/// <summary>
/// Create a new spider to fetch data from some website
/// See use examples on the Github page
/// </summary>
/// <param name="spiderName">A unique name for this spider. Folder will be created with that name</param>
/// <param name="baseUri">The base Uri of the website. Pages outside this Host will not be fetched</param>
/// <param name="params">Additional initialization parameters</param>
public SimpleSpider(string spiderName, Uri baseUri, InitializationParams @params = null)
{
SpiderName = spiderName;
BaseUri = baseUri;
Cacher = @params?.Cacher;
Downloader = @params?.Downloader;
Configuration = @params?.ConfigurationPrototype ?? new Configuration();
initializeConfiguration(spiderName, @params);
LinkCollector = @params?.LinkCollector;
if (Configuration.Auto_AnchorsLinks && LinkCollector == null)
{
LinkCollector = new LinkProcessors.SimpleProcessor();
}
initializeQueues();
// initialize read-only
if (Cacher == null)
{
Cacher = new ContentCacher();
}
if (Downloader == null)
{
Downloader = new HttpClientDownloader();
}
initializeFetchers();
FetchCompleted += fetchCompleted_AutoCollect;
FetchRewrite += fetchRewrite_AutoRewrite;
Parsers = new List <IParserBase>();
if (@params?.Parsers != null)
{
Parsers.AddRange(@params.Parsers);
}
if (@params?.StorageEngine != null)
{
Storage = @params.StorageEngine;
Storage.Initialize(Configuration);
}
logInitialStatus();
}
private void initializeConfiguration(string spiderName, InitializationParams init)
{
var dir = init?.SpiderDirectory;
if (dir == null)
{
dir = new FileInfo(System.Reflection.Assembly.GetEntryAssembly().Location).Directory;
}
var spiderPath = new DirectoryInfo(Path.Combine(dir.FullName, spiderName));
if (!spiderPath.Exists)
{
spiderPath.Create();
}
Configuration.SpiderDirectory = spiderPath;
var dataPath = new DirectoryInfo(Path.Combine(spiderPath.FullName, "Data"));
if (!dataPath.Exists)
{
dataPath.Create();
}
Configuration.SpiderDataDirectory = dataPath;
spiderWorkDataPath = Path.Combine(dataPath.FullName, "privateData.xml");
lock (lockDataPathObject)
{
SpiderWorkData = File.Exists(spiderWorkDataPath)
? SpiderWorkData = XmlSerializerHelper.DeserializeFromFile <SpiderData>(spiderWorkDataPath)
: SpiderWorkData = new SpiderData();
}
if (Configuration.Logger == null)
{
Configuration.Spider_LogFile = Path.Combine(spiderPath.FullName, $"{ spiderName }.log");
Configuration.Logger = new LoggerConfiguration()
.MinimumLevel.Debug()
.WriteTo.Console()
.WriteTo.File(Configuration.Spider_LogFile, rollingInterval: RollingInterval.Day)
.CreateLogger();
}
}
private void initializeConfiguration(string spiderName, InitializationParams init)
{
var dir = init?.SpiderDirectory;
if (dir == null)
{
dir = new FileInfo(System.Reflection.Assembly.GetEntryAssembly().Location).Directory;
}
var spiderPath = new DirectoryInfo(Path.Combine(dir.FullName, spiderName));
if (!spiderPath.Exists)
{
spiderPath.Create();
}
Configuration.SpiderDirectory = spiderPath;
var dataPath = new DirectoryInfo(Path.Combine(spiderPath.FullName, "Data"));
if (!dataPath.Exists)
{
dataPath.Create();
}
Configuration.SpiderDataDirectory = dataPath;
spiderWorkDataPath = Path.Combine(dataPath.FullName, "privateData.xml");
lock (lockDataPathObject)
{
SpiderWorkData = File.Exists(spiderWorkDataPath)
? SpiderWorkData = XmlSerializerHelper.DeserializeFromFile <SpiderData>(spiderWorkDataPath)
: SpiderWorkData = new SpiderData();
}
if (Configuration.Logger == null)
{
initializeWithSerilog(spiderName, spiderPath);
}
}
