125 lines
No EOL
3.9 KiB
JavaScript
125 lines
No EOL
3.9 KiB
JavaScript
{
|
|
"translatorID":"9932d1a7-cc6d-4d83-8462-8f6658b13dc0",
|
|
"translatorType":4,
|
|
"label":"Biblio.com",
|
|
"creator":"Adam Crymble and Michael Berkowitz",
|
|
"target":"http://www.biblio.com/",
|
|
"minVersion":"1.0.0b4.r5",
|
|
"maxVersion":"",
|
|
"priority":100,
|
|
"inRepository":true,
|
|
"lastUpdated":"2008-07-07 14:50:00"
|
|
}
|
|
|
|
function detectWeb(doc, url) {
|
|
if (doc.location.href.match("bookseller_search") || doc.location.href.match("bookstores") || doc.location.href.match("textbooks")) {
|
|
|
|
} else if (doc.location.href.match("search")) {
|
|
return "multiple";
|
|
} else if (doc.location.href.match("books")) {
|
|
return "book";
|
|
}
|
|
}
|
|
|
|
//Biblio.com translator. Code by Adam Crymble.
|
|
|
|
function associateData (newItem, dataTags, field, zoteroField) {
|
|
if (dataTags[field]) {
|
|
newItem[zoteroField] = dataTags[field];
|
|
}
|
|
}
|
|
|
|
function scrape(doc, url) {
|
|
|
|
var namespace = doc.documentElement.namespaceURI;
|
|
var nsResolver = namespace ? function(prefix) {
|
|
if (prefix == 'x') return namespace; else return null;
|
|
} : null;
|
|
|
|
var dataTags = new Object();
|
|
var contents = new Array();
|
|
var multiAuthors = new Array();
|
|
var fieldTitle;
|
|
var author1;
|
|
|
|
var newItem = new Zotero.Item("book");
|
|
|
|
var content = doc.evaluate('//div[@class="smalltext1"][@id="infobox"]/ul/li', doc, nsResolver, XPathResult.ANY_TYPE, null);
|
|
var xPathCount = doc.evaluate('count (//div[@class="smalltext1"][@id="infobox"]/ul/li)', doc, nsResolver, XPathResult.ANY_TYPE, null);
|
|
|
|
for (i=0; i<xPathCount.numberValue; i++) {
|
|
|
|
contents = content.iterateNext().textContent.split(": ");
|
|
fieldTitle = contents[0].replace(/\s*/g, '');
|
|
dataTags[fieldTitle] = contents[1].replace(/^\s*|\s*$/g, '');
|
|
}
|
|
|
|
//Authors
|
|
if (doc.evaluate('//td[2]/h3', doc, nsResolver, XPathResult.ANY_TYPE, null).iterateNext()) {
|
|
var authors = doc.evaluate('//td[2]/h3', doc, nsResolver, XPathResult.ANY_TYPE, null).iterateNext().textContent;
|
|
if (authors.match(/\w/)) {
|
|
if (authors.match(/^ by/)) {
|
|
authors = authors.substr(4);
|
|
if (authors.match(/;/)) {
|
|
multiAuthors = authors.split(";");
|
|
Zotero.debug(multiAuthors);
|
|
for each (var aut in multiAuthors) {
|
|
newItem.creators.push(Zotero.Utilities.cleanAuthor(aut, "author", true));
|
|
}
|
|
} else {
|
|
newItem.creators.push(Zotero.Utilities.cleanAuthor(authors, "author", true));
|
|
}
|
|
} else {
|
|
newItem.creators.push(Zotero.Utilities.cleanAuthor(authors, "author"));
|
|
}
|
|
}
|
|
|
|
}
|
|
|
|
//extra
|
|
if (dataTags["Quantityavailable"]) {
|
|
newItem.extra = "Quantity Available: " + dataTags["Quantityavailable"];
|
|
}
|
|
|
|
associateData (newItem, dataTags, "Publisher", "publisher");
|
|
associateData (newItem, dataTags, "Place", "place");
|
|
associateData (newItem, dataTags, "Datepublished", "date");
|
|
associateData (newItem, dataTags, "ISBN10", "ISBN");
|
|
associateData (newItem, dataTags, "ISBN13", "ISBN");
|
|
associateData (newItem, dataTags, "Pages", "pages");
|
|
associateData (newItem, dataTags, "Edition", "edition");
|
|
|
|
newItem.title = doc.evaluate('//tbody/tr[1]/td[2]/h1', doc, nsResolver, XPathResult.ANY_TYPE, null).iterateNext().textContent.replace(/^\s*|\s&+/g, '');
|
|
newItem.url = doc.location.href;
|
|
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('//table[@class="search-result"]/tbody/tr/td[2]/a', doc, nsResolver, XPathResult.ANY_TYPE, null);
|
|
|
|
var next_title;
|
|
while (next_title = titles.iterateNext()) {
|
|
if (next_title.textContent.match(/\w/)) {
|
|
items[next_title.href] = next_title.textContent;
|
|
}
|
|
}
|
|
items = Zotero.selectItems(items);
|
|
for (var i in items) {
|
|
articles.push(i);
|
|
}
|
|
} else {
|
|
articles = [url];
|
|
}
|
|
Zotero.Utilities.processDocuments(articles, scrape, function() {Zotero.done();});
|
|
Zotero.wait();
|
|
} |