public void testRelativeLinksWithLeadingSlash() throws IOException {
String url1 = "http://www.example.com/blah/branch1/index.html";
String url2 = "http://www.example.com/blah/branch2/index.html";
String url3 = "http://www.example.com/journals/american_imago/toc/aim60.1.html";
String url4 = "http://www.example.com/css/foo.css";
String url5 = "http://www.example.com/javascript/bar.js";
String source =
"<html><head><title>Test</title></head><body>"
+ "<a href= branch1/index.html>link1</a>"
+ "Filler, with <b>bold</b> tags and<i>others</i>"
+ "<a href=\" branch2/index.html\">link2</a>"
+ "<a href =\" /journals/american_imago/toc/aim60.1.html\">"
+ "<link rel=\"stylesheet\" href=\"/css/foo.css\" >"
+ "<script type=\"text/javascript\" src=\"/javascript/bar.js\"></script>"
+ "Number 1, Spring 2003</a>";
MockCachedUrl mcu = new MockCachedUrl("http://www.example.com/blah/");
mcu.setContent(source);
extractor.extractUrls(
mau, new StringInputStream(source), ENC, "http://www.example.com/blah/", cb);
Set expected = SetUtil.set(url1, url2, url3, url4, url5);
assertEquals(expected, cb.getFoundUrls());
}
private void singleTagParse(
String url, String startTag, String endTag, ArchivalUnit au, boolean shouldParse)
throws IOException {
MockCachedUrl mcu = new MockCachedUrl("http://www.example.com");
String content = makeContent(url, startTag, endTag);
mcu.setContent(content);
MyLinkExtractorCallback cb = new MyLinkExtractorCallback();
extractor.extractUrls(mau, new StringInputStream(content), ENC, "http://www.example.com", cb);
if (shouldParse) {
Set expected = SetUtil.set(url);
assertEquals("Misparsed: " + content, expected, cb.getFoundUrls());
} else {
Set expected = SetUtil.set();
assertEquals("Misparsed: " + content, expected, cb.getFoundUrls());
}
}
private Set parseSingleSource(String source) throws IOException {
MockArchivalUnit mau = new MockArchivalUnit();
LinkExtractor ue = new RegexpCssLinkExtractor();
mau.setLinkExtractor("text/css", ue);
MockCachedUrl mcu = new MockCachedUrl("http://www.example.com", mau);
mcu.setContent(source);
cb.reset();
extractor.extractUrls(mau, new StringInputStream(source), ENC, "http://www.example.com", cb);
return cb.getFoundUrls();
}
public void testDoCrawlImageWithSrcInAltTagAfterSrcProper() throws IOException {
String url = "http://www.example.com/link3.html";
String source =
"<html><head><title>Test</title></head><body>" + "<img src=" + url + " alt=src>link3</a>";
MockCachedUrl mcu = new MockCachedUrl(startUrl);
mcu.setContent(source);
extractor.extractUrls(mau, new StringInputStream(source), ENC, startUrl, cb);
Set expected = SetUtil.set(url);
assertEquals(expected, cb.getFoundUrls());
}
public void testRelativeLinksWithSameName() throws IOException {
String url1 = "http://www.example.com/branch1/index.html";
String url2 = "http://www.example.com/branch2/index.html";
String source =
"<html><head><title>Test</title></head><body>"
+ "<a href=branch1/index.html>link1</a>"
+ "Filler, with <b>bold</b> tags and<i>others</i>"
+ "<a href=branch2/index.html>link2</a>";
MockCachedUrl mcu = new MockCachedUrl("http://www.example.com");
mcu.setContent(source);
extractor.extractUrls(mau, new StringInputStream(source), ENC, "http://www.example.com", cb);
Set expected = SetUtil.set(url1, url2);
assertEquals(expected, cb.getFoundUrls());
}