/** Image with the highest priority will be last in NavigableSet. */
public NavigableSet<ExtractedImage> extract(ExtractorContext ctx) {
final NavigableSet<ExtractedImage> images = new TreeSet<>();
int curImgPriority = MAX_PRIORITY;
String baseUrl = ctx.getUrl();
String imgUrlStr = null;
try {
Document doc = Jsoup.parse(ctx.getOriginalHtml());
for (Element singleImgLink : doc.select("img")) {
String srcImgPath = singleImgLink.attr("src");
if (srcImgPath.contains("logo")) {
continue;
}
String classAtrValue = singleImgLink.attr("class");
if (classAtrValue != null && classAtrValue.equals("thumb")) {
continue;
}
if (!IMG_PATTERN.matcher(srcImgPath).matches()) {
continue;
}
imgUrlStr = URLCanonicalizer.getCanonicalURL(srcImgPath, baseUrl);
URL imageUrl = new URL(imgUrlStr);
PlanarImage planarImage = createPlanarImage(imgUrlStr);
int curArea = planarImage.getWidth() * planarImage.getHeight();
if (curArea >= MIN_AREA && curArea <= MAX_AREA) {
images.add(createImageObject(imageUrl, planarImage, curImgPriority));
--curImgPriority;
}
}
// check META tag
String imgUrlFromTag = ctx.getMetaTags().get("image");
if (imgUrlFromTag != null) {
imgUrlStr = URLCanonicalizer.getCanonicalURL(imgUrlFromTag, ctx.getUrl());
PlanarImage planarImage =
createPlanarImage(URLCanonicalizer.getCanonicalURL(imgUrlFromTag, ctx.getUrl()));
int curArea = planarImage.getWidth() * planarImage.getHeight();
if (curArea >= MIN_AREA && curArea <= MAX_AREA) {
images.add(createImageObject(new URL(imgUrlStr), planarImage, curImgPriority));
--curImgPriority;
}
}
} catch (Exception ex) {
LOG.error(
"Can't download image from url '"
+ imgUrlStr
+ "', exception message: "
+ ex.getMessage(),
ex);
}
return images;
}
