httputil and url plugin improvements, see ChangeLog

3 files changed, 170 insertions, 86 deletions

diff --git a/ChangeLog b/ChangeLog
index 94795136..c4b0750b 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,15 @@
+2007-03-24  Giuseppe Bilotta <giuseppe.bilotta@gmail.com>
+
+	* HttpUtil: new get_response method to follow redirects and get an
+	Net::HTTPResponse object with an open connection, allowing partial
+	body retrieval.
+	* HttpUtil: extend Net::HTTPResponse with a #partial_body() method. The
+	method take an argument (max bytes to download) and an optional block;
+	the method yields all the partial content that it's being downloaded,
+	instead of the single chunks yield by #read_body().
+	* url plugin: use the new HttpUtil methods.
+	* url plugin: do not block while trying to get link info.
+
 2007-03-14  Giuseppe Bilotta <giuseppe.bilotta@gmail.com>
 
 	* Socket filtering: socket data, both input and output, can now be
diff --git a/data/rbot/plugins/url.rb b/data/rbot/plugins/url.rb
index 84ee7e43..0d85d473 100644
--- a/data/rbot/plugins/url.rb
+++ b/data/rbot/plugins/url.rb
@@ -27,89 +27,56 @@ class UrlPlugin < Plugin
     return unless TITLE_RE.match(pagedata)
     title = $1.strip.gsub(/\s*\n+\s*/, " ")
     title = Utils.decode_html_entities title
-    title = title[0..255] if title.length > 255
-    "[Link Info] title: #{title}"
+    "title: #{title}"
   end
 
-  def read_data_from_response(response, amount)
+  def get_title_for_url(uri_str)
 
-    amount_read = 0
-    chunks = []
-
-    response.read_body do |chunk|   # read body now
-
-      amount_read += chunk.length
-
-      # if amount_read > amount
-      #   amount_of_overflow = amount_read - amount
-      #   chunk = chunk[0...-amount_of_overflow]
-      # end
-
-      chunks << chunk
-
-      break if amount_read >= amount
-
-    end
-
-    chunks.join('')
-
-  end
-
-  def get_title_for_url(uri_str, depth=@bot.config['http.max_redir'])
-    # This god-awful mess is what the ruby http library has reduced me to.
-    # Python's HTTP lib is so much nicer. :~(
-
-    if depth == 0
-        raise "Error: Maximum redirects hit."
-    end
-
-    debug "+ Getting #{uri_str.to_s}"
     url = uri_str.kind_of?(URI) ? uri_str : URI.parse(uri_str)
     return if url.scheme !~ /https?/
 
     title = nil
 
-    debug "+ connecting to #{url.host}:#{url.port}"
-    http = @bot.httputil.get_proxy(url)
-    http.start { |http|
-
-      http.request_get(url.request_uri(), @bot.httputil.headers) { |response|
-
+    begin
+      @bot.httputil.get_response(url) { |response|
         case response
-          when Net::HTTPRedirection
-            # call self recursively if this is a redirect
-            redirect_to = response['location']  || '/'
-            debug "+ redirect location: #{redirect_to.inspect}"
-            url = URI.join(url.to_s, redirect_to)
-            debug "+ whee, redirecting to #{url.to_s}!"
-            return get_title_for_url(url, depth-1)
-          when Net::HTTPSuccess
-            if response['content-type'] =~ /^text\//
-              # since the content is 'text/*' and is small enough to
-              # be a webpage, retrieve the title from the page
-              debug "+ getting #{url.request_uri}"
-              # was 5*10^4 ... seems to much to me ... 4k should be enough for everybody ;)
-              data = read_data_from_response(response, 4096)
-              return get_title_from_html(data)
-            else
-              unless @bot.config['url.titles_only']
-                # content doesn't have title, just display info.
-                size = response['content-length'].gsub(/(\d)(?=\d{3}+(?:\.|$))(\d{3}\..*)?/,'\1,\2')
-                size = size ? ", size: #{size} bytes" : ""
-                return "[Link Info] type: #{response['content-type']}#{size}"
-              end
-            end
+        when Net::HTTPSuccess
+          if response['content-type'] =~ /^text\//
+            # since the content is 'text/*' and is small enough to
+            # be a webpage, retrieve the title from the page
+            debug "+ getting #{url.request_uri}"
+
+            # we look for the title in the first 4k bytes
+            # TODO make the amount of data configurable
+            response.partial_body(4096) { |part|
+              title = get_title_from_html(part)
+              return title if title
+            }
+            # if nothing was found, return nothing
+            return
           else
-            return "[Link Info] Error getting link (#{response.code} - #{response.message})"
-          end # end of "case response"
-
-      } # end of request block
-    } # end of http start block
-
-    return title
+            unless @bot.config['url.titles_only']
+              # content doesn't have title, just display info.
+              size = response['content-length'].gsub(/(\d)(?=\d{3}+(?:\.|$))(\d{3}\..*)?/,'\1,\2')
+              size = size ? ", size: #{size} bytes" : ""
+              return "type: #{response['content-type']}#{size}"
+            end
+          end
+        when Net::HTTPResponse
+          return "Error getting link (#{response.code} - #{response.message})"
+        else
+          raise response
+        end
+      }
+    rescue Object => e
+      if e.class <= StandardError
+        error e.inspect
+        debug e.backtrace.join("\n")
+      end
 
-  rescue SocketError => e
-    return "[Link Info] Error connecting to site (#{e.message})"
+      msg = e.respond_to?(:message) ? e.message : e.to_s
+      return "Error connecting to site (#{e.message})"
+    end
   end
 
   def listen(m)
@@ -122,17 +89,19 @@ class UrlPlugin < Plugin
         list = @registry[m.target]
 
         if @bot.config['url.display_link_info']
-          debug "Getting title for #{urlstr}..."
-          begin
-          title = get_title_for_url urlstr
-          if title
-            m.reply title
-            debug "Title found!"
-          else
-            debug "Title not found!"
-          end
-          rescue => e
-            debug "Failed: #{e}"
+          Thread.start do
+            debug "Getting title for #{urlstr}..."
+            begin
+              title = get_title_for_url urlstr
+              if title
+                m.reply "[Link Info] #{title}"
+                debug "Title found!"
+              else
+                debug "Title not found!"
+              end
+            rescue => e
+              debug "Failed: #{e}"
+            end
           end
         end
 
diff --git a/lib/rbot/core/utils/httputil.rb b/lib/rbot/core/utils/httputil.rb
index ff2fb6c6..6ca12d5b 100644
--- a/lib/rbot/core/utils/httputil.rb
+++ b/lib/rbot/core/utils/httputil.rb
@@ -10,9 +10,6 @@
 # Copyright:: (C) 2006 Tom Gilbert, Giuseppe Bilotta
 # Copyright:: (C) 2006,2007 Giuseppe Bilotta
 
-module ::Irc
-module Utils
-
 require 'resolv'
 require 'net/http'
 begin
@@ -22,8 +19,30 @@ rescue LoadError => e
   error "Secured HTTP connections will fail"
 end
 
+module ::Net
+  class HTTPResponse
+    # Read chunks from the body until we have at least _size_ bytes, yielding
+    # the partial text at each chunk. Return the partial body.
+    def partial_body(size, &block)
+
+      partial = String.new
+
+      self.read_body { |chunk|
+        partial << chunk
+        yield partial
+        break if size and partial.length >= size
+      }
+
+      return partial
+    end
+  end
+end
+
 Net::HTTP.version_1_2
 
+module ::Irc
+module Utils
+
 # class for making http requests easier (mainly for plugins to use)
 # this class can check the bot proxy configuration to determine if a proxy
 # needs to be used, which includes support for per-url proxy configuration.
@@ -264,6 +283,90 @@ class HttpUtil
     return nil
   end
 
+  # uri::         uri to query (Uri object or String)
+  # opts::        options. Currently used:
+  # :open_timeout::     open timeout for the proxy
+  # :read_timeout::     read timeout for the proxy
+  # :cache::            should we cache results?
+  #
+  # This method is used to get responses following redirections.
+  #
+  # It will return either a Net::HTTPResponse or an error.
+  #
+  # If a block is given, it will yield the response or error instead of
+  # returning it
+  #
+  def get_response(uri_or_str, opts={}, &block)
+    if uri_or_str.kind_of?(URI)
+      uri = uri_or_str
+    else
+      uri = URI.parse(uri_or_str.to_s)
+    end
+    debug "Getting #{uri}"
+
+    options = {
+      :read_timeout => 10,
+      :open_timeout => 5,
+      :max_redir => @bot.config["http.max_redir"],
+      :cache => false,
+      :yield => :none
+    }.merge(opts)
+
+    cache = options[:cache]
+
+    proxy = get_proxy(uri)
+    proxy.open_timeout = options[:open_timeout]
+    proxy.read_timeout = options[:read_timeout]
+
+    begin
+      proxy.start() {|http|
+        req = Net::HTTP::Get.new(uri.request_uri(), @headers)
+        if uri.user and uri.password
+          req.basic_auth(uri.user, uri.password)
+        end
+        http.request(req) { |resp|
+          case resp
+          when Net::HTTPSuccess
+            if cache
+              debug "Caching #{uri.to_s}"
+              cache_response(uri.to_s, resp)
+            end
+          when Net::HTTPRedirection
+            if resp.key?('location')
+              new_loc = URI.join(uri, resp['location']) rescue URI.parse(resp['location'])
+              debug "Redirecting #{uri} to #{new_loc}"
+              if options[:max_redir] > 0
+                new_opts = options.dup
+                new_opts[:max_redir] -= 1
+                return get_response(new_loc, new_opts, &block)
+              else
+                raise "Too many redirections"
+              end
+            end
+          end
+          if block_given?
+            yield resp
+          else
+            return resp
+          end
+        }
+      }
+    rescue StandardError, Timeout::Error => e
+      error "HttpUtil.get_response exception: #{e.inspect}, while trying to get #{uri}"
+      debug e.backtrace.join("\n")
+      def e.body
+        nil
+      end
+      if block_given?
+        yield e
+      else
+        return e
+      end
+    end
+
+    raise "This shouldn't happen"
+  end
+
   def cache_response(k, resp)
     begin
       if resp.key?('pragma') and resp['pragma'] == 'no-cache'