Merge pull request #2371 from sgoley/docs/duckdb-localhost

duckdb docs (v1.1) - scrape v1

assets/javascripts/news.json

@@ -1,4 +1,8 @@
 [
+  [
+    "2024-11-23",
+    "New documentation: <a href=\"/duckdb/\">DuckDB</a>"
+  ],
   [
     "2024-08-20",
     "New documentation: <a href=\"/man/\">Linux man pages</a>"

lib/docs/filters/duckdb/attribution.rb

@@ -0,0 +1,12 @@
+# frozen_string_literal: true
+
+module Docs
+  class Duckdb
+    class AttributionFilter < Docs::AttributionFilter
+      def attribution_link
+        url = current_url.to_s.sub! 'http://localhost:8000', 'https://duckdb.org'
+        %(<a href="#{url}" class="_attribution-link">#{url}</a>)
+      end
+    end
+  end
+end

lib/docs/filters/duckdb/clean_html.rb

@@ -0,0 +1,40 @@
+module Docs
+  class Duckdb
+    class CleanHtmlFilter < Filter
+      def call
+        # First extract the main content
+        @doc = at_css('#main_content_wrap', 'main')
+        return doc if @doc.nil?
+
+        doc.prepend_child at_css('.title').remove
+        at_css('.title').name = 'h1'
+
+        # Remove navigation and header elements
+        css('.headerline', '.headlinebar', '.landingmenu', '.search_icon', '#sidebar', '.pagemeta', '.toc_menu', '.section-nav').remove
+
+        # Clean up code blocks
+        css('div.highlighter-rouge').each do |node|
+          node['data-language'] = node['class'][/language-(\w+)/, 1] if node['class']
+          node.content = node.content.strip
+          node.name = 'pre'
+        end
+
+        # Remove unnecessary attributes
+        css('div, span, p').each do |node|
+          node.remove_attribute('style')
+          node.remove_attribute('class')
+        end
+
+        # Remove empty elements
+        css('div, span').each do |node|
+          node.remove if node.content.strip.empty?
+        end
+
+        # Remove script tags
+        css('script').remove
+
+        doc
+      end
+    end
+  end
+end

lib/docs/filters/duckdb/entries.rb

@@ -0,0 +1,45 @@
+module Docs
+  class Duckdb
+    class EntriesFilter < Docs::EntriesFilter
+      def get_name
+        at_css('h1', '.title').content
+      end
+
+      def get_type
+        case subpath
+        when /\Asql\//
+          'SQL Reference'
+        when /\Aapi\//
+          'Client APIs'
+        when /\Aguides\//
+          'How-to Guides'
+        when /\Adata\//
+          'Data Import'
+        when /\Aoperations_manual\//
+          'Operations Manual'
+        when /\Adev\//
+          'Development'
+        when /\Ainternals\//
+          'Internals'
+        when /\Aextensions\//
+          'Extensions'
+        when /\Aarchive\//
+          'Archive'
+        else
+          'Documentation'
+        end
+      end
+
+      def additional_entries
+        entries = []
+        css('h2[id]', 'h3[id]').each do |node|
+          name = node.content.strip
+          # Clean up the name
+          name = name.gsub(/[\r\n\t]/, ' ').squeeze(' ')
+          entries << [name, node['id'], get_type]
+        end
+        entries
+      end
+    end
+  end
+end

lib/docs/scrapers/duckdb.rb

@@ -0,0 +1,42 @@
+module Docs
+  class Duckdb < UrlScraper
+    self.name = 'DuckDB'
+    self.type = 'duckdb'
+    self.root_path = 'index.html'
+    self.links = {
+      home: 'https://duckdb.org/',
+      code: 'https://github.com/duckdb/duckdb'
+    }
+
+    # https://duckdb.org/docs/guides/offline-copy.html
+    # curl -O https://duckdb.org/duckdb-docs.zip; bsdtar xf duckdb-docs.zip; cd duckdb-docs; python -m http.server
+    self.release = '1.1.3'
+    self.base_url = 'http://localhost:8000/docs/'
+
+    html_filters.push 'duckdb/entries', 'duckdb/clean_html'
+    text_filters.replace 'attribution', 'duckdb/attribution'
+
+    options[:container] = '.documentation'
+    
+    options[:skip_patterns] = [
+      /installation/,
+      /archive/,
+      /reference/,
+    ]
+
+    options[:skip] = %w(
+      docs/archive/
+      docs/installation/
+      docs/api/
+    )
+
+    options[:attribution] = <<-HTML
+      &copy; Copyright 2018&ndash;2024 Stichting DuckDB Foundation<br>
+      Licensed under the MIT License.
+    HTML
+
+    def get_latest_version(opts)
+      get_github_tags('duckdb', 'duckdb', opts)[0]['name']
+    end
+  end
+end

public/icons/docs/duckdb/SOURCE

@@ -0,0 +1 @@
+https://github.com/duckdb/duckdb/tree/main/logo