Uri TryGetInlineBaseUrl()
{
var baseRef = QuerySelector("html > head > base[href]")?.GetAttributeValue("href");
if (baseRef == null)
{
return(null);
}
var baseUrl = TryParse.Uri(baseRef, UriKind.Absolute);
return(baseUrl.Scheme == Uri.UriSchemeHttp || baseUrl.Scheme == Uri.UriSchemeHttps
? baseUrl : null);
}
public static Config FromDictionary(IDictionary dictionary)
{
var url = TryParse.Uri(dictionary.Get(ConfigKeys.Url), UriKind.RelativeOrAbsolute);
var host = dictionary.Get(ConfigKeys.Host);
if (url == null && string.IsNullOrEmpty(host))
{
throw new ApplicationException("The \"url\" or \"host\" setting is not found in configuration");
}
var project = dictionary.Get(ConfigKeys.Project);
if (string.IsNullOrEmpty(project))
{
throw new ApplicationException("The \"project\" setting is not found in configuration");
}
if (url != null)
{
return(new Config(url.Host, project)
{
Port = url.Port,
UseSsl = url.Scheme.Contains("https"),
Path = url.AbsolutePath,
Username = dictionary.Get(ConfigKeys.Username),
Password = dictionary.Get(ConfigKeys.Passwrod),
ReportAsynchronously = TryParse.Boolean(dictionary.Get(ConfigKeys.ReportAsynchronously)).GetValueOrDefault(true)
});
}
else
{
return(new Config(host, project)
{
Port = TryParse.Int32(dictionary.Get(ConfigKeys.Port)).GetValueOrDefault(80),
UseSsl = TryParse.Boolean(dictionary.Get(ConfigKeys.UseSsl)).GetValueOrDefault(),
Path = dictionary.Get(ConfigKeys.Path),
Username = dictionary.Get(ConfigKeys.Username),
Password = dictionary.Get(ConfigKeys.Passwrod),
ReportAsynchronously = TryParse.Boolean(dictionary.Get(ConfigKeys.ReportAsynchronously)).GetValueOrDefault(true)
});
}
}
static IEnumerable <QueryResultItem <HttpFetch <HttpContent> > > CrawlImpl(QueryContext context, Uri rootUrl, int depth, Func <Uri, bool> followPredicate)
{
var linkSet = new HashSet <Uri> {
rootUrl
};
var queue = new Queue <KeyValuePair <int, Uri> >();
queue.Enqueue(0.AsKeyTo(rootUrl));
while (queue.Count > 0)
{
var dequeued = queue.Dequeue();
var url = dequeued.Value;
var level = dequeued.Key;
// TODO retry intermittent errors?
var fetchResult = Http.ReturnErrorneousFetch().Get(url).GetResult(context).Single();
var fetch = fetchResult.Value;
if (!fetch.IsSuccessStatusCode)
{
continue;
}
yield return(fetchResult);
context = fetchResult.Context;
if (level >= depth)
{
continue;
}
// If content is HTML then sniff links and add them to the
// queue assuming they are from the same domain and pass the
// user-supplied condition to follow.
var contentMediaType = fetch.Content.Headers.ContentType?.MediaType;
if (!"text/html".Equals(contentMediaType, StringComparison.OrdinalIgnoreCase))
{
continue;
}
var lq =
from e in Query.Singleton(fetch).Links().Content()
select TryParse.Uri(e, UriKind.Absolute) into e
where e != null &&
(e.Scheme == Uri.UriSchemeHttp || e.Scheme == Uri.UriSchemeHttps) &&
!linkSet.Contains(e) &&
rootUrl.Host.Equals(e.Host, StringComparison.OrdinalIgnoreCase) &&
followPredicate(e)
select e;
var links = lq.GetResult(context);
foreach (var e in links)
{
if (linkSet.Add(e))
{
queue.Enqueue((level + 1).AsKeyTo(e.Value));
}
context = e.Context;
}
}
}
public string TryBaseHref(string href) =>
BaseUrl != null
? TryParse.Uri(BaseUrl, href)?.OriginalString ?? href
: href;
static void QueenSongs()
{
var q =
from t in Http.Get(new Uri("https://en.wikipedia.org/wiki/Queen_discography")).Tables().Content()
.Where(t => t.HasClass("wikitable"))
.Take(1)
from tr in t.TableRows((_, trs) => trs)
select tr.FirstOrDefault(e => e?.AttributeValueEquals("scope", "row") == true) into th
where th != null
let a = th.QuerySelector("a[href]")
select new
{
Title = a.GetAttributeValue("title")?.Trim(),
Href = a.Owner.TryBaseHref(a.GetAttributeValue("href")?.Trim()),
}
into e
select new
{
e.Title,
Url = TryParse.Uri(e.Href, UriKind.Absolute),
}
into e
where !string.IsNullOrEmpty(e.Title) && e.Url != null
select e
into album
from html in Http.Get(album.Url).Html().Content()
from tb in html.Tables(".tracklist").Take(2)
let trs = tb.QuerySelectorAll("tr")
let hdrs =
trs.FirstOrDefault(tr => tr.QuerySelectorAll("th").Take(4).Count() >= 3)
?.QuerySelectorAll("th")
.Select(th => th.InnerTextSource.Decoded.Trim())
.ToArray()
where hdrs != null
let idxs =
new[] { "Title", "Writer(s)", "Length" }
.Select(h => Array.FindIndex(hdrs, he => he == h))
.ToArray()
let his = new
{
Title = idxs[0],
Writers = idxs[1],
Length = idxs[2],
}
from tr in trs
let tds =
tr.QuerySelectorAll("td")
.Select(td => td.InnerTextSource.Decoded)
.ToArray()
where tds.Length >= 3
select new
{
Album = album.Title,
Title = tds[his.Title],
Author = his.Writers >= 0 ? tds[his.Writers] : null,
Duration = tds[his.Length],
};
q.Dump();
}
