Add the possibility to transform the URL in "OpenLink".

Marcel0024 · Sep 26, 2024 · 59357ac · 59357ac
1 parent eb5d18e
commit 59357ac
Show file tree

Hide file tree

Showing 5 changed files with 37 additions and 13 deletions.
diff --git a/CocoCrawler/Builders/PageCrawlJobBuilder.cs b/CocoCrawler/Builders/PageCrawlJobBuilder.cs
@@ -1,4 +1,5 @@
-using CocoCrawler.CrawlOutputs;
+using AngleSharp.Dom;
+using CocoCrawler.CrawlOutputs;
 using CocoCrawler.Exceptions;
 using CocoCrawler.Job;
 using CocoCrawler.Job.PageBrowserActions;
@@ -60,14 +61,17 @@ public PageCrawlJobBuilder ConfigurePageActions(Action<PageActionsBuilder> optio
         return this;
     }
 
+
+
     /// <summary>
     /// Adds a task to open a page and perform openLinks tasks.
     /// </summary>
     /// <param name="linksSelector">The CSS selector to select the element to openLinks.</param>
     /// <param name="tasks">The array of openLinks tasks to perform.</param>
     /// <param name="options">The action to configure the page actions for the openLinks tasks.</param>
+    /// <param name="linksSelectorFunc">A function to execute for each matching element, that produces the URL to follow.</param>
     /// <returns>The updated <see cref="PageCrawlJobBuilder"/> instance.</returns>
-    public PageCrawlJobBuilder OpenLinks(string linksSelector, Action<PageCrawlJobBuilder> jobOptions, Action<PageActionsBuilder>? options = null)
+    public PageCrawlJobBuilder OpenLinks(string linksSelector, Action<PageCrawlJobBuilder> jobOptions, Action<PageActionsBuilder>? options = null, Func<IElement, string?>? linksSelectorFunc = null)
     {
         PageActionsBuilder? pageActionsBuilder = null;
 
@@ -82,7 +86,7 @@ public PageCrawlJobBuilder OpenLinks(string linksSelector, Action<PageCrawlJobBu
 
         jobOptions(builder);
 
-        Tasks.Add(new CrawlPageOpenLinksTask(linksSelector, builder, pageActionsBuilder?.Build()));
+        Tasks.Add(new CrawlPageOpenLinksTask(linksSelector, builder, pageActionsBuilder?.Build(), linksSelectorFunc));
 
         return this;
     }

diff --git a/CocoCrawler/CrawlJob/PageTasks/CrawlPageOpenLinksTask.cs b/CocoCrawler/CrawlJob/PageTasks/CrawlPageOpenLinksTask.cs
@@ -1,12 +1,29 @@
-using CocoCrawler.Builders;
+using AngleSharp.Dom;
+using CocoCrawler.Builders;
 using CocoCrawler.Job.PageBrowserActions;
 
 namespace CocoCrawler.Job.PageTasks;
 
-public class CrawlPageOpenLinksTask(string paginationSelector, PageCrawlJobBuilder builder, PageActions? pageActions = null) : IPageCrawlTask
+public class CrawlPageOpenLinksTask : IPageCrawlTask
 {
-    public string OpenLinksSelector { get; init; } = paginationSelector;
-    public PageActions? PageActions { get; init; } = pageActions;
-    public PageCrawlJobBuilder JobBuilder { get; init; } = builder;
+    public string OpenLinksSelector { get; init; }
+    public PageActions? PageActions { get; init; }
+    public Func<IElement, string?>? LinkProcessor { get; }
+    public PageCrawlJobBuilder JobBuilder { get; init; }
+
+    public CrawlPageOpenLinksTask(string linksSelector, PageCrawlJobBuilder builder, PageActions? pageActions = null)
+    {
+        OpenLinksSelector = linksSelector;
+        PageActions = pageActions;
+        JobBuilder = builder;
+    }
+
+    public CrawlPageOpenLinksTask(string linksSelector, PageCrawlJobBuilder builder, PageActions? pageActions = null, Func<IElement, string?>? linkProcessor = null)
+    {
+        OpenLinksSelector = linksSelector;
+        PageActions = pageActions;
+        LinkProcessor = linkProcessor;
+        JobBuilder = builder;
+    }
 }
 
diff --git a/CocoCrawler/Crawler/PuppeteerCrawler.cs b/CocoCrawler/Crawler/PuppeteerCrawler.cs
@@ -115,7 +115,7 @@ protected virtual void HandlePaginateTask(CrawlPagePaginateTask paginate, PageCr
 
     protected virtual void HandleOpenLinksTask(CrawlPageOpenLinksTask openLinks, PageCrawlJob job, List<PageCrawlJob> newJobs)
     {
-        var urls = Parser!.ParseForLinks(openLinks.OpenLinksSelector);
+        var urls = Parser!.ParseForLinks(openLinks.OpenLinksSelector, openLinks.LinkProcessor);
 
         Logger?.LogDebug("OpenLinks selector returned {Count} Urls found in openLinks task.", urls.Length);
 

diff --git a/CocoCrawler/Parser/AngleSharpParser.cs b/CocoCrawler/Parser/AngleSharpParser.cs
@@ -17,10 +17,12 @@ public virtual async Task Init(string html)
         _document = await context.OpenAsync(req => req.Content(html));
     }
 
-    public virtual string[] ParseForLinks(string linksSelector)
+    public virtual string[] ParseForLinks(string linksSelector, Func<IElement, string?>? linkProcessor = null)
     {
+        linkProcessor ??= (element) => element.GetAttribute("href");
+
         return _document!.QuerySelectorAll(linksSelector)
-            .Select(link => link.GetAttribute("href"))
+            .Select(link => linkProcessor(link))
             .Where(link => link is not null)
             .Select(link => link!)
             .ToArray();

diff --git a/CocoCrawler/Parser/IParser.cs b/CocoCrawler/Parser/IParser.cs
@@ -1,12 +1,13 @@
-using CocoCrawler.Job.PageTasks;
+using AngleSharp.Dom;
+using CocoCrawler.Job.PageTasks;
 using Newtonsoft.Json.Linq;
 
 namespace CocoCrawler.Parser;
 
 public interface IParser
 {
     Task Init(string html);
-    string[] ParseForLinks(string linksSelector);
+    string[] ParseForLinks(string linksSelector, Func<IElement, string?>? linkProcessor = null);
     JArray ExtractList(CrawlPageExtractListTask scrapeList);
     JObject ExtractObject(CrawlPageExtractObjectTask task);
 }