@Override
public MediaMetadata getMetadata(MediaScrapeOptions options) throws Exception {
LOGGER.debug("getMetadata() " + options.toString());
if (options.getType() != MediaType.MOVIE) {
throw new UnsupportedMediaTypeException(options.getType());
}
String id = "";
if (StringUtils.isNotBlank(options.getId(providerInfo.getId()))) {
id = options.getId(providerInfo.getId());
}
if (StringUtils.isBlank(id) && options.getResult() != null) {
if (StringUtils.isEmpty(options.getResult().getId())) {
id = StrgUtils.substr(options.getResult().getUrl(), "id=(.*?)");
} else {
id = options.getResult().getId();
}
}
// we can not scrape without zelluloid id and url
if (StringUtils.isBlank(id) && StringUtils.isBlank(options.getResult().getUrl())) {
throw new Exception("cannot scrape without id and url");
}
String detailurl = BASE_URL + "/filme/index.php3?id=" + id;
if (StringUtils.isBlank(id)) {
detailurl = options.getResult().getUrl();
}
MediaMetadata md = new MediaMetadata(providerInfo.getId());
Url url;
try {
url = new CachedUrl(detailurl);
InputStream in = url.getInputStream();
Document doc = Jsoup.parse(in, PAGE_ENCODING, "");
in.close();
// parse title
String title = doc.getElementsByAttributeValue("property", "og:title").attr("content").trim();
md.setTitle(title);
// parse plot
String plot = doc.getElementsByAttributeValue("class", "bigtext").text();
md.setPlot(plot);
md.setTagline(plot.length() > 150 ? plot.substring(0, 150) : plot);
// parse poster
Elements el = doc.getElementsByAttributeValueStarting("src", "/images/poster");
if (el.size() == 1) {
// Poster
MediaArtwork ma =
new MediaArtwork(providerInfo.getId(), MediaArtwork.MediaArtworkType.POSTER);
ma.setPreviewUrl(BASE_URL + el.get(0).attr("src"));
ma.setDefaultUrl(BASE_URL + el.get(0).attr("src"));
ma.setLanguage(options.getLanguage().getLanguage());
md.addMediaArt(ma);
}
// parse year
el = doc.getElementsByAttributeValueContaining("href", "az.php3?j=");
if (el.size() == 1) {
try {
md.setYear(Integer.parseInt(el.get(0).text()));
} catch (Exception ignored) {
}
}
// parse cinema release
el = doc.getElementsByAttributeValueContaining("href", "?v=w");
if (el.size() > 0) {
try {
SimpleDateFormat sdf = new SimpleDateFormat("dd.MM.yyyy");
Date d = sdf.parse(el.get(0).text());
md.setReleaseDate(d);
} catch (Exception e) {
LOGGER.warn("cannot parse cinema release date: " + el.get(0).text());
}
}
// parse original title
md.setOriginalTitle(StrgUtils.substr(doc.toString(), "Originaltitel: (.*?)\\<"));
if (StringUtils.isEmpty(md.getOriginalTitle())) {
md.setOriginalTitle(md.getTitle());
}
// parse runtime
String rt = (StrgUtils.substr(doc.toString(), "ca. (.*?) min"));
if (!rt.isEmpty()) {
try {
md.setRuntime(Integer.valueOf(rt));
} catch (Exception e2) {
LOGGER.warn("cannot convert runtime: " + rt);
}
}
// parse genres
el = doc.getElementsByAttributeValueContaining("href", "az.php3?g=");
for (Element g : el) {
String gid = g.attr("href").substring(g.attr("href").lastIndexOf('=') + 1);
md.addGenre(getTmmGenre(gid));
}
// parse cert
// FSK: ab 12, $230 Mio. Budget
String fsk = StrgUtils.substr(doc.toString(), "FSK: (.*?)[,<]");
if (!fsk.isEmpty()) {
md.addCertification(Certification.findCertification(fsk));
}
// parse rating
Elements ratings = doc.getElementsByAttributeValue("class", "ratingBarTable");
if (ratings.size() == 2) { // get user rating
Element e = ratings.get(1);
// <div>87%</div>
String r = e.getElementsByTag("div").text().replace("%", "");
try {
md.setRating(Float.valueOf(r) / 10); // only 0-10
} catch (Exception e2) {
LOGGER.warn("cannot convert rating: " + r);
}
}
// details page
doc = null;
String detailsUrl = BASE_URL + "/filme/details.php3?id=" + id;
try {
url = new CachedUrl(detailsUrl);
in = url.getInputStream();
doc = Jsoup.parse(in, PAGE_ENCODING, "");
in.close();
} catch (Exception e) {
LOGGER.error("failed to get details: " + e.getMessage());
}
if (doc != null) {
Element tab = doc.getElementById("ccdetails");
int header = 0;
String lastRole = "";
for (Element tr : tab.getElementsByTag("tr")) {
if (tr.toString().contains("dyngfx")) { // header gfx
if (tr.toString().contains("Besetzung")) {
header = 1;
} else if (tr.toString().contains("Crew")) {
header = 2;
} else if (tr.toString().contains("Produktion")) { // company, not producers
header = 3;
} else if (tr.toString().contains("Verleih")) {
header = 4;
} else if (tr.toString().contains("Alternativtitel")) {
header = 5;
}
continue;
} else {
// no header gfx, so data
MediaCastMember mcm = new MediaCastMember();
el = tr.getElementsByTag("td");
if (header == 1) {
// actors
if (el.size() == 2) {
String role = "" + el.get(0).text().trim();
// text() decodes to \u00a0
if (role.equals("\u00a0") || StringUtils.isBlank(role)) {
continue;
}
mcm.setCharacter(role);
mcm.setName(el.get(1).getElementsByTag("a").text());
mcm.setId(
StrgUtils.substr(el.get(1).getElementsByTag("a").attr("href"), "id=(\\d+)"));
mcm.setType(MediaCastMember.CastType.ACTOR);
md.addCastMember(mcm);
// parsing actor pages would we too heavy here just for actor images..
}
} else if (header == 2) {
// crew
if (el.size() == 2) {
String crewrole = el.get(0).html().trim();
mcm.setName(el.get(1).getElementsByTag("a").text());
if (crewrole.equals(" ")) {
crewrole = lastRole; // pop previous
} else {
lastRole = crewrole; // push new
}
mcm.setPart(crewrole);
switch (crewrole) {
case "Regie":
mcm.setType(MediaCastMember.CastType.DIRECTOR);
break;
case "Drehbuch":
mcm.setType(MediaCastMember.CastType.WRITER);
break;
case "Produktion":
mcm.setType(MediaCastMember.CastType.PRODUCER);
break;
default:
mcm.setType(MediaCastMember.CastType.OTHER);
break;
}
mcm.setId(
StrgUtils.substr(el.get(1).getElementsByTag("a").attr("href"), "id=(\\d+)"));
md.addCastMember(mcm);
}
} else if (header == 3) {
// production
md.addProductionCompany(el.get(0).text());
}
}
}
}
// get links page
doc = null;
String linksUrl = BASE_URL + "/filme/links.php3?id=" + id;
try {
url = new CachedUrl(linksUrl);
in = url.getInputStream();
doc = Jsoup.parse(in, PAGE_ENCODING, "");
in.close();
} catch (Exception e) {
LOGGER.error("failed to get links page: " + e.getMessage());
}
if (doc != null) {
el = doc.getElementsByAttributeValueContaining("href", "german.imdb.com");
if (el != null && el.size() > 0) {
String imdb = StrgUtils.substr(el.get(0).attr("href"), "(tt\\d{7})");
if (imdb.isEmpty()) {
imdb = "tt" + StrgUtils.substr(el.get(0).attr("href"), "\\?(\\d+)");
}
md.setId(MediaMetadata.IMDB, imdb);
}
}
} catch (Exception e) {
LOGGER.error("Error parsing " + detailurl);
throw e;
}
return md;
}
@Override
public List<MediaSearchResult> search(MediaSearchOptions options) throws Exception {
LOGGER.debug("search() " + options.toString());
if (options.getMediaType() != MediaType.MOVIE) {
throw new UnsupportedMediaTypeException(options.getMediaType());
}
int year = 0;
if (options.getYear() != 0) {
year = options.getYear();
}
ArrayList<MediaSearchResult> resultList = new ArrayList<>();
String searchUrl = "";
String searchTerm = "";
String imdb = "";
// only title search
if (StringUtils.isNotEmpty(options.getQuery())) {
searchTerm = options.getQuery();
searchUrl = BASE_URL + "/suche/index.php3?qstring=" + URLEncoder.encode(searchTerm, "UTF-8");
LOGGER.debug("search for : " + searchTerm);
} else {
LOGGER.debug("empty searchString");
return resultList;
}
searchTerm = MetadataUtil.removeNonSearchCharacters(searchTerm);
Document doc = null;
try {
Url url = new CachedUrl(searchUrl);
InputStream in = url.getInputStream();
doc = Jsoup.parse(in, PAGE_ENCODING, "");
in.close();
} catch (Exception e) {
LOGGER.error("failed to search for " + searchTerm + ": " + e.getMessage());
}
if (doc == null || doc.text().contains("Interner Fehler")) {
// FIXME: we are using the one which comes with zelluloid - NOT the global one
SearchTitleWithGoogle gs = new SearchTitleWithGoogle();
List<MediaSearchResult> gr = gs.search("zelluloid.de", this.getProviderInfo(), options);
for (MediaSearchResult msr : gr) {
// filter google results - only movie links
if (msr.getUrl().contains("/filme/index.php3")) {
String id = StrgUtils.substr(msr.getUrl(), "id=(.*)");
msr.setId(id);
resultList.add(msr);
}
}
return resultList;
}
// only look for movie links
// Elements filme = doc.getElementsByAttributeValueStarting("href", "hit.php");
// <TR><TD ALIGN=CENTER><IMG SRC="/gfx/icoMovie.gif" WIDTH=26 HEIGHT=26
// ALT="Film"></TD><TD><B><a
// href="hit.php3?hit=3700de0676109950820a042115e98d99-movie-886-23126993-2"
// class="normLight">Twelve
// Monkeys</B> <nobr>(1995)</nobr></a><div class="smallBlur">R: Terry Gilliam</div></TD>
Elements filme = doc.getElementsByTag("tr");
for (Element tr : filme) {
// no nesting trs
if (tr.getElementsByTag("tr").size() > 1) {
continue;
}
// only tr with movie links
Elements as = tr.getElementsByAttributeValueStarting("href", "hit.php3?hit=");
if (as.isEmpty()) {
continue;
}
// and only movies
if (tr.text().contains("TV-Serie")) {
continue;
}
try {
Element a = as.first();
String id = StrgUtils.substr(a.attr("href"), "-movie-(.*?)-");
MediaSearchResult sr = new MediaSearchResult(providerInfo.getId(), options.getMediaType());
sr.setId(id);
if (StringUtils.isEmpty(sr.getTitle())) {
if (a.html().contains("nobr")) {
sr.setTitle(a.ownText());
} else {
sr.setTitle(a.text());
}
}
LOGGER.debug("found movie " + sr.getTitle());
sr.setOriginalTitle(a.getElementsByTag("span").text());
try {
sr.setYear(
Integer.parseInt(
StrgUtils.substr(
tr.getElementsByTag("nobr").text(), ".*(\\d{4}).*"))); // any 4 digit
} catch (Exception ignored) {
}
sr.setUrl(BASE_URL + "/filme/index.php3?id=" + id);
// sr.setPosterUrl(BASE_URL + "/images" + StrgUtils.substr(a.toString(),
// "images(.*?)\\""));
if (imdb.equals(sr.getIMDBId())) {
// perfect match
sr.setScore(1);
} else {
// compare score based on names
float score = MetadataUtil.calculateScore(searchTerm, sr.getTitle());
if (yearDiffers(year, sr.getYear())) {
float diff = (float) Math.abs(year - sr.getYear()) / 100;
LOGGER.debug(
"parsed year does not match search result year - downgrading score by " + diff);
score -= diff;
}
sr.setScore(score);
}
resultList.add(sr);
} catch (Exception e) {
LOGGER.warn("error parsing movie result: " + e.getMessage());
}
}
LOGGER.debug("found " + resultList.size() + " search results");
// didn't we find anything? we may have been redirected to the details page
if (resultList.isEmpty()) {
if (!doc.getElementsByTag("title").text().contains("Suche nach")) {
// redirected to detail page
MediaSearchResult msr = new MediaSearchResult(providerInfo.getId(), options.getMediaType());
Elements el = doc.getElementsByAttributeValueStarting("href", "index.php3?id=");
if (el.size() > 0) {
msr.setId(StrgUtils.substr(el.get(0).attr("href"), "id=(\\d+)"));
}
msr.setTitle(StrgUtils.substr(doc.getElementsByTag("title").text(), "(.*?)\\|").trim());
el = doc.getElementsByAttributeValueContaining("href", "az.php3?j=");
if (el.size() == 1) {
try {
msr.setYear(Integer.parseInt(el.get(0).text()));
} catch (Exception ignored) {
}
}
resultList.add(msr);
}
return resultList;
}
Collections.sort(resultList);
Collections.reverse(resultList);
return resultList;
}
