privoxy-regression-parser: Allow '|' in tokens and values

[privoxy.git] / tools / privoxy-log-parser.pl

diff --git a/tools/privoxy-log-parser.pl b/tools/privoxy-log-parser.pl
index 6749543..c7312e8 100755 (executable)
--- a/tools/privoxy-log-parser.pl
+++ b/tools/privoxy-log-parser.pl
@@ -52,6 +52,7 @@ use constant {
 
     CLI_OPTION_DEFAULT_TO_HTML_OUTPUT => 0,
     CLI_OPTION_TITLE => 'Privoxy-Log-Parser in da house',
 
     CLI_OPTION_DEFAULT_TO_HTML_OUTPUT => 0,
     CLI_OPTION_TITLE => 'Privoxy-Log-Parser in da house',

+    CLI_OPTION_KEEP_DATE => 0,

     CLI_OPTION_NO_EMBEDDED_CSS => 0,
     CLI_OPTION_NO_MSECS => 0,
     CLI_OPTION_NO_SYNTAX_HIGHLIGHTING => 0,
     CLI_OPTION_NO_EMBEDDED_CSS => 0,
     CLI_OPTION_NO_MSECS => 0,
     CLI_OPTION_NO_SYNTAX_HIGHLIGHTING => 0,


@@ -107,6 +108,7 @@ my %h_colours;
 my $header_highlight_regex = '';
 
 my $html_output_mode;
 my $header_highlight_regex = '';
 
 my $html_output_mode;

+my $keep_date_mode;

 my $no_msecs_mode; # XXX: should probably be removed
 my $shorten_thread_ids;
 my $line_end;
 my $no_msecs_mode; # XXX: should probably be removed
 my $shorten_thread_ids;
 my $line_end;


@@ -1061,7 +1063,7 @@ sub handle_loglevel_re_filter($) {
                 return '';
         }
 
                 return '';
         }
 

-        $c =~ s@(?<=\(size )(\d+)\)(?= with)@$h{'Number'}$1$h{'Standard'}@;
+        $c =~ s@(?<=\(size )(\d+)@$h{'Number'}$1$h{'Standard'}@;

         $c =~ s@(?<=\(new size )(\d+)@$h{'Number'}$1$h{'Standard'}@;
         $c =~ s@(?<=produced )(\d+)(?= hits)@$h{'Number'}$1$h{'Standard'}@;
 
         $c =~ s@(?<=\(new size )(\d+)@$h{'Number'}$1$h{'Standard'}@;
         $c =~ s@(?<=produced )(\d+)(?= hits)@$h{'Number'}$1$h{'Standard'}@;
 


@@ -1210,6 +1212,12 @@ sub handle_loglevel_re_filter($) {
 
         return '' unless SHOW_FILTER_READIN_IN;
 
 
         return '' unless SHOW_FILTER_READIN_IN;
 

+    } elsif ($c =~ m/^Decompression didn't result/) {
+
+        # Decompression didn't result in any content.
+
+        # Nothing to highlight.
+

     } else {
 
         found_unknown_content($content);
     } else {
 
         found_unknown_content($content);


@@ -1627,6 +1635,7 @@ sub handle_loglevel_connect($) {
         # Reusing server socket 7 connected to www.privoxy.org. Total requests: 2.
         # Closing server socket 6 connected to d.asset.soup.io. Keep-alive: 0.\
         #  Tainted: 1. Socket alive: 1. Timeout: 60. Configuration file change detected: 0.
         # Reusing server socket 7 connected to www.privoxy.org. Total requests: 2.
         # Closing server socket 6 connected to d.asset.soup.io. Keep-alive: 0.\
         #  Tainted: 1. Socket alive: 1. Timeout: 60. Configuration file change detected: 0.

+        # Reusing server socket 35 connected to nl.wikipedia.org. Requests already sent: 5.

 
         $c =~ s@(?<= socket )(\d+)@$h{'Number'}$1$h{'Standard'}@;
         $c = highlight_matched_host($c, '(?<=for )[^\s]+(?=\.)');
 
         $c =~ s@(?<= socket )(\d+)@$h{'Number'}$1$h{'Standard'}@;
         $c = highlight_matched_host($c, '(?<=for )[^\s]+(?=\.)');


@@ -1634,6 +1643,7 @@ sub handle_loglevel_connect($) {
         for my $number_pattern ('requests', 'Keep-alive', 'Tainted', ' alive', 'Timeout', 'detected') {
             $c = highlight_matched_pattern($c, 'Number', '(?<='. $number_pattern . ': )\d+');
         }
         for my $number_pattern ('requests', 'Keep-alive', 'Tainted', ' alive', 'Timeout', 'detected') {
             $c = highlight_matched_pattern($c, 'Number', '(?<='. $number_pattern . ': )\d+');
         }

+        $c =~ s@(?<=already sent: )(\d+)@$h{'Number'}$1$h{'Standard'}@;

 
     } elsif ($c =~ m/^Connected to /) {
 
 
     } elsif ($c =~ m/^Connected to /) {
 


@@ -1754,6 +1764,33 @@ sub handle_loglevel_connect($) {
         # Forwarding 1954 bytes of encrypted POST data
         $c =~ s@(?<=Forwarding )(\d+)@$h{'Number'}$1$h{'Standard'}@;
 
         # Forwarding 1954 bytes of encrypted POST data
         $c =~ s@(?<=Forwarding )(\d+)@$h{'Number'}$1$h{'Standard'}@;
 

+    } elsif ($c =~ m/^Forwarded the last \d+ bytes/) {
+
+        # Forwarded the last 1954 bytes
+        $c =~ s@(?<=the last )(\d+)@$h{'Number'}$1$h{'Standard'}@;
+
+    } elsif ($c =~ m/^Waiting for the next client connection. Currently active threads:/) {
+
+        # Waiting for the next client connection. Currently active threads: 30
+        $c =~ s@(?<=threads: )(\d+)@$h{'Number'}$1$h{'Standard'}@;
+
+    } elsif ($c =~ m/^Data arrived in time on client socket/) {
+
+        # Data arrived in time on client socket 6. Requests so far: 3
+        $c =~ s@(?<=client socket )(\d+)@$h{'Number'}$1$h{'Standard'}@;
+        $c =~ s@(?<=Requests so far: )(\d+)@$h{'Number'}$1$h{'Standard'}@;
+
+    } elsif ($c =~ m/^Dropping the client connection on socket/) {
+
+        # Dropping the client connection on socket 71. The server connection has not been established yet.
+        $c =~ s@(?<=on socket )(\d+)@$h{'Number'}$1$h{'Standard'}@;
+
+    } elsif ($c =~ m/^The client socket \d+ has become unusable while the server/) {
+
+        # The client socket 16 has become unusable while the server socket 24 is still open.
+        $c =~ s@(?<=client socket )(\d+)@$h{'Number'}$1$h{'Standard'}@;
+        $c =~ s@(?<=server socket )(\d+)@$h{'Number'}$1$h{'Standard'}@;
+

     } elsif ($c =~ m/^Looks like we / or
              $c =~ m/^Unsetting keep-alive flag/ or
              $c =~ m/^No connections to wait/ or
     } elsif ($c =~ m/^Looks like we / or
              $c =~ m/^Unsetting keep-alive flag/ or
              $c =~ m/^No connections to wait/ or


@@ -2007,14 +2044,19 @@ sub gather_loglevel_clf_stats($) {
     our %cli_options;
 
     # +0200] "GET https://www.youtube.com/watch?v=JmcA9LIIXWw HTTP/1.1" 200 68004
     our %cli_options;
 
     # +0200] "GET https://www.youtube.com/watch?v=JmcA9LIIXWw HTTP/1.1" 200 68004

-    $content =~ m/^[+-]\d{4}\] "(\w+) (.+) (HTTP\/\d\.\d)" (\d+) (\d+)/;
+    # +0200] "VERSION-CONTROL http://p.p/ HTTP/1.1" 200 2787
+    $content =~ m/^[+-]\d{4}\] "([^ ]+) (.+) (HTTP\/\d\.\d)" (\d+) (\d+)/;

     $method       = $1;
     $resource     = $2;
     $http_version = $3;
     $status_code  = $4;
     $size         = $5;
 
     $method       = $1;
     $resource     = $2;
     $http_version = $3;
     $status_code  = $4;
     $size         = $5;
 

+    $stats{requests_clf}++;
+

     unless (defined $method) {
     unless (defined $method) {

+        # +0200] "Invalid request" 400 0
+        return if ($content =~ m/^[+-]\d{4}\] "Invalid request"/);

         print("Failed to parse: $content\n");
         return;
     }
         print("Failed to parse: $content\n");
         return;
     }


@@ -2025,12 +2067,11 @@ sub gather_loglevel_clf_stats($) {
     $stats{'http-version'}{$http_version}++;
 
     if ($cli_options{'host-statistics-threshold'} != 0) {
     $stats{'http-version'}{$http_version}++;
 
     if ($cli_options{'host-statistics-threshold'} != 0) {

-        $resource =~ m@(?:http[s]://)([^/]+)/?@;
+        $resource =~ m@(?:https?://)?([^/]+)/?@;

         $stats{'hosts'}{$1}++;
     }
     $stats{'content-size-total'} += $size;
     $stats{'status-code'}{$status_code}++;
         $stats{'hosts'}{$1}++;
     }
     $stats{'content-size-total'} += $size;
     $stats{'status-code'}{$status_code}++;

-    $stats{requests_clf}++;

 }
 
 sub gather_loglevel_request_stats($$) {
 }
 
 sub gather_loglevel_request_stats($$) {


@@ -2349,12 +2390,14 @@ sub print_clf_message() {
 sub print_non_clf_message($) {
 
     my $content = shift;
 sub print_non_clf_message($) {
 
     my $content = shift;

+    my $date_string = $keep_date_mode ? $req{$t}{'day'} . ' ' : '';

     my $msec_string = $no_msecs_mode ? '' : '.' . $req{$t}{'msecs'};
     my $line_start = $html_output_mode ? '' : $h{"Standard"};
 
     return if DEBUG_SUPPRESS_LOG_MESSAGES;
 
     print $line_start
     my $msec_string = $no_msecs_mode ? '' : '.' . $req{$t}{'msecs'};
     my $line_start = $html_output_mode ? '' : $h{"Standard"};
 
     return if DEBUG_SUPPRESS_LOG_MESSAGES;
 
     print $line_start

+        . $date_string

         . $time_colours[$time_colour_index % 2]
         . $req{$t}{'time-stamp'}
         . $msec_string
         . $time_colours[$time_colour_index % 2]
         . $req{$t}{'time-stamp'}
         . $msec_string


@@ -2572,6 +2615,7 @@ sub get_cli_options() {
     our %cli_options = (
         'html-output'              => CLI_OPTION_DEFAULT_TO_HTML_OUTPUT,
         'title'                    => CLI_OPTION_TITLE,
     our %cli_options = (
         'html-output'              => CLI_OPTION_DEFAULT_TO_HTML_OUTPUT,
         'title'                    => CLI_OPTION_TITLE,

+        'keep-date'                => CLI_OPTION_KEEP_DATE,

         'no-syntax-highlighting'   => CLI_OPTION_NO_SYNTAX_HIGHLIGHTING,
         'no-embedded-css'          => CLI_OPTION_NO_EMBEDDED_CSS,
         'no-msecs'                 => CLI_OPTION_NO_MSECS,
         'no-syntax-highlighting'   => CLI_OPTION_NO_SYNTAX_HIGHLIGHTING,
         'no-embedded-css'          => CLI_OPTION_NO_EMBEDDED_CSS,
         'no-msecs'                 => CLI_OPTION_NO_MSECS,


@@ -2588,6 +2632,7 @@ sub get_cli_options() {
     GetOptions (
         'html-output'              => \$cli_options{'html-output'},
         'title'                    => \$cli_options{'title'},
     GetOptions (
         'html-output'              => \$cli_options{'html-output'},
         'title'                    => \$cli_options{'title'},

+        'keep-date'                => \$cli_options{'keep-date'},

         'no-syntax-highlighting'   => \$cli_options{'no-syntax-highlighting'},
         'no-embedded-css'          => \$cli_options{'no-embedded-css'},
         'no-msecs'                 => \$cli_options{'no-msecs'},
         'no-syntax-highlighting'   => \$cli_options{'no-syntax-highlighting'},
         'no-embedded-css'          => \$cli_options{'no-embedded-css'},
         'no-msecs'                 => \$cli_options{'no-msecs'},


@@ -2605,6 +2650,7 @@ sub get_cli_options() {
 
    $html_output_mode = cli_option_is_set('html-output');
    $no_msecs_mode = cli_option_is_set('no-msecs');
 
    $html_output_mode = cli_option_is_set('html-output');
    $no_msecs_mode = cli_option_is_set('no-msecs');

+   $keep_date_mode = cli_option_is_set('keep-date');

    $shorten_thread_ids = cli_option_is_set('shorten-thread-ids');
    $line_end = get_line_end();
 }
    $shorten_thread_ids = cli_option_is_set('shorten-thread-ids');
    $line_end = get_line_end();
 }


@@ -2704,6 +2750,9 @@ omitted, ANSI escape sequences are used unless B<--no-syntax-highlighting> is ac
 This option is only intended to make embedding log excerpts in web pages easier.
 It does not escape any input!
 
 This option is only intended to make embedding log excerpts in web pages easier.
 It does not escape any input!
 

+[B<--keep-date>] Don't remove the date when printing highlighted log messages.
+Useful when parsing multiple log files at once.
+

 [B<--no-msecs>] Don't expect milisecond resolution
 
 [B<--no-syntax-highlighting>] Disable syntax-highlighting. Useful when
 [B<--no-msecs>] Don't expect milisecond resolution
 
 [B<--no-syntax-highlighting>] Disable syntax-highlighting. Useful when