cvs2cl.pl

java写的html的解析器parser

        my $tag_name = $1;
        my $tag_rev  = $2;

        # A branch number either has an odd number of digit sections
        # (and hence an even number of dots), or has ".0." as the
        # second-to-last digit section.  Test for these conditions.
        my $real_branch_rev = "";
        if (($tag_rev =~ /^(\d+\.\d+\.)+\d+$/)   # Even number of dots...
            and (! ($tag_rev =~ /^(1\.)+1$/)))   # ...but not "1.[1.]1"
        {
          $real_branch_rev = $tag_rev;
        }
        elsif ($tag_rev =~ /(\d+\.(\d+\.)+)0.(\d+)/)  # Has ".0."
        {
          $real_branch_rev = $1 . $3;
        }
        # If we got a branch, record its number.
        if ($real_branch_rev)
        {
          $branch_names{$real_branch_rev} = $tag_name;
          if (@Follow_Branches) {
            if (grep ($_ eq $tag_name, @Follow_Branches)) {
              $branch_numbers{$tag_name} = $real_branch_rev;
            }
          }
        }
        else {
          # Else it's just a regular (non-branch) tag.
          push (@{$symbolic_names{$tag_rev}}, $tag_name);
        }
      }
    }
    # End of code for collecting tag names.

    # If have file name, but not revision, and see revision, then grab
    # it.  (We collect unconditionally, even though we may or may not
    # ever use it.)
    if ((! (defined $revision)) and (/^revision (\d+\.[\d.]+)/))
    {
      $revision = $1;

      if (@Follow_Branches)
      {
        foreach my $branch (@Follow_Branches)
        {
          # Special case for following trunk revisions
          if (($branch =~ /^trunk$/i) and ($revision =~ /^[0-9]+\.[0-9]+$/))
          {
            goto dengo;
          }

          my $branch_number = $branch_numbers{$branch};
          if ($branch_number)
          {
            # Are we on one of the follow branches or an ancestor of
            # same?
            #
            # If this revision is a prefix of the branch number, or
            # possibly is less in the minormost number, OR if this
            # branch number is a prefix of the revision, then yes.
            # Otherwise, no.
            #
            # So below, we determine if any of those conditions are
            # met.

            # Trivial case: is this revision on the branch?
            # (Compare this way to avoid regexps that screw up Emacs
            # indentation, argh.)
            if ((substr ($revision, 0, ((length ($branch_number)) + 1)))
                eq ($branch_number . "."))
            {
              goto dengo;
            }
            # Non-trivial case: check if rev is ancestral to branch
            elsif ((length ($branch_number)) > (length ($revision))
                   and
                   $No_Ancestors)
            {
              $revision =~ /^((?:\d+\.)+)(\d+)$/;
              my $r_left = $1;          # still has the trailing "."
              my $r_end = $2;

              $branch_number =~ /^((?:\d+\.)+)(\d+)\.\d+$/;
              my $b_left = $1;  # still has trailing "."
              my $b_mid  = $2;   # has no trailing "."

              if (($r_left eq $b_left)
                  && ($r_end <= $b_mid))
              {
                goto dengo;
              }
            }
          }
        }
      }
      else    # (! @Follow_Branches)
      {
        next;
      }

      # Else we are following branches, but this revision isn't on the
      # path.  So skip it.
      undef $revision;
    dengo:
      next;
    }

    # If we don't have a revision right now, we couldn't possibly
    # be looking at anything useful.
    if (! (defined ($revision))) {
      $detected_file_separator = /^$file_separator$/o;
      if ($detected_file_separator) {
        # No revisions for this file; can happen, e.g. "cvs log -d DATE"
        goto CLEAR;
      }
      else {
        next;
      }
    }

    # If have file name but not date and author, and see date or
    # author, then grab them:
    unless (defined $time)
    {
      if (/^date: .*/)
      {
        ($time, $author, $state, $lines) =
          &parse_date_author_and_state ($_);
        if (defined ($usermap{$author}) and $usermap{$author}) {
          $author = $usermap{$author};
        } elsif($Domain ne "" or $Gecos == 1) {
          my $email = $author;
          if($Domain ne "") {
            $email = $author."@".$Domain;
          }
          my $pw = getpwnam($author);
          my $fullname;
          my $office;
          my $workphone;
          my $homephone;
          for (($fullname, $office, $workphone, $homephone) = split /\s*,\s*/, $pw->gecos) {
            s/&/ucfirst(lc($pw->name))/ge;
          }
          if($fullname ne "") {
            $author = $fullname . "  <" . $email . ">";
          }
        }
      }
      else {
        $detected_file_separator = /^$file_separator$/o;
        if ($detected_file_separator) {
          # No revisions for this file; can happen, e.g. "cvs log -d DATE"
          goto CLEAR;
        }
      }
      # If the date/time/author hasn't been found yet, we couldn't
      # possibly care about anything we see.  So skip:
      next;
    }

    # A "branches: ..." line here indicates that one or more branches
    # are rooted at this revision.  If we're showing branches, then we
    # want to show that fact as well, so we collect all the branches
    # that this is the latest ancestor of and store them in
    # @branch_roots.  Just for reference, the format of the line we're
    # seeing at this point is:
    #
    #    branches:  1.5.2;  1.5.4;  ...;
    #
    # Okay, here goes:

    if (/^branches:\s+(.*);$/)
    {
      if ($Show_Branches)
      {
        my $lst = $1;
        $lst =~ s/(1\.)+1;|(1\.)+1$//;  # ignore the trivial branch 1.1.1
        if ($lst) {
          @branch_roots = split (/;\s+/, $lst);
        }
        else {
          undef @branch_roots;
        }
        next;
      }
      else
      {
        # Ugh.  This really bothers me.  Suppose we see a log entry
        # like this:
        #
        #    ----------------------------
        #    revision 1.1
        #    date: 1999/10/17 03:07:38;  author: jrandom;  state: Exp;
        #    branches:  1.1.2;
        #    Intended first line of log message begins here.
        #    ----------------------------
        #
        # The question is, how we can tell the difference between that
        # log message and a *two*-line log message whose first line is
        #
        #    "branches:  1.1.2;"
        #
        # See the problem?  The output of "cvs log" is inherently
        # ambiguous.
        #
        # For now, we punt: we liberally assume that people don't
        # write log messages like that, and just toss a "branches:"
        # line if we see it but are not showing branches.  I hope no
        # one ever loses real log data because of this.
        next;
      }
    }

    # If have file name, time, and author, then we're just grabbing
    # log message texts:
    $detected_file_separator = /^$file_separator$/o;
    if ($detected_file_separator && ! (defined $revision)) {
      # No revisions for this file; can happen, e.g. "cvs log -d DATE"
      goto CLEAR;
    }
    unless ($detected_file_separator || /^$logmsg_separator$/o)
    {
      $msg_txt .= $_;   # Normally, just accumulate the message...
      next;
    }
    # ... until a msg separator is encountered:
    # Ensure the message contains something:
    if ((! $msg_txt)
        || ($msg_txt =~ /^\s*\.\s*$|^\s*$/)
        || ($msg_txt =~ /\*\*\* empty log message \*\*\*/))
    {
      if ($Prune_Empty_Msgs) {
        goto CLEAR;
      }
      # else
      $msg_txt = "[no log message]\n";
    }

    ### Store it all in the Grand Poobah:
    {
      my $dir_key;        # key into %grand_poobah
      my %qunk;           # complicated little jobbie, see below

      # Each revision of a file has a little data structure (a `qunk')
      # associated with it.  That data structure holds not only the
      # file's name, but any additional information about the file
      # that might be needed in the output, such as the revision
      # number, tags, branches, etc.  The reason to have these things
      # arranged in a data structure, instead of just appending them
      # textually to the file's name, is that we may want to do a
      # little rearranging later as we write the output.  For example,
      # all the files on a given tag/branch will go together, followed
      # by the tag in parentheses (so trunk or otherwise non-tagged
      # files would go at the end of the file list for a given log
      # message).  This rearrangement is a lot easier to do if we
      # don't have to reparse the text.
      #
      # A qunk looks like this:
      #
      #   {
      #     filename    =>    "hello.c",
      #     revision    =>    "1.4.3.2",
      #     time        =>    a timegm() return value (moment of commit)
      #     tags        =>    [ "tag1", "tag2", ... ],
      #     branch      =>    "branchname" # There should be only one, right?
      #     branchroots =>    [ "branchtag1", "branchtag2", ... ]
      #   }

      if ($Distributed) {
        # Just the basename, don't include the path.
        ($qunk{'filename'}, $dir_key, undef) = fileparse ($file_full_path);
      }
      else {
        $dir_key = "./";
        $qunk{'filename'} = $file_full_path;
      }

      # This may someday be used in a more sophisticated calculation
      # of what other files are involved in this commit.  For now, we
      # don't use it much except for delta mode, because the
      # common-commit-detection algorithm is hypothesized to be
      # "good enough" as it stands.
      $qunk{'time'} = $time;

      # We might be including revision numbers and/or tags and/or
      # branch names in the output.  Most of the code from here to
      # loop-end deals with organizing these in qunk.

      $qunk{'revision'} = $revision;
      $qunk{'state'} = $state;
      if ( defined( $lines )) {
        $qunk{'lines'} = $lines;
      }

      # Grab the branch, even though we may or may not need it:
      $qunk{'revision'} =~ /((?:\d+\.)+)\d+/;
      my $branch_prefix = $1;
      $branch_prefix =~ s/\.$//;  # strip off final dot
      if ($branch_names{$branch_prefix}) {
        $qunk{'branch'} = $branch_names{$branch_prefix};
      }

      # Keep a record of the file's cvs state.
      $qunk{'cvsstate'} = $state;

      # If there's anything in the @branch_roots array, then this
      # revision is the root of at least one branch.  We'll display
      # them as branch names instead of revision numbers, the
      # substitution for which is done directly in the array:
      if (@branch_roots) {
        my @roots = map { $branch_names{$_} } @branch_roots;
        $qunk{'branchroots'} = \@roots;
      }

      # Save tags too.
      if (defined ($symbolic_names{$revision})) {
        $qunk{'tags'} = $symbolic_names{$revision};
        delete $symbolic_names{$revision};

	# If we're in 'delta' mode, update the latest observed
	# times for the beginning and ending tags, and
	# when we get around to printing output, we will simply restrict
	# ourselves to that timeframe...
	
	if ($Delta_Mode) {
	  if (($time > $Delta_StartTime) &&
	      (grep { $_ eq $Delta_From } @{$qunk{'tags'}}))
	  {
	    $Delta_StartTime = $time;
	  }
	
	  if (($time > $Delta_EndTime) &&
	      (grep { $_ eq $Delta_To } @{$qunk{'tags'}}))
	  {
	    $Delta_EndTime = $time;
	  }
	}
      }

      # Add this file to the list
      # (We use many spoonfuls of autovivication magic. Hashes and arrays
      # will spring into existence if they aren't there already.)

      &debug ("(pushing log msg for ${dir_key}$qunk{'filename'})\n");

      # Store with the files in this commit.  Later we'll loop through
      # again, making sure that revisions with the same log message
      # and nearby commit times are grouped together as one commit.
      push (@{$grand_poobah{$dir_key}{$author}{$time}{$msg_txt}}, \%qunk);
    }

  CLEAR:
    # Make way for the next message
    undef $msg_txt;
    undef $time;
    undef $revision;
    undef $author;
    undef @branch_roots;

    # Maybe even make way for the next file:
    if ($detected_file_separator) {
      undef $file_full_path;
      undef %branch_names;
      undef %branch_numbers;
      undef %symbolic_names;
    }
  }

  close (LOG_SOURCE);

  ### Process each ChangeLog

  while (my ($dir,$authorhash) = each %grand_poobah)
  {
    &debug ("DOING DIR: $dir\n");

    # Here we twist our hash around, from being
    #   author => time => message => filelist
    # in %$authorhash to
    #   time => author => message => filelist
    # in %changelog.
    #
    # This is also where we merge entries.  The algorithm proceeds
    # through the timeline of the changelog with a sliding window of
    # $Max_Checkin_Duration seconds; within that window, entries that
    # have the same log message are merged.
    #
    # (To save space, we zap %$authorhash after we've copied
    # everything out of it.)

    my %changelog;
    while (my ($author,$timehash) = each %$authorhash)
    {
      my $lasttime;
      my %stamptime;
      foreach my $time (sort {$main::a <=> $main::b} (keys %$timehash))
      {
        my $msghash = $timehash->{$time};
        while (my ($msg,$qunklist) = each %$msghash)
        {
 	  my $stamptime = $stamptime{$msg};
          if ((defined $stamptime)
              and (($time - $stamptime) < $Max_Checkin_Duration)
              and (defined $changelog{$stamptime}{$author}{$msg}))
          {
 	    push(@{$changelog{$stamptime}{$author}{$msg}}, @$qunklist);
          }
          else {
            $changelog{$time}{$author}{$msg} = $qunklist;
            $stamptime{$msg} = $time;
          }
        }
      }
    }
    undef (%$authorhash);

    ### Now we can write out the ChangeLog!

    my ($logfile_here, $logfile_bak, $tmpfile);

    if (! $Output_To_Stdout) {
      $logfile_here =  $dir . $Log_File_Name;
      $logfile_here =~ s/^\.\/\//\//;   # fix any leading ".//" problem
      $tmpfile      = "${logfile_here}.cvs2cl$$.tmp";
      $logfile_bak  = "${logfile_here}.bak";