blob: 55c3aa9b50f3bdbfe804deb4f61539882f46af1b [file] [log] [blame]
// Copyright 2019 The Chromium Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#include "components/page_image_annotation/content/renderer/content_page_annotator_driver.h"
#include "base/base64.h"
#include "base/bind.h"
#include "base/optional.h"
#include "content/public/renderer/render_frame.h"
#include "crypto/sha2.h"
#include "services/image_annotation/public/mojom/image_annotation.mojom.h"
#include "services/service_manager/public/cpp/interface_provider.h"
#include "third_party/blink/public/platform/web_string.h"
#include "third_party/blink/public/web/web_document.h"
#include "third_party/blink/public/web/web_local_frame.h"
#include "third_party/blink/public/web/web_node.h"
namespace page_image_annotation {
namespace {
namespace ia_mojom = image_annotation::mojom;
// The number of milliseconds to wait before traversing the DOM to find image
// elements.
constexpr int kDomCrawlDelayMs = 3000;
// Attempts to produce image metadata for the given element. Will produce a null
// value if the element has a missing or malformed src attribute.
base::Optional<PageAnnotator::ImageMetadata> ProduceMetadata(
const GURL& page_url,
const blink::WebElement element,
const uint64_t node_id) {
const std::string source_id = ContentPageAnnotatorDriver::GenerateSourceId(
page_url, element.GetAttribute("src").Utf8());
if (source_id.empty())
return base::nullopt;
return PageAnnotator::ImageMetadata{node_id, source_id};
ia_mojom::AnnotatorPtr RequestAnnotator(
content::RenderFrame* const render_frame) {
ia_mojom::AnnotatorPtr ptr;
return ptr;
} // namespace
content::RenderFrame* const render_frame)
: RenderFrameObserver(render_frame),
weak_ptr_factory_(this) {}
ContentPageAnnotatorDriver::~ContentPageAnnotatorDriver() {}
// static
ContentPageAnnotatorDriver* ContentPageAnnotatorDriver::GetOrCreate(
content::RenderFrame* const render_frame) {
ContentPageAnnotatorDriver* const existing = Get(render_frame);
if (existing)
return existing;
return new ContentPageAnnotatorDriver(render_frame);
PageAnnotator& ContentPageAnnotatorDriver::GetPageAnnotator() {
return page_annotator_;
blink::WebElement ContentPageAnnotatorDriver::GetElement(
const uint64_t node_id) {
const auto lookup = tracked_elements_.find(node_id);
if (lookup == tracked_elements_.end())
return blink::WebElement();
return lookup->second.second;
// static
std::string ContentPageAnnotatorDriver::GenerateSourceId(
const GURL& page_url,
const std::string& uri_fragment) {
if (uri_fragment.empty())
return std::string();
const GURL src_url = page_url.Resolve(uri_fragment);
if (!src_url.is_valid())
return std::string();
// Assign a source ID: either the URL of this image (if it can be resolved) or
// a hash of its data URI.
if (src_url.SchemeIs("data")) {
const std::string& content = src_url.GetContent();
if (!content.empty()) {
// We use SHA256 since it has comparable (<2x) speed to e.g. crc32, but
// has no known collisions (which could lead to cached results for another
// image being returned for this one).
std::string source_id;
base::Base64Encode(crypto::SHA256HashString(content), &source_id);
return source_id;
} else if (src_url.SchemeIs("http") || src_url.SchemeIs("https")) {
return src_url.spec();
return std::string();
void ContentPageAnnotatorDriver::DidFinishDocumentLoad() {
if (!render_frame()->IsMainFrame())
// Cancel any pending DOM crawl.
// Stop tracking old elements. After a page refresh, we're the only thing
// keeping these elements alive (i.e. they are not still being displayed to
// the user); we need to let Blink garbage collect them.
for (const auto& entry : tracked_elements_) {
// Schedule new DOM crawl after page has likely reached a stable state.
// TODO( this approach is ad-hoc (e.g. uses a heuristic
// delay to wait for a stable DOM) and can cause jank;
// reinvestigate it once we are done prototyping the
// feature.
void ContentPageAnnotatorDriver::OnDestruct() {
delete this;
void ContentPageAnnotatorDriver::FindImages(const GURL& page_url,
blink::WebElement element) {
if (element.ImageContents().isNull()) {
// This element is not an image but it could have children that are.
for (blink::WebNode child = element.FirstChild(); !child.IsNull();
child = child.NextSibling()) {
if (child.IsElementNode())
FindImages(page_url, child.To<blink::WebElement>());
} else {
// This element is an image; attempt to produce metadata for it and begin
// tracking.
const base::Optional<PageAnnotator::ImageMetadata> metadata =
ProduceMetadata(page_url, element, next_node_id_);
if (metadata.has_value())
tracked_elements_.insert({next_node_id_++, {*metadata, element}});
void ContentPageAnnotatorDriver::FindAndTrackImages() {
const blink::WebDocument doc = render_frame()->GetWebFrame()->GetDocument();
if (doc.IsNull() || doc.Body().IsNull())
const GURL page_url(doc.Url().GetString().Utf8(), doc.Url().GetParsed(),
FindImages(page_url, doc.Body());
// Inform the PageAnnotator of the new images.
for (const auto& entry : tracked_elements_) {
base::Unretained(this), entry.first));
SkBitmap ContentPageAnnotatorDriver::GetBitmapForId(const uint64_t node_id) {
return GetElement(node_id).ImageContents();
} // namespace page_image_annotation