#!/usr/bonsaitools/bin/perl
# $Id: Local.pm,v 1.5 1999-04-30 19:00:47 endico%mozilla.org Exp $
# Local.pm -- Subroutines that need to be customized for each installation
#
#	Dawn Endico <endico@mozilla.org>
#
######################################################################
# This package is for placing subroutines that are likely to need
# to be customized for each installation. In particular, the file
# and directory description snarfing mechanism is likely to be
# different for each project.

package Local;

require Exporter;
@ISA = qw(Exporter);
@EXPORT = qw(&fdescexpand &descexpand &dirdesc &convertwhitespace );

use lib 'lib/';
use LXR::Common;

# dme: Create descriptions for a file in a directory listing
# If no description, return the string "\&nbsp\;" to keep the
# table looking pretty.
#
# In mozilla search the beginning of a source file for a short 
# description. Not all files have them and the ones that do use 
# many different formats. Try to find as many of these without
# printing gobbeldygook or something silly like a file name or a date.
#
# Read in the beginning of the file into a string. I chose 60 because the 
# Berkeley copyright notice is around 40 lines long so we need a bit more 
# than this.
#
# Its common for file descriptions to be delimited by the file name or
# the word "Description" which preceeds the description. Search the entire
# string for these. Sometimes they're put in odd places such as inside
# the copyright notice or after the code begins. The file name should be
# followed by a colon or some pattern of dashes. 
#
# If no such description is found then use the contents of the "first"
# comment as the description. First, strip off the copyright notice plus
# anything before it. Remove rcs comments. Search for the first bit of
# code (usually #include) and remove it plus anything after it. In what's
# left, find the contents of the first comment, and get the first paragraph.
# If that's too long, use only the first sentence up to a period. If that's 
# still too long then we probably have a list or something that will look 
# strange if we print it out so give up and return null.
#
# Yes, this is a lot of trouble to go through but its easier than getting
# people to use the same format and re-writing thousands of comments. Not
# everything printed will really be a summary of the file, but still the
# signal/noise ratio seems pretty high.
#
# Yea, though I walk through the valley of the shadow of pattern
# matching, I shall fear no regex.
sub fdescexpand {
    # use global vars here because the expandtemplate subroutine makes
    # passing parameters impossible. Use $filename from source and
    # $Path from Common.pm
    my $filename = $main::filename;
    my $linecount=0;
    my $copy= "";
    local $desc= "";
    my $maxlines = 40; #only look at the beginning of the file

    #ignore files that aren't source code
    if (!(
	    ($filename =~ /\.c$/) |
	    ($filename =~ /\.h$/) | 
	    ($filename =~ /\.cc$/) |
	    ($filename =~ /\.cp$/) | 
	    ($filename =~ /\.cpp$/) | 
	    ($filename =~ /\.idl$/) | 
	    ($filename =~ /\.java$/)
	    )){
	return("\&nbsp\;");
	}

    if (open(FILE, $Path->{'real'}."/".$filename)) {
        while(<FILE>){
	    $desc = $desc . $_ ;
	    if($linecount++ > 60) {
		last;
	    }
	}
	close(FILE);
    } 

    # sanity check: if there's no description then stop
    if (!($desc =~ /\w/)){
	return("\&nbsp\;");;
	}

    # save a copy for later
    $copy = $desc;

    # Look for well behaved <filename><seperator> formatted 
    # descriptions before we go to the trouble of looking for
    # one in the first comment. The whitespace between the 
    # delimeter and the description may include a newline.
    if (($desc =~ s/(?:.*?$filename\s*?- ?-*\s*)([^\n]*)(?:.*)/$1/sgi) || 
        ($desc =~ s/(?:.*?$filename\s*?:\s*)([^\n]*)(?:.*)/$1/sgi) ||
        ($desc =~ s/(?:.*?Description:\s*)([^\n]*)(?:.*)/$1/sgi) 
	){
        # if the description is non-empty then clean it up and return it
        if ($desc =~ /\w/) {
            #strip trailing asterisks and "*/"
            $desc =~ s#\*/?\s*$##;
            $desc =~ s#^[^\S]*\**[^\S]*#\n#gs;

            # Strip beginning and trailing whitespace
            $desc =~ s/^\s+//;
            $desc =~ s/\s+$//;

            # Strip junk from the beginning
            $desc =~ s#[^\w]*##ms;

            #htmlify the comments making links to symbols and files
            $desc = markupstring($desc, $Path->{'virt'});
            return($desc);
           } 
    }

    # we didn't find any well behaved descriptions above so start over 
    # and look for one in the first comment
    $desc = $copy;

    # Strip off code from the end, starting at the first cpp directive
    $desc =~ s/\n#.*//s;

    # Strip off code from the end, starting at typedef
    $desc =~ s/\ntypedef.*//s;

    # Strip off license
    $desc =~ s#(?:/\*.*license.*?\*/)(.*)#$1#is;

    # Strip off copyright notice
    $desc =~ s#(?:/\*.*copyright.*?\*/)(.*)#$1#is;

    # Strip off emacs line
    $desc =~ s#(/\*.*tab-width.*?\*/)(.*)#$2#isg;

    # excise rcs crud
    $desc =~ s#Id: $filename.*?Exp \$##g;

    # Yuck, nuke these silly comments in js/jsj /* ** */
    $desc =~ s#\n\s*/\*+[\s\*]+\*/\n#\n#sg;

    # Don't bother to continue if there aren't any comments here
    if(!($desc =~ m#/\*#)) {
	return("&nbsp;");
    }

    # Remove lines generated by jmc
    $desc =~ s#\n.*?Source date:.*\n#\n#;
    $desc =~ s#\n.*?Generated by jmc.*\n#\n#;

    # Extract the first comment
    $desc =~ s#(?:.*?/\*+)(.*?)(?:(?:\*+/.*)|(?:$))#$1#s;

    # Strip silly borders
    $desc =~ s#\n\s*[\*\=\-\s]+#\n#sg;

    # Strip beginning and trailing whitespace
    $desc =~ s/^\s+//;
    $desc =~ s/\s+$//;

    # Strip out file name
    $desc =~ s#$filename##i;

    # Strip By line
    $desc =~ s#By [^\n]*##;

    # Strip out dates
    $desc =~ s#\d{1,2}/\d{1,2}/\d\d\d\d##;
    $desc =~ s#\d{1,2}/\d{1,2}/\d\d##;
    $desc =~ s#\d{1,2} \w\w\w \d\d\d\d##;

    # Strip junk from the beginning
    $desc =~ s#[^\w]*##;

    # Extract the first paragraph
    $desc =~ s#(\n\s*?\n.*)##s;

    # If the description is too long then just use the first sentence
    # this will fail if no period was used.
    if (length($desc) > 200 ) {
        $desc =~ s#([^\.]+\.)\s.*#$1#s;
    }

    # If the description is still too long then assume it will look
    # like gobbeldygook and give up
    if (length($desc) > 200 ) {
	return("&nbsp;");
    }

    # htmlify the comments, making links to symbols and files
    $desc = markupstring($desc, $Path->{'virt'});

    if ($desc) {
	return($desc);
    }else{
	return("\&nbsp\;");
    }
}


# dme: create a short description for a subdirectory in a directory listing
# If no description, return the string "\&nbsp\;" to keep the
# table looking pretty.
#
# In Mozilla, if the directory has a README file look in it for lines 
# like the ones used in source code: "directoryname --- A short description"
sub descexpand {
    # use global vars here because the expandtemplate subroutine makes
    # passing parameters impossible. Use $filename from source and
    # $Path from Common.pm
    my $filename = $main::filename;
    my $linecount=0;
    local $desc= "";

    if (open(DESC, $Path->{'real'}. $filename."/README.html")) {
        undef $/;
        $desc = <DESC>;
        $/ = "\n";
        close(DESC);

        # Make sure there is no <span> embedded in our string. If so 
        # then we've matched against the wrong /span and this string is junk
        # so we'll throw it away and refrain from writing a descrioption.
        # Disallowing embedded spans theoretically removes some flexibility
        # but this seems to be a little used tag and doing this makes lxr 
        # a lot faster.
        if ($desc =~ /<SPAN CLASS=LXRSHORTDESC>(.*?)<\/SPAN>/is) {
            $short = $1;
            if (!($short =~ /\<span/is)) {
                return ($short);
            }
        }
    }

    $desc = ""; 
    if (open(FILE, $Path->{'real'}. $filename."README")) {
	$path = $Path->{'virt'}.$filename;
	$path =~ s#/(.+)/#$1#;
        while(<FILE>){
            if($linecount++ > 10) {
                last;
            }elsif (/\s*$path\s*-\s*-*\s*/i){
                $desc = (split(/\s*$path\s*-\s*-*\s*/i))[1];
                if ($desc) {last};
            }elsif (/\s*$filename\s*-\s*-*\s*/i){
                $desc = (split(/\s*$filename\s*-\s*-*\s*/i))[1];
                if ($desc) {last};
            }elsif (/$path\s*:\s*/i){
                $desc = (split(/ $path\s*:\s*/i))[1];
                if ($desc) {last};
            }elsif (/$filename\s*:\s*/i){
                $desc = (split(/ $filename\s*:\s*/i))[1];
                if ($desc) {last};
            }
        }
        close(FILE);
    }

    #strip trailing asterisks and "*/"
    $desc =~ s#\*/?\s*$##;

    if ($desc){
        #htmlify the comments making links to symbols and files
        $desc = markupstring($desc, $Path->{'virt'});

        return($desc);
    } else {
        return("\&nbsp\;");
    }
}

# dme: Print a descriptive blurb in directory listings between 
# the document heading and the table containing the actual listing.
#
# For Mozilla, we extract this information from the README file if
# it exists. If the file is short then just print the whole thing.
# For longer files print the first paragraph or so. As much as 
# possible make this work for randomly formatted files rather than 
# inventing strict rules which create gobbeldygook when they're broken.
sub dirdesc {
    my ($path) = @_;

    if (-f $Path->{'real'}."/README") {
	    descreadme($path);
    } elsif (-f $Path->{'real'}."/README.html") {
	    descreadmehtml($path);
    }
}


sub descreadmehtml {
    my ($path) = @_;

    my $string = ""; 

    if (!(open(DESC, $Path->{'real'}."/README.html"))) {
	return;
        }
    undef $/;
    $string = <DESC>;
    $/ = "\n";
    close(DESC);

    # if the README is 0 length then give up
    if (!$string) {
        return;
    }

    # check if there's a short desc nested inside the long desc. If not, do
    # a non-greedy search for a long desc. assume there are no other stray
    # spans within the description.
    if ($string =~ /<SPAN CLASS=LXRLONGDESC>(.*?<SPAN CLASS=LXRSHORTDESC>.*?<\/SPAN>.*?)<\/SPAN>/is) {
        $long = $1;
        if (!($long =~ /<span.*?\<span/is)) {
            print($long . "<P>\nSEE ALSO: <A HREF=\"README.html\">README</A>\n");
        }
    } elsif ($string =~ /<SPAN CLASS=LXRLONGDESC>(.*?)<\/SPAN>/is) {
        $long = $1;
        if (!($long =~ /\<span/is)) {
            print($long . "<P>\nSEE ALSO: <A HREF=\"README.html\">README</A>\n");
        }
    }
}

sub descreadme {
    my ($path) = @_;

    my $string = ""; 
#    $string =~ s#(</?([^>^\s]+[^>]*)>.*$)#($2~/B|A|IMG|FONT|BR|EM|I|TT/i)?$1:""#sg;
    my $n; 
    my $count;
    my $temp;

    my $maxlines = 20;  # If file is less than this then just print it all
    my $minlines = 5;   # Too small. Go back and add another paragraph.
    my $chopto = 10;    # Truncate long READMEs to this length

    if (!(open(DESC, $Path->{'real'}."/README"))) {
	return;
        }

    undef $/;
    $string = <DESC>;
    $/ = "\n";
    close(DESC);

    # if the README is 0 length then give up
    if (!$string){
	return;
    }
    # strip the emacs tab line
    $string =~ s/.*tab-width:[ \t]*([0-9]+).*\n//;

    # strip the npl
    $string =~ s/.*The contents of this .* All Rights.*Reserved\.//s;

    # strip the short description from the beginning
    $path =~ s#/(.+)/#$1#;
    $string =~ s/.*$path\/*\s+--- .*//;

    # strip away junk
    $string =~ s/#+\s*\n/\n/;
    $string =~ s/---+\s*\n/\n/g;
    $string =~ s/===+\s*\n/\n/g;

    # strip blank lines at beginning and end of file.
    $string =~ s/^\s*\n//gs;
    $string =~ s/\s*\n$//gs;
    chomp($string);
    $_ = $string;
    $count = tr/\n//;

    # If the file is small there's not much use splitting it up.
    # Just print it all
    if ($count <= $maxlines) {
        $string = markupstring($string, $Path->{'virt'});
	$string = convertwhitespace($string);
        print($string);
    } else {
        # grab the first n paragraphs, with n decreasing until the
        # string is 10 lines or shorter or until we're down to 
	# one paragraph.
	$n = 6;
	$temp = $string;
	while ( ($count > $chopto) && ($n-- > 1) ) {
            $string =~ s/^((?:(?:[\S\t ]*?\n)+?[\t ]*\n){$n}?)(.*)/$1/s;
	    $_ = $string;
	    $string =~ s/\s*\n$//gs;
	    $count = tr/\n//;
	}

	# if we have too few lines then back up and grab another paragraph
	$_ = $string;
	$count = tr/\n//;
	if ($count < $minlines) {
	    $n = $n+1;
	    $temp =~ s/^((?:(?:[\S\t ]*?\n)+?[\t ]*\n){$n}?)(.*)/$1/s;
	    $string = $temp;
	}

	# if we have more than $maxlines then truncate to $chopto
	# and add an elipsis. 
	if ($count > $maxlines) {
	    $string =~ s/^((?:[\S \t]*\n){$chopto}?)(.*)/$1/s;
	    chomp($string);
	    $string = $string . "\n...";
	} 
	
        # since not all of the README is displayed here,
        # add a link to it.
        chomp($string);
        if ($string =~ /SEE ALSO/) {
            $string = $string . ", README";
        } else {
            $string = $string . "\n\nSEE ALSO: README";
        } 

        $string = markupstring($string, $Path->{'virt'});
	$string = convertwhitespace($string);

        # strip blank lines at beginning and end of file again
        $string =~ s/^\s*\n//gs;
        $string =~ s/\s*\n$//gs;
        chomp($string);

        print($string . "<P>\n");
    }
}

# dme: substitute carraige returns and spaces in original text
# for html equivalent so we don't need to use <pre> and can
# use variable width fonts but preserve the formatting
sub convertwhitespace {
    my ($string) = @_;

    # handle ascii bulleted lists
    $string =~ s/<p>\n\s+o\s/<p>\n\&nbsp\;\&nbsp\;o /sg;
    $string =~ s/\n\s+o\s/&nbsp\;\n<br>\&nbsp\;\&nbsp\;o /sg;

    #find paragraph breaks and replace with <P>
    $string =~ s/\n\s*\n/<p>\n/sg;

    return($string);
}