/ 
PDF-Make-0.11
/ PDF::Make::Markup::Diff

NAME

PDF::Make::Markup::Diff - which pages changed between two renders

SYNOPSIS

my $d = PDF::Make::Markup::Diff->diff($bytes_before, $bytes_after);
printf "%d of %d pages changed\n", $d->{changed}, $d->{pages_b};
for my $p (@{ $d->{pages} }) {
    next unless $p->{changed};
    printf "page %d: -%s +%s\n", $p->{page},
        "@{$p->{removed}}", "@{$p->{added}}";
    printf "page %d: %s moved\n", $p->{page}, "@{$p->{moved}}"
        if @{ $p->{moved} };
}

DESCRIPTION

The diff behind the engine version promise. A layout change that moves a corpus page is a deliberate engine version bump or it does not merge - this dist's own corpus test asks exactly that - and a hosted template's upgrade review asks it again per fixture before a customer opts in.

The unit is the positioned word from structured extraction, coordinates rounded to the point: a moved column flags its page, a re-encoded but identical page does not, and no rasterizer is required. When the rasterizer lands a pixel mode joins; this interface stays.

WHAT IS REPORTED

Words are kept in reading order - lines top to bottom, words left to right - and the two pages are aligned by a longest common subsequence over the text. Each page then reports three lists, all in reading order:

removed

Words present before and gone after.

added

Words new after.

moved

Words whose text survived but whose position did not - and only when nothing was removed or added. Editing a word shifts every word after it on the line, so once the text has changed, movement is the consequence rather than the news. A page whose every word survived in a new place is the reflow an engine upgrade has to be checked for, and that is the case this list exists to name.

changed is true if any of the three is non-empty, so a page that only reflowed is still flagged. Each list is capped at forty words: a page that changed wholesale says so with its first few dozen.

METHODS

pages

my $pages = PDF::Make::Markup::Diff->pages($pdf_bytes);

Per-page words in reading order, each { text, x, y } with the coordinates rounded to the point.

Before 0.11 this returned per-page bags of "text@x,y" counts.

diff

my $d = PDF::Make::Markup::Diff->diff($bytes_a, $bytes_b);

{ pages_a, pages_b, changed, pages => [ { page, changed, removed, added, moved } ] }.