of name to be '\0' - SkASSERT(key->fStr.fBuffer[key->fStr.fBytes] == '\0'); + //// we rewrite all delimiters and white spaces with '\0', so we expect the end of name to be '\0' + //SkASSERT(key->fStr.fBuffer[key->fStr.fBytes] == '\0'); return set(key->fStr.fBuffer, key->fStr.fBytes, value); } @@ -411,7 +411,7 @@ public: return NULL; } - SkASSERT(key->fStr.fBuffer[key->fStr.fBytes] == '\0'); + //SkASSERT(key->fStr.fBuffer[key->fStr.fBytes] == '\0'); return get(key->fStr.fBuffer, key->fStr.fBytes); } @@ -441,7 +441,7 @@ public: return NULL; } - SkASSERT(key->fStr.fBuffer[key->fStr.fBytes] == '\0'); + //SkASSERT(key->fStr.fBuffer[key->fStr.fBytes] == '\0'); return get(key->fStr.fBuffer, key->fStr.fBytes); } @@ -637,6 +637,19 @@ public: return (const char*)fStr.fBuffer; } + inline NotOwnedString strRef() { + switch (fObjectType) { + case kString_PdfObjectType: + case kHexString_PdfObjectType: + case kKeyword_PdfObjectType: + return fStr; + + default: + // TODO(edisonn): report/warning + return NotOwnedString(); + } + } + // TODO(edisonn): nameValue2 and stringValue2 are used to make code generation easy, // but it is not a performat way to do it, since it will create an extra copy // remove these functions and make code generated faster @@ -647,7 +660,7 @@ public: // TODO(edisonn): log err return ""; } - return (const char*)fStr.fBuffer; + return std::string((const char*)fStr.fBuffer, fStr.fBytes); } inline std::string stringValue2() const { @@ -657,7 +670,7 @@ public: // TODO(edisonn): log err return ""; } - return (const char*)fStr.fBuffer; + return std::string((const char*)fStr.fBuffer, fStr.fBytes); } inline bool boolValue() const { @@ -713,23 +726,23 @@ public: return SkMatrixFromPdfMatrix(array); } - bool filterStream(SkPdfAllocator* allocator); + bool filterStream(); - bool GetFilteredStreamRef(unsigned char** buffer, size_t* len, SkPdfAllocator* allocator) { + bool GetFilteredStreamRef(unsigned char const** buffer, size_t* len) { // TODO(edisonn): add params that couls let the last filter in place if it is jpeg or png to fast load images if (!hasStream()) { return false; } - filterStream(allocator); + filterStream(); if (buffer) { *buffer = fStr.fBuffer; } if (len) { - *len = fStr.fBytes >> 1; // last bit + *len = fStr.fBytes >> 2; // last 2 bits } return true; @@ -739,7 +752,11 @@ public: return hasStream() && ((fStr.fBytes & 1) == kFilteredStreamBit); } - bool GetUnfilteredStreamRef(unsigned char** buffer, size_t* len) const { + bool isStreamOwned() const { + return hasStream() && ((fStr.fBytes & 2) == kOwnedStreamBit); + } + + bool GetUnfilteredStreamRef(unsigned char const** buffer, size_t* len) const { if (isStreamFiltered()) { return false; } @@ -753,13 +770,13 @@ public: } if (len) { - *len = fStr.fBytes >> 1; // remove slast bit + *len = fStr.fBytes >> 2; // remove last 2 bits } return true; } - bool addStream(unsigned char* buffer, size_t len) { + bool addStream(const unsigned char* buffer, size_t len) { SkASSERT(!hasStream()); SkASSERT(isDictionary()); @@ -849,15 +866,15 @@ public: } private: - static void makeStringCore(unsigned char* start, SkPdfObject* obj, ObjectType type) { + static void makeStringCore(const unsigned char* start, SkPdfObject* obj, ObjectType type) { makeStringCore(start, strlen((const char*)start), obj, type); } - static void makeStringCore(unsigned char* start, unsigned char* end, SkPdfObject* obj, ObjectType type) { + static void makeStringCore(const unsigned char* start, const unsigned char* end, SkPdfObject* obj, ObjectType type) { makeStringCore(start, end - start, obj, type); } - static void makeStringCore(unsigned char* start, size_t bytes, SkPdfObject* obj, ObjectType type) { + static void makeStringCore(const unsigned char* start, size_t bytes, SkPdfObject* obj, ObjectType type) { SkASSERT(obj->fObjectType == kInvalid_PdfObjectType); obj->fObjectType = type; @@ -865,9 +882,9 @@ private: obj->fStr.fBytes = bytes; } - bool applyFilter(const char* name, SkPdfAllocator* allocator); - bool applyFlateDecodeFilter(SkPdfAllocator* allocator); - bool applyDCTDecodeFilter(SkPdfAllocator* allocator); + bool applyFilter(const char* name); + bool applyFlateDecodeFilter(); + bool applyDCTDecodeFilter(); }; class SkPdfStream : public SkPdfObject {}; @@ -880,11 +897,11 @@ class SkPdfNumber : public SkPdfObject {}; class SkPdfName : public SkPdfObject { SkPdfName() : SkPdfObject() { - SkPdfObject::makeName((unsigned char*)"", this); + SkPdfObject::makeName((const unsigned char*)"", this); } public: SkPdfName(char* name) : SkPdfObject() { - this->makeName((unsigned char*)name, this); + this->makeName((const unsigned char*)name, this); } }; |