zotero/translators/fr-online.de.js
2011-04-20 22:16:29 +00:00

176 lines
6.8 KiB
JavaScript

{
"translatorID": "488fe1e0-b7d2-406f-8257-5060418ce9b2",
"label": "fr-online.de",
"creator": "Martin Meyerhoff",
"target": "^http://www\\.fr-online\\.de",
"minVersion": "1.0",
"maxVersion": "",
"priority": 100,
"inRepository": "1",
"translatorType": 4,
"lastUpdated": "2011-03-26 15:45:54"
}
/*
fr-online.de Translator
Copyright (C) 2011 Martin Meyerhoff
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
/*
Works w/ search and overviews. I had to include the ugly hack stopping non-articles (photo-streams) to make the multiple item import return an error. Test on:
http://www.fr-online.de/politik/spezials/wikileaks---die-enthuellungsplattform/-/4882932/4882932/-/index.html
http://www.fr-online.de/page/search/fr-online/home/suche/-/1473784/1473784/-/view/asSearch/-/index.html?contextsIds=1472660&docTypes=%22MauArticle,MauGallery,DMBrightcoveVideo,CMDownload,DMMovie,DMEvent,DMVenue%22&offset=5&pageNumber=2&searchMode=SIMPLEALL&sortBy=maupublicationdate&userQuery=Wikileaks
http://www.fr-online.de/wirtschaft/krise/-/1471908/1471908/-/index.html
http://www.fr-online.de/wirtschaft/krise/portugal-koennte-rettungspaket-benoetigen/-/1471908/8251842/-/index.html
*/
function detectWeb(doc, url) {
// I use XPaths. Therefore, I need the following block.
var namespace = doc.documentElement.namespaceURI;
var nsResolver = namespace ? function(prefix) {
if (prefix == 'x') return namespace; else return null;
} : null;
var FR_article_XPath = ".//div[contains(@class, 'ArticleToolBoxIcons')]";
var FR_multiple_XPath = ".//*[@id='ContainerContent']/div/div[contains(@class, 'Headline2')]/a"
if (doc.evaluate(FR_article_XPath, doc, null, XPathResult.ANY_TYPE, null).iterateNext() ){
Zotero.debug("newspaperArticle");
return "newspaperArticle";
} else if (doc.location.href.match(/^http\:\/\/www\.fr-online\.de\/.*?page\/search/) ) {
Zotero.debug("multiple");
return "multiple";
} else if (doc.evaluate(FR_multiple_XPath, doc, null, XPathResult.ANY_TYPE, null).iterateNext() ){
Zotero.debug("multiple");
return "multiple";
}
}
function authorCase(author) { // Turns All-Uppercase-Authors to normally cased Authors
var words = author.split(/\s/);
var authorFixed = '';
for (var i in words) {
words[i] = words[i][0].toUpperCase() + words[i].substr(1).toLowerCase();
authorFixed = authorFixed + words[i] + ' ';
}
return(authorFixed);
}
function scrape(doc, url) {
var namespace = doc.documentElement.namespaceURI;
var nsResolver = namespace ? function(prefix) {
if (prefix == 'x') return namespace; else return null;
} : null;
var FR_article_XPath = ".//div[contains(@class, 'ArticleToolBoxIcons')]"; // this protects against galleries...
if (doc.evaluate(FR_article_XPath, doc, null, XPathResult.ANY_TYPE, null).iterateNext() ){
var newItem = new Zotero.Item("newspaperArticle");
newItem.url = doc.location.href;
// This is for the title!
var title_XPath = '//title'
var title = doc.evaluate(title_XPath, doc, nsResolver, XPathResult.ANY_TYPE, null).iterateNext().textContent;
newItem.title = title.split("|")[0].replace(/^\s*|\s*$/g, '');
// This is for the author!
var author_XPath = '//meta[@name="author"]';
var author= doc.evaluate(author_XPath, doc, nsResolver, XPathResult.ANY_TYPE, null).iterateNext().content;
author = author.split(/\,\s|\sund\s/g);
if (author[0].match(/Rundschau/)) { // Frankfurter Rundschau is no author.
author[0] = "";
}
for (var i in author) {
if (author[i].match(/\s/)) { // only names that contain a space!
author[i] = Zotero.Utilities.trim(author[i]);
author[i] = authorCase(author[i]);
newItem.creators.push(Zotero.Utilities.cleanAuthor(author[i], "author"));
}
}
//Summary
var summary_XPath = '//meta[@name="description"]';
if (doc.evaluate(summary_XPath, doc, null, XPathResult.ANY_TYPE, null).iterateNext() ){
var summary= doc.evaluate(summary_XPath, doc, nsResolver, XPathResult.ANY_TYPE, null).iterateNext().content;
newItem.abstractNote = Zotero.Utilities.trim(summary);
}
//Date
var date_XPath = ".//div[contains(@class, 'TB_Date')]";
var date = doc.evaluate(date_XPath, doc, nsResolver, XPathResult.ANY_TYPE, null).iterateNext().textContent;
date = date.replace(/^\s*Datum\:\s|\s/g, ''); // remove "Datum: " and " "
date = date.split("|");
var realdate = "";
realdate = realdate.concat(date[2], "-", date[1], "-", date[0]);
newItem.date = realdate;
// No Tags. FR does not provide consistently meaningful ones.
// Publikation
newItem.publicationTitle = "fr-online.de"
// Section
var section_XPath = '//title'
var section = doc.evaluate(section_XPath, doc, nsResolver, XPathResult.ANY_TYPE, null).iterateNext().textContent;
section = section.split(/\||-/);
newItem.section = section[1].replace(/^\s*|\s*$/g, '');
// Attachment
var printurl = doc.location.href;
if (printurl.match("asFirstTeaser")) {
printurl = printurl.replace("asFirstTeaser", "printVersion");
} else {
printurl = printurl.replace(/\-\/index.html$/, "-/view/printVersion/-/index.html");
}
newItem.attachments.push({url:printurl, title:doc.title, mimeType:"text/html"});
newItem.complete()
}
}
function doWeb(doc, url) {
var namespace = doc.documentElement.namespaceURI;
var nsResolver = namespace ? function(prefix) {
if (prefix == 'x') return namespace; else return null;
} : null;
var articles = new Array();
if (detectWeb(doc, url) == "multiple") {
var items = new Object();
var titles = doc.evaluate(".//*[@id='ContainerContentLinie']/div/h2/a|.//*[@id='ContainerContent']/div/div[contains(@class, 'Headline2')]/a|.//*[@id='ContainerContent']/div/div/div[contains(@class, 'link_article')]/a|.//*[@id='Main']/div[contains(@class, '2ColHP')]/div/div/div[contains(@class, 'Headline2')]/a", doc, nsResolver, XPathResult.ANY_TYPE, null);
var next_title;
while (next_title = titles.iterateNext()) {
// This excludes the videos, whos link terminates in a hash.
if (next_title.href.match(/.*html$/)) {
items[next_title.href] = next_title.textContent;
}
}
items = Zotero.selectItems(items);
for (var i in items) {
articles.push(i);
}
Zotero.Utilities.processDocuments(articles, scrape, function() {Zotero.done();});
Zotero.wait();
} else {
scrape(doc, url);
}
}