# Configuration of latexmk so that it works with the hyperxmp package.

# Configuration of latexmk so that it works with the hyperxmp package.
# Hyperxmp puts document xmp metadata in the pdf file, and one of the
# fields in that metadata is byteCount, which is the size of the pdf file.
# This configuration ensures that the byteCount is present and correctly
# equals the pdf file's size.
#
# Method of operation:
# For any command that might produce a pdf file, prepend "internal mycmd3 "
# to the previously configured command, so that a subroutine mycmd3 is
# called. It runs the expected command and then does post-processing if
# the output file is a pdf file.

foreach my $cmd ( "latex", "lualatex", "pdflatex", "xelatex",
"dvipdf", "xdvipdfmx", "ps2pdf" ) {
${$cmd} = "internal mycmd ${$cmd}";
}

#======================================================

sub mycmd {
# Run command, and do extra processing of output file, if it is a pdf
# file.
# This subroutine is run in a rule context, so that $rule is the name of
# the rule and $$Pdest is the destination file name.

# Identification string for messages:
local $name = "$My_name.mycmd";

my $retval = system @_;

if ( $$Pdest =~ /\.pdf$/ ) {
fix_pdf( $$Pdest );
}
return $retval;
}

#======================================================

sub fix_pdf {
# Change/insert byteCount field with correct file length, while preserving
# the file size and the length of the stream containing xmp metadata.
# Return 1 on success, else 0.

local $pdf_name = shift;
local $tmp_name = "$pdf_name.new.pdf";
local $pdf_size = (stat($pdf_name))[7];
warn "Inserting/correcting byteCount field in '$pdf_name' ...\n";

# Strings surrounding (and identifying) the byteCount field, and other
# parts of the xmp packet:
local $xmp_start = '<x:xmpmeta xmlns:x="adobe:ns:meta/">';
local $decl_bC = '<pdfaProperty:name>byteCount</pdfaProperty:name>';
local $pre_bC = '<prism:byteCount>';
local $post_bC = '</prism:byteCount>';
local $pC = '<prism:pageCount>';
local $rd_end = '</rdf:Description>';
local $xmp_end = '</x:xmpmeta>';

local *PDF;
local *TMP;

if (! open PDF, "<", $pdf_name ) {
warn " Cannot read '$pdf_name'\n";
return 0;
}
if ( ! open TMP, ">", $tmp_name ) {
warn " Cannot write temporary file '$tmp_name'\n";
close PDF;
return 0;
}
local $status = 0; # 0 = no XMP packet, 1 = success, >= errors
while ( <PDF> ) {
# Only examine first XMP packet:
if ( ($status == 0) && /^\s*\Q$xmp_start\E/ ) {
local @xmp = $_;
local $len_padding = 0;
local $xmp_after_line = '';
&xmp_get_mod;
print TMP @xmp;
# Insert correct padding to leave file size unchanged:
while ( $len_padding > 0 ) {
my $len_line = 64;
if ( $len_line > $len_padding ) { $len_line = $len_padding; }
$len_padding -= $len_line;
print TMP (' ' x ($len_line - 1) ), "\n";
}
print TMP $xmp_after_line;
$xmp_after_line = '';
}
else {
print TMP "$_";
}
}
close PDF;
close TMP;

if ($status == 0) {
warn " Could not insert/modify byteCount, since no XMP packet was found.\n";
warn " So '$pdf_name' is unchanged,\n",
" and I will delete temporary file '$tmp_name'.\n";
unlink $tmp_name;
} elsif ($status == 1) {
rename $tmp_name, $pdf_name
or die " Cannot move temporary file '$tmp_name' to '$pdf_name'.\n",
" Error is '$!'\n";
} else {
warn " Could not insert correct byteCount. See above for reason.\n";
warn " So '$pdf_name' is unchanged,\n",
" and I will delete temporary file '$tmp_name'.\n";
unlink $tmp_name;
}
return ($status == 1);
}

#======================================================

sub xmp_get_mod {
# Get xmp packet, given that @xmp contains its first line.
# Get amount of trailing padding, and line after that.
# If possible, insert a byteCount field:
# Either replace existing specification, if it exists,
# or insert one in expected place for hyperxmp, if the XMP packet
# matches what hyperxmp would produce.
# Return xmp packet in @xmp, amount of padding needed in $len_padding,
# line after that in $xmp_after_line, and error code in $error.
# Set $status appropriately: 1 for success; >=1 for failure.

$len_padding = 0;
$xmp_after_line = '';

my $bC_index = -1;
my $xmp_end_found = 0;
my $decl_bC_found = 0;
while ( <PDF> ) {
push @xmp, $_;
if ( /^\s*\Q$xmp_end\E/ ) {
$xmp_end_found = 1;
# Get amount of padding;
while (<PDF>) {
if ( /^\s*$/ ) {
$len_padding += length($_);
} else {
$xmp_after_line = $_;
last;
}
}
last;
}
elsif ( $bC_index >= 0 ){
next;
}
# Rest of conditions only apply if no place yet found for byteCount
# specification.
elsif ( /^(\s*)\Q$pre_bC\E.*?\Q$post_bC\E\s*$/ ) {
$bC_index = $#xmp;
}
elsif ( /^\s*\Q$decl_bC\E/ ) {
$decl_bC_found = 1;
}
elsif ( /^(\s*)\Q$rd_end\E/ ){
# End of rdf:Description block.
# So having previous declaration of byteCount is irrelevant.
$decl_bC_found = 0;
}
elsif ( $decl_bC_found && /^(\s*)\Q$pC\E/ ){
$bC_index = $#xmp;
pop @xmp;
push @xmp, '', $_;
}

} # End reading of XMP

if ($bC_index < 0) {
if ( ! $xmp_end_found ) {
warn " End of XMP packet not found.\n";
$status = 2;
}
elsif ( ! $decl_bC_found ) {
warn " XMP packet not in appropriate hyperxmp-compatible format.\n";
$status = 3;
}
return;
}
my $new_line = ' ' . $pre_bC . $pdf_size . $post_bC . "\n";
my $old_line = $xmp[$bC_index];
my $delta_len = length($new_line) - length($old_line);
if ($delta_len > $len_padding) {
warn " Cannot get padding correct for '$pdf_name'.\n",
" Length change of bC line = $delta_len; ",
" Padding bytes available = $len_padding.\n";
$status = 4;
return;
} else {
$len_padding -= $delta_len;
$xmp[$bC_index] = $new_line;
$status = 1;
}
}

#======================================================