#! /usr/bin/perl
-# $Cambridge: exim/doc/doc-docbook/Tidytxt,v 1.1 2005/06/16 10:32:31 ph10 Exp $
+use strict;
+use warnings;
+use Getopt::Long;
-# Script to tidy up the output of w3m when it makes a text file. We convert
-# sequences of blank lines into a single blank line.
-$blanks = 0;
-while (<>)
+# For now we can't rely on a perl >= 5.14 on
+# the build sites, thus we throw away all unicode
+# awareness and do the matching byte by byte
+binmode STDIN;
+binmode STDOUT;
+
+GetOptions(
+ 'u|utf8!' => \my $want_utf8, # do not replace unicode characters
+) or die "Usage: $0 [-u|--utf8]\n";
+
+# Script to tidy up the output of w3m when it makes a text file. First we
+# convert sequences of blank lines into a single blank line, to get everything
+# uniform. Then we go through and insert blank lines before chapter and
+# sections, also converting chapter titles to uppercase.
+
+# We also have to do some character translation in the first pass. It seems
+# that xmlto now generates Unicode in its HTML pages. This gives three problems:
+# (1) It inserts the byte sequence C2 A0 (U+00A0) as a fixed-width space;
+# (2) It uses U+25CF as its bullet character.
+# (3) It inserts a whole slew of "box drawing" characters round the heading.
+
+my @lines = <>;
+my $lastwasblank = 0;
+
+foreach my $line (@lines)
{
- if (/^\s*$/)
+ # (1) non-break space -> normal space
+ $line =~ s/\x{c2}\x{a0}/ /g;
+
+ unless ($want_utf8)
{
- $blanks++;
+ # (2) bullet -> asterisk
+ $line =~ s/\x{e2}\x{97}\x{8f}/*/g;
+ $line =~ s/\x{e2}\x{80}\x{a2}/*/g; # OpenSUSE
+ $line =~ s/\x{e2}\x{96}\x{a1}/*/g; # OpenSUSE
+ # (3a) horizontal box drawing -> hyphen
+ $line =~ s/\x{e2}\x{94}[\x{80}\x{81}\x{84}\x{85}\x{88}\x{89}]/-/g;
+ $line =~ s/\x{e2}\x{95}[\x{8c}\x{8d}\x{90}]/-/g;
+ $line =~ s/\x{e2}\x{95}[\x{b4}\x{b6}\x{b8}\x{ba}\x{bc}\x{be}]/-/g;
+ # (3b) vertical box drawing -> bar
+ $line =~ s/\x{e2}\x{94}[\x{82}\x{83}\x{86}\x{87}\x{8a}\x{8b}]/|/g;
+ $line =~ s/\x{e2}\x{95}[\x{8e}\x{8f}\x{91}]/|/g;
+ $line =~ s/\x{e2}\x{95}[\x{b5}\x{b7}\x{b9}\x{bb}\x{bd}\x{bf}]/|/g;
+ # (3c) corner box drawing -> plus
+ $line =~ s/\x{e2}\x{94}[\x{8c}-\x{bf}]/+/g;
+ $line =~ s/\x{e2}\x{95}[\x{80}-\x{8b}\x{92}-\x{b0}]/+/g;
+ # other
+ $line =~ s/\x{e2}\x{95}\x{b1}/\//g;
+ $line =~ s/\x{e2}\x{95}\x{b2}/\\/g;
+ $line =~ s/\x{e2}\x{95}\x{b3}/X/g;
+ }
+
+ # w3m rendering issue apparently only seen by pdp
+ # affects section numbers after the ToC, some info on spool-file -lines, etc
+ # always appears to be a spurious extra character, safely just dropped.
+ $line =~ s/\x{c2}//g;
+
+ if ($line =~ /^\s*$/)
+ {
+ $line = "" if $lastwasblank;
+ $lastwasblank = 1;
next;
}
- print "\n" if $blanks > 0;
- $blanks = 0;
- print;
+ $lastwasblank = 0;
}
+# Find start of TOC, uppercasing its title
+
+my $i = 0;
+for ($i = 0; $i < scalar @lines; $i++)
+ {
+ $lines[$i] = "TABLE OF CONTENTS\n" if $lines[$i] =~ /^Table of Contents/;
+ last if $lines[$i] =~ /^1\. /;
+ }
+
+# Find start of first chapter
+
+for ($i++; $i < scalar @lines; $i++)
+ { last if $lines[$i] =~ /^1\. /; }
+
+# Process the body. We can detect the starts of chapters and sections by
+# looking for preceding and following blank lines, and then matching against
+# the numbers.
+
+my $chapter = 0;
+my $section;
+for (; $i < scalar @lines; $i++)
+ {
+ next if $lines[$i-1] !~ /^$/ || $lines[$i+1] !~ /^$/;
+
+ # Start of chapter
+
+ if ($lines[$i] =~ /^(\d+)\. / && $1 == $chapter + 1)
+ {
+ $chapter++;
+ $section = 0;
+ $lines[$i] = "\n\n" . ("=" x 79) . "\n" . uc($lines[$i]);
+ }
+
+ # Start of next section
+
+ elsif ($lines[$i] =~ /^(\d+)\.(\d+) / && $1 == $chapter && $2 == $section + 1)
+ {
+ $section++;
+ $lines[$i] = "\n$lines[$i]" . "-" x (length($lines[$i]) - 1) . "\n";
+ }
+ }
+
+print @lines;
+
# End
git://git.exim.org / users / jgh / exim.git / blobdiff