public void TestBaidu()
{
// tests <meta http-equiv="Content-Type" content="text/html;charset=gb2312">
string @in = GetFilePath("/htmltests/baidu-cn-home.html");
Document doc = Dcsoup.ParseFile(@in, null,
"http://www.baidu.com/"); // http charset is gb2312, but NOT specifying it, to test http-equiv parse
Element submit = doc.Select("#su").First;
Assert.AreEqual("百度一下", submit.Attr("value"));
// test from attribute match
submit = doc.Select("input[value=百度一下]").First;
Assert.AreEqual("su", submit.Id);
Element newsLink = doc.Select("a:Contains(新)").First;
Assert.AreEqual("http://news.baidu.com/", newsLink.AbsUrl("href")); // trailing slash to domain name
// check auto-detect from meta
Assert.AreEqual("gb2312", doc.OutputSettings.Charset.WebName); // charset name is lower case
Assert.AreEqual("<title>百度一下，你就知道 </title>", doc.Select("title").OuterHtml);
doc.OutputSettings.Charset = Encoding.ASCII;
Assert.AreEqual("<title>百度一下，你就知道 </title>",
doc.Select("title").OuterHtml);
}
private bool TestValidProtocol(Element el, Nodes.Attribute attr, ICollection <Whitelist.Protocol> protocols)
{
// try to resolve relative urls to abs, and optionally update the attribute so output html has abs.
// rels without a baseuri get removed
string value = el.AbsUrl(attr.Key);
if (value.Length == 0)
{
value = attr.Value;
}
// if it could not be made abs, run as-is to allow custom unknown protocols
if (!preserveRelativeLinks)
{
attr.Value = value;
}
foreach (Whitelist.Protocol protocol in protocols)
{
string prot = protocol.ToString() + ":";
if (value.ToLower().StartsWith(prot, StringComparison.Ordinal))
{
return(true);
}
}
return(false);
}
public void HandleAbsOnLocalhostFileUris()
{
Document doc = Dcsoup.Parse("<a href='password'>One/a><a href='/var/log/messages'>Two</a>", "file://localhost/etc/");
Element one = doc.Select("a").First;
Assert.AreEqual("file://localhost/etc/password", one.AbsUrl("href"));
}
public void testBaidu()
{
// tests <meta http-equiv="Content-Type" content="text/html;charset=gb2312">
using (Stream input = getFile("Test.htmltests.baidu-cn-home.html"))
{
Document doc = NSoup.NSoupClient.Parse(input, null,
"http://www.baidu.com/"); // http charset is gb2312, but NOT specifying it, to test http-equiv parse
Element submit = doc.Select("#su").First;
Assert.AreEqual("百度一下", submit.Attr("value"));
// test from attribute match
submit = doc.Select("input[value=百度一下]").First;
Assert.AreEqual("su", submit.Id);
Element newsLink = doc.Select("a:contains(新)").First;
//Assert.AreEqual("http://news.baidu.com", newsLink.AbsUrl("href"));
Assert.AreEqual("http://news.baidu.com/", newsLink.AbsUrl("href")); // This was changed due to System.Uri's behavior. I think this change is acceptable.
// check auto-detect from meta
Assert.AreEqual("GB2312", doc.OutputSettings().Encoding.WebName.ToUpperInvariant());
Assert.AreEqual("<title>百度一下，你就知道 </title>", doc.Select("title").OuterHtml());
doc.OutputSettings().SetEncoding("ascii");
Assert.AreEqual("<title>百度一下，你就知道 </title>",
doc.Select("title").OuterHtml());
}
}
private bool TestValidProtocol(Element el, NSoup.Nodes.Attribute attr, HashSet <Protocol> protocols)
{
// try to resolve relative urls to abs, and optionally update the attribute so output html has abs.
// rels without a baseuri get removed
string value = el.AbsUrl(attr.Key);
if (value.Length == 0)
{
value = attr.Value; // if it could not be made abs, run as-is to allow custom unknown protocols
}
if (!_preserveRelativeLinks)
{
attr.Value = value;
}
foreach (Protocol protocol in protocols)
{
string prot = protocol.ToString() + ":";
if (value.ToLowerInvariant().StartsWith(prot))
{
return(true);
}
}
return(false);
}
public void handlesAbsOnImage()
{
Document doc = NSoup.NSoupClient.Parse("<p><img src=\"/rez/osi_logo.png\" /></p>", "http://jsoup.org/");
Element img = doc.Select("img").First;
Assert.AreEqual("http://jsoup.org/rez/osi_logo.png", img.Attr("abs:src"));
Assert.AreEqual(img.AbsUrl("src"), img.Attr("abs:src"));
}
public void HandlesProtocolRelativeUrl()
{
string @base = "https://example.com/";
string html = "<img src='//example.net/img.jpg'>";
Document doc = Dcsoup.Parse(html, @base);
Element el = doc.Select("img").First;
Assert.AreEqual("https://example.net/img.jpg", el.AbsUrl("src"));
}
public void HandleAbsOnFileUris()
{
Document doc = Dcsoup.Parse("<a href='password'>One/a><a href='/var/log/messages'>Two</a>", "file:///etc/"); // double slash // is required
Element one = doc.Select("a").First;
Assert.AreEqual("file:///etc/password", one.AbsUrl("href")); // double slash // is required
Element two = doc.Select("a")[1];
Assert.AreEqual("file:///var/log/messages", two.AbsUrl("href")); // double slash // is required
}
public void handlesBaseUri()
{
Tag tag = Tag.ValueOf("a");
Attributes attribs = new Attributes();
attribs.Add("relHref", "/foo");
attribs.Add("absHref", "http://bar/qux");
Element noBase = new Element(tag, "", attribs);
Assert.AreEqual("", noBase.AbsUrl("relHref")); // with no base, should NOT fallback to href attrib, whatever it is
Assert.AreEqual("http://bar/qux", noBase.AbsUrl("absHref")); // no base but valid attrib, return attrib
Element withBase = new Element(tag, "http://foo/", attribs);
Assert.AreEqual("http://foo/foo", withBase.AbsUrl("relHref")); // construct abs from base + rel
Assert.AreEqual("http://bar/qux", withBase.AbsUrl("absHref")); // href is abs, so returns that
Assert.AreEqual("", withBase.AbsUrl("noval"));
Element dodgyBase = new Element(tag, "wtf://no-such-protocol/", attribs);
Assert.AreEqual("http://bar/qux", dodgyBase.AbsUrl("absHref")); // base fails, but href good, so get that
Assert.AreEqual("", dodgyBase.AbsUrl("relHref")); // base fails, only rel href, so return nothing
}
public void absHandlesRelativeQuery()
{
Document doc = NSoup.NSoupClient.Parse("<a href='?foo'>One</a> <a href='bar.html?foo'>Two</a>", "http://jsoup.org/path/file?bar");
Element a1 = doc.Select("a").First;
Assert.AreEqual("http://jsoup.org/path/file?foo", a1.AbsUrl("href"));
Element a2 = doc.Select("a")[1];
Assert.AreEqual("http://jsoup.org/path/bar.html?foo", a2.AbsUrl("href"));
}
public void HandlesAbsOnProtocolessAbsoluteUris()
{
Document doc1 = Dcsoup.Parse("<a href='//example.net/foo'>One</a>", "http://example.com/");
Document doc2 = Dcsoup.Parse("<a href='//example.net/foo'>One</a>", "https://example.com/");
Element one = doc1.Select("a").First;
Element two = doc2.Select("a").First;
Assert.AreEqual("http://example.net/foo", one.AbsUrl("href"));
Assert.AreEqual("https://example.net/foo", two.AbsUrl("href"));
Document doc3 = Dcsoup.Parse("<img src=//www.google.com/images/errors/logo_sm.gif alt=Google>", "https://google.com");
Assert.AreEqual("https://www.google.com/images/errors/logo_sm.gif", doc3.Select("img").Attr("abs:src"));
}
public void MaybeSetBaseUri(Element baseEl)
{
if (_baseUriSetFromDoc) // only listen to the first <base href> in parse
{
return;
}
string href = baseEl.AbsUrl("href");
if (href.Length != 0)
{ // ignore <base target> etc
_baseUri = href;
_baseUriSetFromDoc = true;
_doc.BaseUri = href; // set on the doc so doc.createElement(Tag) will get updated base, and to update all descendants
}
}
public void handlesBaseUri()
{
Tag tag = Tag.ValueOf("a");
Attributes attribs = new Attributes();
attribs.Add("relHref", "/foo");
attribs.Add("absHref", "http://bar/qux");
Element noBase = new Element(tag, "", attribs);
Assert.AreEqual("", noBase.AbsUrl("relHref")); // with no base, should NOT fallback to href attrib, whatever it is
Assert.AreEqual("http://bar/qux", noBase.AbsUrl("absHref")); // no base but valid attrib, return attrib
Element withBase = new Element(tag, "http://foo/", attribs);
Assert.AreEqual("http://foo/foo", withBase.AbsUrl("relHref")); // construct abs from base + rel
Assert.AreEqual("http://bar/qux", withBase.AbsUrl("absHref")); // href is abs, so returns that
Assert.AreEqual("", withBase.AbsUrl("noval"));
Element dodgyBase = new Element(tag, "wtf://no-such-protocol/", attribs);
Assert.AreEqual("http://bar/qux", dodgyBase.AbsUrl("absHref")); // base fails, but href good, so get that
Assert.AreEqual("", dodgyBase.AbsUrl("relHref")); // base fails, only rel href, so return nothing
}
public void handlesAbsPrefixOnHasAttr()
{
// 1: no abs url; 2: has abs url
Document doc = NSoup.NSoupClient.Parse("<a id=1 href='/foo'>One</a> <a id=2 href='http://jsoup.org/'>Two</a>");
Element one = doc.Select("#1").First;
Element two = doc.Select("#2").First;
Assert.IsFalse(one.HasAttr("abs:href"));
Assert.IsTrue(one.HasAttr("href"));
Assert.AreEqual("", one.AbsUrl("href"));
Assert.IsTrue(two.HasAttr("abs:href"));
Assert.IsTrue(two.HasAttr("href"));
Assert.AreEqual("http://jsoup.org/", two.AbsUrl("href"));
}
internal void MaybeSetBaseUri(Element @base)
{
if (baseUriSetFromDoc)
{
// only listen to the first <base href> in parse
return;
}
string href = @base.AbsUrl("href");
if (href.Length != 0)
{
// ignore <base target> etc
baseUri = href;
baseUriSetFromDoc = true;
doc.BaseUri = href;
}
}
private bool TestValidProtocol(Element el, NSoup.Nodes.Attribute attr, HashSet<Protocol> protocols)
{
// try to resolve relative urls to abs, and optionally update the attribute so output html has abs.
// rels without a baseuri get removed
string value = el.AbsUrl(attr.Key);
if (value.Length == 0)
{
value = attr.Value; // if it could not be made abs, run as-is to allow custom unknown protocols
}
if (!_preserveRelativeLinks)
{
attr.Value = value;
}
foreach (Protocol protocol in protocols)
{
string prot = protocol.ToString() + ":";
if (value.ToLowerInvariant().StartsWith(prot))
{
return true;
}
}
return false;
}
public void MaybeSetBaseUri(Element baseEl)
{
if (_baseUriSetFromDoc) // only listen to the first <base href> in parse
{
return;
}
string href = baseEl.AbsUrl("href");
if (href.Length != 0)
{ // ignore <base target> etc
_baseUri = href;
_baseUriSetFromDoc = true;
_doc.BaseUri = href; // set on the doc so doc.createElement(Tag) will get updated base, and to update all descendants
}
}
//Saves all elements in argument
public void DownloadElements(Element[] elements, FileInfo[] fileInfo = null)
{
int totalDownloaded = 0;
int totalBounced = 0;
//Returns if there are no files to be downloaded
if (fileInfo != null && fileInfo.Length == 0)
{
return;
}
for (int i = 0; i < elements.Length; i++)
{
Element content = elements[i];
string absURL;
string tag = content.Tag.ToString();
switch (tag)
{
case "img":
absURL = content.AbsUrl("src");
break;
case "a":
absURL = content.AbsUrl("href");
break;
default:
absURL = content.AbsUrl("src");
break;
}
data.mediaTrie.InsertURL(absURL);
FileInfo file;
//Doesn't recaculate file info if it doesn't have to
if (fileInfo == null)
{
int nameIndex = absURL.LastIndexOf('/');
//Name of the element
string elementName = Regex.Replace(absURL.Substring(nameIndex + 1), "[^A-Za-z.]", "");
//File location of the element
string elementLocation = absURL.Substring(0, nameIndex);
if (elementName.Length > 20)
{
elementName = elementName.Substring(elementName.Length - 20);
}
//Inserts hash into filename to avoid duplicates
string hashCode = Convert.ToString(content.GetHashCode());
elementName = elementName.Insert(0, hashCode);
if (!data.gallery)
{
file = new FileInfo(webStringUtils.UrlToDir(elementLocation) + elementName);
}
else
{
file = new FileInfo(data.outputFolder + elementName);
}
}
else
{
file = fileInfo[i];
}
//Defers downloading to the saver
Save(absURL, file);
//Sleeps to slow down image requests
Thread.Sleep(data.delay);
totalDownloaded++;
}
string report = "Downloaded " + totalDownloaded + " media files, denied " + totalBounced;
CU.WCol(CU.nl + report + CU.nl, CU.c);
}
