Search: Add support for Youtube URLs (#4146)

Closes issue 3300

src/invidious/routes/search.cr

@@ -51,6 +51,12 @@ module Invidious::Routes::Search
     else
       user = env.get? "user"
 
+      # An URL was copy/pasted in the search box.
+      # Redirect the user to the appropriate page.
+      if query.is_url?
+        return env.redirect UrlSanitizer.process(query.text).to_s
+      end
+
       begin
         items = query.process
       rescue ex : ChannelSearchException

src/invidious/search/query.cr

@@ -20,6 +20,9 @@ module Invidious::Search
     property region : String?
     property channel : String = ""
 
+    # Flag that indicates if the smart search features have been disabled.
+    @inhibit_ssf : Bool = false
+
     # Return true if @raw_query is either `nil` or empty
     private def empty_raw_query?
       return @raw_query.empty?
@@ -48,10 +51,18 @@ module Invidious::Search
     )
       # Get the raw search query string (common to all search types). In
       # Regular search mode, also look for the `search_query` URL parameter
-      if @type.regular?
+      _raw_query = params["q"]?
-        @raw_query = params["q"]? || params["search_query"]? || ""
+      _raw_query ||= params["search_query"]? if @type.regular?
-      else
+      _raw_query ||= ""
-        @raw_query = params["q"]? || ""
+
+      # Remove surrounding whitespaces. Mostly useful for copy/pasted URLs.
+      @raw_query = _raw_query.strip
+
+      # Check for smart features (ex: URL search) inhibitor (backslash).
+      # If inhibitor is present, remove it.
+      if @raw_query.starts_with?('\\')
+        @inhibit_ssf = true
+        @raw_query = @raw_query[1..]
       end
 
       # Get the page number (also common to all search types)
@@ -85,7 +96,7 @@ module Invidious::Search
           @filters = Filters.from_iv_params(params)
           @channel = params["channel"]? || ""
 
-          if @filters.default? && @raw_query.includes?(':')
+          if @filters.default? && @raw_query.index(/\w:\w/)
             # Parse legacy filters from query
             @filters, @channel, @query, subs = Filters.from_legacy_filters(@raw_query)
           else
@@ -136,5 +147,22 @@ module Invidious::Search
 
       return params
     end
+
+    # Checks if the query is a standalone URL
+    def is_url? : Bool
+      # If the smart features have been inhibited, don't go further.
+      return false if @inhibit_ssf
+
+      # Only supported in regular search mode
+      return false if !@type.regular?
+
+      # If filters are present, that's a regular search
+      return false if !@filters.default?
+
+      # Simple heuristics: domain name
+      return @raw_query.starts_with?(
+        /(https?:\/\/)?(www\.)?(m\.)?youtu(\.be|be\.com)\//
+      )
+    end
   end
 end

src/invidious/yt_backend/url_sanitizer.cr

@@ -0,0 +1,121 @@
+require "uri"
+
+module UrlSanitizer
+  extend self
+
+  ALLOWED_QUERY_PARAMS = {
+    channel:  ["u", "user", "lb"],
+    playlist: ["list"],
+    search:   ["q", "search_query", "sp"],
+    watch:    [
+      "v",                                  # Video ID
+      "list", "index",                      # Playlist-related
+      "playlist",                           # Unnamed playlist (id,id,id,...) (embed-only?)
+      "t", "time_continue", "start", "end", # Timestamp
+      "lc",                                 # Highlighted comment (watch page only)
+    ],
+  }
+
+  # Returns whether the given string is an ASCII word. This is the same as
+  # running the following regex in US-ASCII locale: /^[\w-]+$/
+  private def ascii_word?(str : String) : Bool
+    return false if str.bytesize != str.size
+
+    str.each_byte do |byte|
+      next if 'a'.ord <= byte <= 'z'.ord
+      next if 'A'.ord <= byte <= 'Z'.ord
+      next if '0'.ord <= byte <= '9'.ord
+      next if byte == '-'.ord || byte == '_'.ord
+
+      return false
+    end
+
+    return true
+  end
+
+  # Return which kind of parameters are allowed based on the
+  # first path component (breadcrumb 0).
+  private def determine_allowed(path_root : String)
+    case path_root
+    when "watch", "w", "v", "embed", "e", "shorts", "clip"
+      return :watch
+    when .starts_with?("@"), "c", "channel", "user", "profile", "attribution_link"
+      return :channel
+    when "playlist", "mix"
+      return :playlist
+    when "results", "search"
+      return :search
+    else # hashtag, post, trending, brand URLs, etc..
+      return nil
+    end
+  end
+
+  # Create a new URI::Param containing only the allowed parameters
+  private def copy_params(unsafe_params : URI::Params, allowed_type) : URI::Params
+    new_params = URI::Params.new
+
+    ALLOWED_QUERY_PARAMS[allowed_type].each do |name|
+      if unsafe_params[name]?
+        # Only copy the last parameter, in case there is more than one
+        new_params[name] = unsafe_params.fetch_all(name)[-1]
+      end
+    end
+
+    return new_params
+  end
+
+  # Transform any user-supplied youtube URL into something we can trust
+  # and use across the code.
+  def process(str : String) : URI
+    # Because URI follows RFC3986 specifications, URL without a scheme
+    # will be parsed as a relative path. So we have to add a scheme ourselves.
+    str = "https://#{str}" if !str.starts_with?(/https?:\/\//)
+
+    unsafe_uri = URI.parse(str)
+    unsafe_host = unsafe_uri.host
+    unsafe_path = unsafe_uri.path
+
+    new_uri = URI.new(path: "/")
+
+    # Redirect to homepage for bogus URLs
+    return new_uri if (unsafe_host.nil? || unsafe_path.nil?)
+
+    breadcrumbs = unsafe_path
+      .split('/', remove_empty: true)
+      .compact_map do |bc|
+        # Exclude attempts at path trasversal
+        next if bc == "." || bc == ".."
+
+        # Non-alnum characters are unlikely in a genuine URL
+        next if !ascii_word?(bc)
+
+        bc
+      end
+
+    # If nothing remains, it's either a legit URL to the homepage
+    # (who does that!?) or because we filtered some junk earlier.
+    return new_uri if breadcrumbs.empty?
+
+    # Replace the original query parameters with the sanitized ones
+    case unsafe_host
+    when .ends_with?("youtube.com")
+      # Use our sanitized path (not forgetting the leading '/')
+      new_uri.path = "/#{breadcrumbs.join('/')}"
+
+      # Then determine which params are allowed, and copy them over
+      if allowed = determine_allowed(breadcrumbs[0])
+        new_uri.query_params = copy_params(unsafe_uri.query_params, allowed)
+      end
+    when "youtu.be"
+      # Always redirect to the watch page
+      new_uri.path = "/watch"
+
+      new_params = copy_params(unsafe_uri.query_params, :watch)
+      new_params["id"] = breadcrumbs[0]
+
+      new_uri.query_params = new_params
+    end
+
+    return new_uri
+  end
+end