/* * Copyright 2013 Google Inc. * * Use of this source code is governed by a BSD-style license that can be * found in the LICENSE file. */ #include "SkBitmap.h" #include "SkImageDecoder.h" #include "SkOSFile.h" #include "SkRunnable.h" #include "SkSize.h" #include "SkStream.h" #include "SkTDict.h" #include "SkTaskGroup.h" // from the tools directory for replace_char(...) #include "picture_utils.h" #include "SkDiffContext.h" #include "SkImageDiffer.h" #include "skpdiff_util.h" SkDiffContext::SkDiffContext() { fDiffers = nullptr; fDifferCount = 0; } SkDiffContext::~SkDiffContext() { if (fDiffers) { delete[] fDiffers; } } void SkDiffContext::setAlphaMaskDir(const SkString& path) { if (!path.isEmpty() && sk_mkdir(path.c_str())) { fAlphaMaskDir = path; } } void SkDiffContext::setRgbDiffDir(const SkString& path) { if (!path.isEmpty() && sk_mkdir(path.c_str())) { fRgbDiffDir = path; } } void SkDiffContext::setWhiteDiffDir(const SkString& path) { if (!path.isEmpty() && sk_mkdir(path.c_str())) { fWhiteDiffDir = path; } } void SkDiffContext::setLongNames(const bool useLongNames) { longNames = useLongNames; } void SkDiffContext::setDiffers(const SkTDArray& differs) { // Delete whatever the last array of differs was if (fDiffers) { delete[] fDiffers; fDiffers = nullptr; fDifferCount = 0; } // Copy over the new differs fDifferCount = differs.count(); fDiffers = new SkImageDiffer* [fDifferCount]; differs.copy(fDiffers); } static SkString get_common_prefix(const SkString& a, const SkString& b) { const size_t maxPrefixLength = SkTMin(a.size(), b.size()); SkASSERT(maxPrefixLength > 0); for (size_t x = 0; x < maxPrefixLength; ++x) { if (a[x] != b[x]) { SkString result; result.set(a.c_str(), x); return result; } } if (a.size() > b.size()) { return b; } else { return a; } } static SkString get_combined_name(const SkString& a, const SkString& b) { // Note (stephana): We must keep this function in sync with // getImageDiffRelativeUrl() in static/loader.js (under rebaseline_server). SkString result = a; result.append("-vs-"); result.append(b); sk_tools::replace_char(&result, '.', '_'); return result; } void SkDiffContext::addDiff(const char* baselinePath, const char* testPath) { // Load the images at the paths SkBitmap baselineBitmap; SkBitmap testBitmap; if (!SkImageDecoder::DecodeFile(baselinePath, &baselineBitmap)) { SkDebugf("Failed to load bitmap \"%s\"\n", baselinePath); return; } if (!SkImageDecoder::DecodeFile(testPath, &testBitmap)) { SkDebugf("Failed to load bitmap \"%s\"\n", testPath); return; } // Setup a record for this diff fRecordMutex.acquire(); DiffRecord* newRecord = fRecords.addToHead(DiffRecord()); fRecordMutex.release(); // compute the common name SkString baseName = SkOSPath::Basename(baselinePath); SkString testName = SkOSPath::Basename(testPath); if (longNames) { newRecord->fCommonName = get_combined_name(baseName, testName); } else { newRecord->fCommonName = get_common_prefix(baseName, testName); } newRecord->fCommonName.append(".png"); newRecord->fBaselinePath = baselinePath; newRecord->fTestPath = testPath; newRecord->fSize = SkISize::Make(baselineBitmap.width(), baselineBitmap.height()); // only generate diff images if we have a place to store them SkImageDiffer::BitmapsToCreate bitmapsToCreate; bitmapsToCreate.alphaMask = !fAlphaMaskDir.isEmpty(); bitmapsToCreate.rgbDiff = !fRgbDiffDir.isEmpty(); bitmapsToCreate.whiteDiff = !fWhiteDiffDir.isEmpty(); // Perform each diff for (int differIndex = 0; differIndex < fDifferCount; differIndex++) { SkImageDiffer* differ = fDiffers[differIndex]; // Copy the results into data for this record DiffData& diffData = newRecord->fDiffs.push_back(); diffData.fDiffName = differ->getName(); if (!differ->diff(&baselineBitmap, &testBitmap, bitmapsToCreate, &diffData.fResult)) { // if the diff failed, record -1 as the result // TODO(djsollen): Record more detailed information about exactly what failed. // (Image dimension mismatch? etc.) See https://bug.skia.org/2710 ('make skpdiff // report more detail when it fails to compare two images') diffData.fResult.result = -1; continue; } if (bitmapsToCreate.alphaMask && SkImageDiffer::RESULT_CORRECT != diffData.fResult.result && !diffData.fResult.poiAlphaMask.empty() && !newRecord->fCommonName.isEmpty()) { newRecord->fAlphaMaskPath = SkOSPath::Join(fAlphaMaskDir.c_str(), newRecord->fCommonName.c_str()); // compute the image diff and output it SkBitmap copy; diffData.fResult.poiAlphaMask.copyTo(©, kN32_SkColorType); SkImageEncoder::EncodeFile(newRecord->fAlphaMaskPath.c_str(), copy, SkImageEncoder::kPNG_Type, 100); // cleanup the existing bitmap to free up resources; diffData.fResult.poiAlphaMask.reset(); bitmapsToCreate.alphaMask = false; } if (bitmapsToCreate.rgbDiff && SkImageDiffer::RESULT_CORRECT != diffData.fResult.result && !diffData.fResult.rgbDiffBitmap.empty() && !newRecord->fCommonName.isEmpty()) { // TODO(djsollen): Rather than taking the max r/g/b diffs that come back from // a particular differ and storing them as toplevel fields within // newRecord, we should extend outputRecords() to report optional // fields for each differ (not just "result" and "pointsOfInterest"). // See https://bug.skia.org/2712 ('allow skpdiff to report different sets // of result fields for different comparison algorithms') newRecord->fMaxRedDiff = diffData.fResult.maxRedDiff; newRecord->fMaxGreenDiff = diffData.fResult.maxGreenDiff; newRecord->fMaxBlueDiff = diffData.fResult.maxBlueDiff; newRecord->fRgbDiffPath = SkOSPath::Join(fRgbDiffDir.c_str(), newRecord->fCommonName.c_str()); SkImageEncoder::EncodeFile(newRecord->fRgbDiffPath.c_str(), diffData.fResult.rgbDiffBitmap, SkImageEncoder::kPNG_Type, 100); diffData.fResult.rgbDiffBitmap.reset(); bitmapsToCreate.rgbDiff = false; } if (bitmapsToCreate.whiteDiff && SkImageDiffer::RESULT_CORRECT != diffData.fResult.result && !diffData.fResult.whiteDiffBitmap.empty() && !newRecord->fCommonName.isEmpty()) { newRecord->fWhiteDiffPath = SkOSPath::Join(fWhiteDiffDir.c_str(), newRecord->fCommonName.c_str()); SkImageEncoder::EncodeFile(newRecord->fWhiteDiffPath.c_str(), diffData.fResult.whiteDiffBitmap, SkImageEncoder::kPNG_Type, 100); diffData.fResult.whiteDiffBitmap.reset(); bitmapsToCreate.whiteDiff = false; } } } void SkDiffContext::diffDirectories(const char baselinePath[], const char testPath[]) { // Get the files in the baseline, we will then look for those inside the test path SkTArray baselineEntries; if (!get_directory(baselinePath, &baselineEntries)) { SkDebugf("Unable to open path \"%s\"\n", baselinePath); return; } sk_parallel_for(baselineEntries.count(), [&](int i) { const char* baseFilename = baselineEntries[i].c_str(); // Find the real location of each file to compare SkString baselineFile = SkOSPath::Join(baselinePath, baseFilename); SkString testFile = SkOSPath::Join(testPath, baseFilename); // Check that the test file exists and is a file if (sk_exists(testFile.c_str()) && !sk_isdir(testFile.c_str())) { this->addDiff(baselineFile.c_str(), testFile.c_str()); } else { SkDebugf("Baseline file \"%s\" has no corresponding test file\n", baselineFile.c_str()); } }); } void SkDiffContext::diffPatterns(const char baselinePattern[], const char testPattern[]) { // Get the files in the baseline and test patterns. Because they are in sorted order, it's easy // to find corresponding images by matching entry indices. SkTArray baselineEntries; if (!glob_files(baselinePattern, &baselineEntries)) { SkDebugf("Unable to get pattern \"%s\"\n", baselinePattern); return; } SkTArray testEntries; if (!glob_files(testPattern, &testEntries)) { SkDebugf("Unable to get pattern \"%s\"\n", testPattern); return; } if (baselineEntries.count() != testEntries.count()) { SkDebugf("Baseline and test patterns do not yield corresponding number of files\n"); return; } sk_parallel_for(baselineEntries.count(), [&](int i) { this->addDiff(baselineEntries[i].c_str(), testEntries[i].c_str()); }); } void SkDiffContext::outputRecords(SkWStream& stream, bool useJSONP) { RecordList::Iter iter(fRecords, RecordList::Iter::kHead_IterStart); DiffRecord* currentRecord = iter.get(); if (useJSONP) { stream.writeText("var SkPDiffRecords = {\n"); } else { stream.writeText("{\n"); } // TODO(djsollen): Would it be better to use the jsoncpp library to write out the JSON? // This manual approach is probably more efficient, but it sure is ugly. // See https://bug.skia.org/2713 ('make skpdiff use jsoncpp library to write out // JSON output, instead of manual writeText() calls?') stream.writeText(" \"records\": [\n"); while (currentRecord) { stream.writeText(" {\n"); SkString baselineAbsPath = get_absolute_path(currentRecord->fBaselinePath); SkString testAbsPath = get_absolute_path(currentRecord->fTestPath); stream.writeText(" \"commonName\": \""); stream.writeText(currentRecord->fCommonName.c_str()); stream.writeText("\",\n"); stream.writeText(" \"differencePath\": \""); stream.writeText(get_absolute_path(currentRecord->fAlphaMaskPath).c_str()); stream.writeText("\",\n"); stream.writeText(" \"rgbDiffPath\": \""); stream.writeText(get_absolute_path(currentRecord->fRgbDiffPath).c_str()); stream.writeText("\",\n"); stream.writeText(" \"whiteDiffPath\": \""); stream.writeText(get_absolute_path(currentRecord->fWhiteDiffPath).c_str()); stream.writeText("\",\n"); stream.writeText(" \"baselinePath\": \""); stream.writeText(baselineAbsPath.c_str()); stream.writeText("\",\n"); stream.writeText(" \"testPath\": \""); stream.writeText(testAbsPath.c_str()); stream.writeText("\",\n"); stream.writeText(" \"width\": "); stream.writeDecAsText(currentRecord->fSize.width()); stream.writeText(",\n"); stream.writeText(" \"height\": "); stream.writeDecAsText(currentRecord->fSize.height()); stream.writeText(",\n"); stream.writeText(" \"maxRedDiff\": "); stream.writeDecAsText(currentRecord->fMaxRedDiff); stream.writeText(",\n"); stream.writeText(" \"maxGreenDiff\": "); stream.writeDecAsText(currentRecord->fMaxGreenDiff); stream.writeText(",\n"); stream.writeText(" \"maxBlueDiff\": "); stream.writeDecAsText(currentRecord->fMaxBlueDiff); stream.writeText(",\n"); stream.writeText(" \"diffs\": [\n"); for (int diffIndex = 0; diffIndex < currentRecord->fDiffs.count(); diffIndex++) { DiffData& data = currentRecord->fDiffs[diffIndex]; stream.writeText(" {\n"); stream.writeText(" \"differName\": \""); stream.writeText(data.fDiffName); stream.writeText("\",\n"); stream.writeText(" \"result\": "); stream.writeScalarAsText((SkScalar)data.fResult.result); stream.writeText(",\n"); stream.writeText(" \"pointsOfInterest\": "); stream.writeDecAsText(data.fResult.poiCount); stream.writeText("\n"); stream.writeText(" }"); // JSON does not allow trailing commas if (diffIndex + 1 < currentRecord->fDiffs.count()) { stream.writeText(","); } stream.writeText(" \n"); } stream.writeText(" ]\n"); stream.writeText(" }"); currentRecord = iter.next(); // JSON does not allow trailing commas if (currentRecord) { stream.writeText(","); } stream.writeText("\n"); } stream.writeText(" ]\n"); if (useJSONP) { stream.writeText("};\n"); } else { stream.writeText("}\n"); } } void SkDiffContext::outputCsv(SkWStream& stream) { SkTDict columns(2); int cntColumns = 0; stream.writeText("key"); RecordList::Iter iter(fRecords, RecordList::Iter::kHead_IterStart); DiffRecord* currentRecord = iter.get(); // Write CSV header and create a dictionary of all columns. while (currentRecord) { for (int diffIndex = 0; diffIndex < currentRecord->fDiffs.count(); diffIndex++) { DiffData& data = currentRecord->fDiffs[diffIndex]; if (!columns.find(data.fDiffName)) { columns.set(data.fDiffName, cntColumns); stream.writeText(", "); stream.writeText(data.fDiffName); cntColumns++; } } currentRecord = iter.next(); } stream.writeText("\n"); double values[100]; SkASSERT(cntColumns < 100); // Make the array larger, if we ever have so many diff types. RecordList::Iter iter2(fRecords, RecordList::Iter::kHead_IterStart); currentRecord = iter2.get(); while (currentRecord) { for (int i = 0; i < cntColumns; i++) { values[i] = -1; } for (int diffIndex = 0; diffIndex < currentRecord->fDiffs.count(); diffIndex++) { DiffData& data = currentRecord->fDiffs[diffIndex]; int index = -1; SkAssertResult(columns.find(data.fDiffName, &index)); SkASSERT(index >= 0 && index < cntColumns); values[index] = data.fResult.result; } const char* filename = currentRecord->fBaselinePath.c_str() + strlen(currentRecord->fBaselinePath.c_str()) - 1; while (filename > currentRecord->fBaselinePath.c_str() && *(filename - 1) != '/') { filename--; } stream.writeText(filename); for (int i = 0; i < cntColumns; i++) { SkString str; str.printf(", %f", values[i]); stream.writeText(str.c_str()); } stream.writeText("\n"); currentRecord = iter2.next(); } }