User:ImplSearchBot/Code

From Rosetta Code

<lang perl>#!/usr/bin/perl -w use strict; use Getopt::Long; use MediaWiki::Bot; use JSON qw/to_json from_json/;

our $editor; our $starttime = time; our $pagesedited = 0; our $categorypulls = 0; our $cachehits = 0; our $task_count_change; our $alltasks; our %impldiff; our %createdomitcategories; our @initialcache; our $alllanguages; our @logoutput;

my %options;

  1. ---------------------
  2. These are all the functions that break up our work
  3. into logical chunks.
  4. ---------------------
  1. Called in a *lot* of places to check the machine for overload.

sub slumber { # Snooze. while(&snooze) {}; }

  1. A reversed snooze buton

sub snooze { open my $loadavg, '<', '/proc/loadavg' or die "Failed to check load average: $!"; my $loadstr = <$loadavg>; close $loadavg;

# Wait one minute for every process in the wait queue. my @avgs = split / +/, $loadstr, 3;

my $sleeptime = 0;

# Sleep at least one minute for each process in the wait queue. $sleeptime = $avgs[0] * 60;


# If we're currently under notable load if($sleeptime > 30) { print "Calculated sleeptime: $sleeptime\n"; print "Load averages: " . join(' ', @avgs) . " ... Sleeping $sleeptime seconds\n"; sleep $sleeptime; return $sleeptime; }

return 0; }

  1. Builds a simple hash ref associating a page name with body.
  2. Used to help us prepare our postings in one pass, then commit them
  3. in a second pass.

sub build_posting { &slumber;

   my $name = shift;
   my $body = shift;
   
   return {$name => $body};

}

  1. Prepares the template body for the unimplemented data.

sub prep_unimp_posting { &slumber;

   my $impldiff = shift;
   my $language = shift;
   my $unimplisting = "";
   
   foreach my $taskname (@$alltasks)
   {
       # If it's a category task, the task name will be slightly different.
       my $baretaskname = $taskname;
       $baretaskname =~ s/^Category://;
       my $implpage = $taskname;
       $implpage =~ s/^Category:(.*)/$1\/$language/;
 #      my $escapedImplPage = $implpage;
#       $escapedImplPage =~ s/ /_/g;
       
       my $link;

$link = "Template:Unimpl task link";

  1. $link = "$implpage";
       # Add the task to the unimplemented list, if it's unimplemented, and if it's not in the omit list.
       
       unless(exists $impldiff->{$language}->{'impl'}->{$implpage} or exists $impldiff->{$language}->{'omit'}->{$implpage})
       {
           $unimplisting .= "* $link\n";
       }
       
   }
   
   
   return &build_posting("Template:" . &get_unimpl_template_name($language), $unimplisting);

}

  1. Prepares the template body for the listings of omitted data for a page.

sub prep_omit_posting { &slumber;

   my $implediff = shift;
   my $language = shift;
   my $omitlisting = "";
   my $omittemplatename = &get_omit_template_name($language);
   
   foreach my $taskname (@$alltasks)
   {
       # We want the task name, not the fully-qualified wiki name.
       my $baretaskname = $taskname;
       $baretaskname =~ s/^Category://;
       # Add the task to the omission list, if it's omitted.
       my $escapedTaskName = $baretaskname;
       $escapedTaskName =~ s/ /_/g;
       $omitlisting .= "* $baretaskname\n"
           if(exists $impldiff{$language}->{'omit'}->{$taskname})
   }
   
   # Note that there's no data in the template.
   $implediff->{$language}->{'omit'} = 0
       if( "" eq $omitlisting );
   
   return &build_posting("Template:$omittemplatename", $omitlisting);

}

  1. Prepares the page body that folks look at to find out what changed for a language.

sub prep_listing_posting { &slumber;

   my $impldiff = shift;
   my $language = shift;
   
   my $targetcount = $impldiff->{$language}->{'target_count'};
   my $unimpcount = $impldiff->{$language}->{'unimp_count'};
   
   # Prepare template fields
   my $langfield = "|$language";
   my $unimpfield = "|$unimpcount";
   my $tcfield = "|$targetcount";
   
   my $impperccalc = 0;
   $impperccalc = (($targetcount - $unimpcount) / $targetcount) * 100

unless ($targetcount == 0);

   my $imppercfield = sprintf "|%u", $impperccalc;
   
   # Prepare the listing page format.
   my $pagedata = 'Template:Unimpl header' . $langfield . $unimpfield . $tcfield . $imppercfield . '';
   $pagedata .= "Template:" . &get unimpl template name($language) . "$unimpfield";
   
   if (0 != $impldiff->{$language}->{'omit'})
   {
       $pagedata .= "Template:Omit header" . "$langfield";
       $pagedata .= "Template:" . &get omit template name($language) ."";
   }
   
   $pagedata .= "Template:Unimpl footer$langfield";
   
   return &build_posting(&get_listing_name($language), $pagedata);

}

  1. Prepare the body of the omit category.

sub prep_omit_cat_posting { &slumber;

   my $implediff = shift;
   my $language = shift;
   return &build_posting(&get_omit_cat_name($language), "Tasks listed here have been marked as "un-implementable" in $language. Solutions may be impossible to do, too complex to be of any valuable instruction in $language, or prohibited by the task definition.

But hey; if you think you can prove us wrong, go for it. :-) "); }

  1. Prepare the page title for the page folks look at to see what change what.

sub get_listing_name { &slumber;

   my $language = shift;
   return "Tasks not implemented in $language";

}

  1. Prepare the template name for the omit listing body

sub get_omit_template_name { &slumber;

   my $language = shift;
   return "unimp_omit_body_$language";

}

  1. Prepare the template name for the unimplimended listing body

sub get_unimpl_template_name { &slumber;

   my $language = shift;
   return "unimp_body_$language";

}

  1. Return the name of the category to find the omitted pages for a particular language.

sub get_omit_cat_name { &slumber;

   my $language = shift;
   return "Category:$language/Omit";

}

  1. Prepare all the pages associated with a specific language.

sub process_language { &slumber;

   my $implediff = shift;
   my $language = shift;
   
   
   my $unimpl = &prep_unimp_posting($implediff, $language);
   my $omit = &prep_omit_posting($implediff, $language);
   my $listing = &prep_listing_posting($implediff, $language);
   
   my %langpostings = ( %$unimpl, %$omit, %$listing );
   
   
   unless ( exists $createdomitcategories{&get_omit_cat_name($language)} )
   {
       my $omit_cat = &prep_omit_cat_posting($implediff, $language);
       %langpostings = ( %langpostings, %$omit_cat );
   }
   
   return \%langpostings;

}

sub wikitxt_pathname { &slumber;

   my $page_name = $_[0];
   "test/" . sanitizenamefs($page_name) . ".wikitxt";

}

  1. Prepare and post all of the core pages for this run.

sub processimplediff { &slumber;

   my $implediff = shift;
   my %work;
   
   # Prepare all our work.
   foreach my $language (keys %$implediff)
   {
       # &out("Preparing data for:$language\n", 4);
       my $workitem = &process_language($implediff, $language);
       %work = ( %work, %$workitem );
   }
   # Now that we've prepared all our work, commit it.
   foreach my $pagename (keys %work)
   {
       &postpage($pagename, $work{$pagename}, "Updating $pagename", 0);
   }

}

  1. Return the Mediawiki editor object.

sub geteditor { &slumber;

   # If we're not posting, and we're only drawing from cache
   # We don't actually need to pull from the wiki.
   if( exists $options{'cacheonly'} )
   {
       return undef
           unless exists $options{'post'};
   }
   
   # Handles interaction with the wiki.
   # Note that I had to modify HTTP::Message to make it work
   # HTTP::Message silently failed when presented by MW
   # with an encoding type of "application/json" or some such.
   &out("Creating editor\n", 3);
   my $editor = MediaWiki::Bot->new('ImpleSearchBot');
   $editor->{debug} = 1;
   
   # Tell the editor to edit Rosetta Code.  I'm sure Wikipedia didn't like
   # my initial attempts from before I added this line.
   &out("Trying to set wiki.\n",3);
   $editor->set_wiki('rosettacode.org','mw');
   
   # If we're not posting, we don't need to log in if we're pulling from cache.
   # Otherwise, attempt to log in.
   &out("Trying to log in.\n", 3);
   my $loginres = $editor->login($options{'username'}, $options{'password'});
   
   die "Unable to login: " . $loginres
       unless $loginres == 0;
   
   return $editor;

}

  1. Simple logging infrastructure. Current sends to STDWARN or STDOUT, which cron
  2. emails to the user, which gets forwarded to Short Circuit. (Anyone want a copy?)

sub out { &slumber;

   my $string = shift;
   my $loglevel = shift;
   
   chomp $string;
   push @logoutput, ($string);
   
   if($options{'verbosity'} >= $loglevel)
   {
       if($loglevel == 1)
       {
           warn $string; # use stderr.
       }
       else
       {
           print "$string\n";
       }
   }

}

  1. Many popular filesystems can't andle : and \ in filenames.
  2. Since I plan to open the SVN repo we save to the rest of the world at some point,
  3. I'm trying to make sure the files are representable.

sub sanitizenamefs { &slumber;

   my $pagename = shift;
   $pagename =~ tr/:\//__/;
   return $pagename;

}

  1. Find all the entries that are in the second list ref, but not the first.

sub diffcat_simple { &slumber;

   my $first = shift;
   my $second = shift;
   my %firsthash = map { $_, 1 } @$first;
   
   my @new = ();
   
   foreach my $secondelement (@$second)
   {
       push @new, $secondelement
           unless exists $firsthash{$secondelement};
   }
   
   return @new;

}

  1. Find all the entries that are in one listref, but not the other.

sub diffcat { &slumber;

   my $first = shift;
   my $second = shift;
   my @onlyinfirst = &diffcat_simple($second, $first);
   my @onlyinsecond = &diffcat_simple($first, $second);
   return (\@onlyinfirst, \@onlyinsecond);

}

sub slurp_file { &slumber;

   my $pathname = $_[0];
   open(my $stream, "<", $pathname) or do {warn"Can't open '$pathname': $!\n";
                                           return};
   my $ret;
   while (1) {
       my $len = read($stream, my $buf, 1024);
       if (!defined($len)) { die "I/O error while reading '$pathname': $!"; }
       elsif ($len == 0) { last }
       else { $ret .= $buf }
   }
   $ret;

}

  1. Post a page (or save it to disk, if we're testing.)

sub postpage { &slumber;

   my $pagename = shift;
   my $pagedata = shift;
   my $remark = shift;
   my $minoredit = shift;
   
   
   # MediaWiki won't let us create blank, empty pages.
   # But since we don't want to query (or cache) to see if
   # the page already exists, we'll just add an HTML
   # non-breaking-space entity if the page is truly empty.
   $pagedata = " "
       if(0 == length $pagedata);
   
   ++$pagesedited;
   
   unless( exists $options{'post'} )
   {
       $pagename = wikitxt_pathname($pagename);
       # save it to disk, and out of the way.
       
       &out("Saving: $pagename\n", 2);
       
       open my $outfile, '>', $pagename
           or &out("Failed to open $pagename: $!", 1);
       
       return unless defined $outfile;
       
       print $outfile $pagedata;
       close $outfile;
   }
   else {
       &out("Posting $pagename\n", 2);
   
   $editor->edit($pagename, $pagedata, "ImplSearchBot:$remark", $minoredit)
       or &out("Failed to post page: " . $editor->{'errstr'}, 1);
   }

}

  1. This allows us to compare site state between now and when we
  2. last ran, so that we can learn to avoid doing unnecessary work.
  3. (Saves on server resources and bloated edit statistics...)
  4. Also, the data gets saved to an SVN repo, so that multiple bots can
  5. use the history. We're using JSON, as JSON has broader
  6. cross-language support than Data::Dumper, making it easier for others
  7. to use the data.

sub cachedata { &slumber; my $dataname = shift; my $data = shift; my $filename = $options{'cachepath'} . &sanitizenamefs($dataname . ".json"); unless( exists $options{'post'} ) { # This is just a test run. Best not to overwrite our cache. &out("Skipping caching of $dataname to $filename...", 4); return; } my $outfile; unless(open $outfile, '>', $filename) { &out("Failed to open $filename for caching: $!\n", 1); return; }

print $outfile to_json($data); close $outfile; &out(scalar @$data . " members cached to $filename.\n", 5); }

  1. Return data we cached previously.

sub getcacheddata { &slumber; my $dataname = shift; my $filename = $options{'cachepath'} . &sanitizenamefs($dataname . ".json"); &out("Getting cached data for $dataname from $filename...", 4); my $infile; unless (open $infile, '<', $filename) { &out("Failed to load cached data $filename: $!\n", 5); return []; }

my $jsondata; $jsondata .= $_ while <$infile>; close $infile;

my $cacheddata = from_json($jsondata); &out(scalar @$cacheddata . " cache members retrieved from $filename\n", 5);

++$cachehits; return $cacheddata; }

  1. Report the changes between two categories.
  2. More interesting than "x added, y removed"

sub reportcatchanges { &slumber; my $category = shift; my $old = shift; my $new = shift;

my ($removed, $added) = &diffcat($old, $new);

my $out = "";

$out .= "Removed from $category:\n" if( scalar @$removed > 0 ); $out .= "$_\n" foreach (@$removed);

$out .= "Added to $category:\n" if( scalar @$added > 0 ); $out .= "$_\n" foreach (@$added);

&out($out, 2) if("" ne $out); }

  1. Pull the category data, or cached data if we're not pulling from the wiki.

sub getcategory { &slumber; my $categoryname = shift; &out("Getting category contents for $categoryname...", 4);

my $old; my $new;

if( exists $options{'cacheonly'} ) { # Return the cache data if we're not supposed to query the database. $old = []; $new = &getcacheddata($categoryname); } else { $old = &getcacheddata($categoryname); $new = [$editor->get_pages_in_category($categoryname)]; ++$categorypulls; }

# &reportcatchanges($categoryname, $old, $new);

&cachedata("$categoryname", $new);

&out(scalar @$new . " members returned for $categoryname\n", 5);

return $new; }

  1. Find if this category changed, report its contents if it has.

sub getwork { &slumber; my $categoryname = shift; &out("Getting work for $categoryname.\n",4); my $cacheddata = &getcacheddata($categoryname);

# If we're on a cache-only basis, we'll just say we have no old data, # and that our cached data is our new data. my $olddata; my $newdata; if (exists $options{'cacheonly'}) { $olddata = []; $newdata = $cacheddata; } else { $olddata = $cacheddata; $newdata = &getcategory($categoryname); }

my ($removed, $added) = &diffcat($olddata, $newdata); my $impl_count_change = scalar @$removed + @$added; my $work = []; $work = $newdata if(($impl_count_change + $task_count_change) > 0);

&out(scalar @$work . " items to process for $categoryname.\n", 3);

return @$work; }

  1. If changes occurred, the info is more important than if they didn't.

sub getloglevelfromdiff { &slumber; my $base = shift; my $first = shift; my $second = shift; return $base if((scalar @$first + scalar @$second) > 0); return $base + 1; }

  1. Find all the work items for a given language.

sub getlangwork { &slumber; my $language = shift; &out("Getting lang work for $language\n", 4); my %impl; foreach my $workitem (&getwork("Category:$language")) { $impl{$workitem} = 1; } my %omit; foreach my $workitem (&getwork("Category:$language/Omit")) { $omit{$workitem} = 1; }

# No work to do? return undef if( ( 0 == scalar keys %omit) && (0 == scalar keys %impl));

# Language metadata my $unimpcount = scalar @$alltasks - scalar keys %impl; my $targetcount = (scalar @$alltasks - scalar keys %omit);


# work to do. return {'impl' => \%impl, 'omit' => \%omit,

      		'unimp_count' => $unimpcount,

'target_count' => $targetcount }; }

  1. Commit the cache.

sub commitcache { &slumber; # First, find out if we've added any files. my $cachepath = $options{'cachepath'}; opendir(CACHEDIR, $cachepath);

my @current = readdir(CACHEDIR); close(CACHEDIR);

# We need to run svn adds if we've created any new files. # Maybe we'll use SVN::Client some day. Not right now.

my ($added, $removed) = &diffcat(\@current, \@initialcache); &out("Detected " . scalar @$added . " new cache files and " . scalar @$removed . " removed\n",&getloglevelfromdiff(2, $added, $removed));

if ((scalar @$added + scalar @$removed ) > 0) { foreach my $cachefile (@$added) { &svn('add', $cachepath . $cachefile);; }

foreach my $cachefile (@$removed) { &svn('remove', $cachepath . $cachefile);; } }

&svn('ci', '--message="ImplSearchBot run"', $cachepath); &svn('update', $cachepath); }


  1. Wrap svn commands so we can log them.

sub svn { &slumber; return if(exists $options{'nosvn'}); my @args = @_;

my $string = "system 'svn'"; $string .= ", '$_'" foreach (@args);

$string .= "\n";

&out($string, 3);

system 'svn', @args if(exists $options{'post'}); }

sub getopt { &slumber;

   #I don't care to pollute my global scope.
   my $wiki = 'rosettacode.org';
   my $username;
   my $password;
   my $verbosity = 2; # verbosity level.  0 is silent. 1 is error only. 2 is updates. 3 is process, more is noisy.
   my $post; # Is this an actual run?
   my $cachepath = "/tmp/";
   my $cacheonly; # Don't query the wiki for data. Just pull from cache.
   my $nosvn;
   my $rebuild_all;
   
   my $opt_matrix = {
       "wiki=s"	=> \$wiki,
       "username=s"	=> \$username,
       "password=s"	=> \$password,
       "verbosity=s"	=> \$verbosity,
       "post"		=> \$post,
       "cacheonly"	=> \$cacheonly,
       "nosvn"		=> \$nosvn,
       "cachepath=s"	=> \$cachepath,
       "rebuildall"	=> \$rebuild_all	};
   
   my $result = GetOptions( %$opt_matrix );
   $options{'wiki'} = $wiki;
   
   $options{'nosvn'} = $nosvn
       if defined $nosvn;
   
   $options{'username'} = $username
       if defined $username;
   
   $options{'password'} = $password
       if defined $password;
   
   $options{'post'} = $post
       if defined $post;
   
   $options{'verbosity'} = $verbosity;
   
   $cachepath .= '/'
       if('/' ne substr($cachepath, -1, 1));
   
   $options{'cachepath'} = $cachepath;
   
   $options{'cacheonly'} = $cacheonly
       if defined $cacheonly;
   
   $options{'rebuild_all'} = $rebuild_all
       if defined $rebuild_all;
   
   
   my $usage = "Usage: $0 (options)\n The available options are:\n";
   $usage .= "\t--$_\n"
       foreach (keys %$opt_matrix);
   
   $usage .= "Username and password are required if you need to pull data from the wiki.  Wiki defaults to Rosetta Code.\n";
   
   
   my $wikineeded;
   
   $wikineeded = "yes"
       unless ( exists $options{'cacheonly'} );
   
   $wikineeded = "yes"
       if ( exists $options{'post'} );
   
   if(defined $wikineeded)
   {
       unless(exists $options{'username'} and exists $options{'password'})
       {
           die $usage;
       }
   }

}

sub main { &slumber;

   unless (@ARGV) { @ARGV = @_; }
   getopt();
   
  1. Statistic tracking.
  1. Tracking for svn checkin at end.
  2. We *should* be the only ones writing to the cache path for now.
  3. Anyone else should have their own checkout of the data,
  4. and test runs are done with --cacheonly and --nosvn.
   opendir(my $cachedir, $options{'cachepath'})
       or die "Unable to open cache directory";
   @initialcache = readdir($cachedir);
   closedir($cachedir);
   
   &out(scalar @initialcache . " categories initially cached\n", 4);
   
  1. Get our editor
   $editor = &geteditor();
   
  1. Get a complete listing of the tasks.
   &out("Getting tasks\n", 3);
   $alltasks = &getcategory('Category:Programming Tasks');
   
  1. Get a complete listing of the languages.
   &out("Getting the languages.\n", 3);
   $alllanguages= &getcategory('Category:Programming Languages');
   
  1. Quick check. Did we add or lose any tasks? If so, we've got to recalc *all*
  2. of the pages. :-/
   unless(exists $options{'rebuild_all'}) {
       my $cacheddata = &getcacheddata('Category:Programming Tasks');
       my ($added, $removed) = &diffcat($alltasks, $cacheddata);
       $task_count_change = scalar @$added + scalar @$removed;
   }
   else {
       $task_count_change = 1; # It just needs to be nonzero...
   }
  1. We want the language name, not the fully-qualified wiki name.
   $_ =~ s/^Category:// foreach (@$alllanguages);
  1. Get a list of the languages for which we've already provided bodies for the related omit categories.
  2. Store it as a hash, so the lookup will be faster.
   my $omitcatcontents = &getcategory('Category:Maintenance/OmitCategoriesCreated');
   my %createdomitcategories = map {$_, 1} @$omitcatcontents;
   &out("Identifying work to do\n", 3);
   foreach my $lang (@$alllanguages) {
       my $val = &getlangwork($lang);
       next unless defined $val;
       $impldiff{$lang} = $val;
   }
   &processimplediff(\%impldiff);


   my $runtime = time() - $starttime;
   my $statsdata = "Pages edited last run: $pagesedited
Time to post all per-language updates: $runtime seconds
Category pulls: $categorypulls
Cache hits: $cachehits
"; &out("Updating stats page. Runtime ($runtime), Pages edited ($pagesedited)\n", 2);
   &postpage("User:ImplSearchBot/Stats", $statsdata, "Updating stats data", 0);


   &out("Updating bot code page\n", 4);
   open my $sourcefile, '<', $0
       or die "Finished without updating bot source page";
   my $botsource;
   $botsource .= $_ while <$sourcefile>;
   close $sourcefile;
   my $tag = "lang";
   &postpage("User:ImplSearchBot/Code", "<$tag perl>$botsource</$tag>", 0);
my $logdata = "
" . join( "\n", @logoutput) . "
";
   &postpage("User:ImplSearchBot/Log", $logdata, 0);
   &out("Updating cache\n", 4);
   &commitcache();
   &out("Done\n", 3);
   exit 0

}

if (!caller) { exit main } </lang>