TL;DR

Taking a look at textual elements in SVG.

I think it’s not that difficult to guess where I’m heading to, considering all the fuss about stitching stuff on top of PDF files (with PDF::Collage) and generally have a way to generate PDFs programmatically. SVG just is another direction of investigation.

So, after figuring out how to deal with px units, it makes sense to figure out where text is placed in SVG files.

Why?

Well… we could go the cairosvg way of course, or even fiddle with InkScape from the command line. But hey! This would be having less fun, right?

So let’s assume we want to look inside the SVG box and deal with the text inserts inside. I feel both lazy and hands-on, so I’ll just observe that my current InkScape installation produces stuff like the following (redacted for conciseness):

<text     style="..." x="100" y="100"
  ><tspan style="..." x="100" y="100"
    >whatever</tspan></text>

I can just guess that tspan elements can be inside text elements, but I don’t know if text elements can fit inside other text elements or the like. Besides, I’ll only be using simple stuff.

Assuming that text elements might be embedded in other ones, it makes sense to have a stack of them, inheriting unknown stuff from the parent. While parsing each of them we can take the upper one as default, which is what’s done in the following function (using $upper_frame to this extent). The $cv is a conversion hash from SVG px conversion in Perl.

sub parse_textish ($el, $cv, $upper_frame) {
   my $frame = {$upper_frame->%*};    # shallow copy will do

   # x attribute just needs to be converted from px
   if (defined(my $x = $el->getAttribute('x'))) {
      $frame->{x} = sprintf '%.2f', $x * $cv->{factor};
   }

   # y attribute needs to take into account that PDF's viewport has the
   # origin set on the lower-left corner, not upper-left like SVG.
   if (defined(my $y = $el->getAttribute('y'))) {
      $frame->{y} = sprintf '%.2f', ($cv->{Y_span} - $y) * $cv->{factor};
   }

   # the style attribute can be a trove of useful information
   if (defined(my $style = $el->getAttribute('style'))) {
      if (defined(my $font_size = $style->{'font-size'})) {
         $font_size =~ s{px\z}{}mxs or die "font-size...";
         $frame->{font_size} = sprintf '%.2f', $font_size * $cv->{factor};
      }
      if (defined(my $ta = $style->{'text-align'})) {
         $frame->{align} = $ta;
      }
      if (defined(my $ff = $style->{'font-family'})) {
         $frame->{font_family} = $ff;
      }
   } ## end if (defined(my $style ...))

   # cut to integers if it makes sense
   s{\.00\z}{}mxs for $frame->@{qw< x y font_size >};

   # CDATA might or might not be present, whatever is good
   $frame->{cdata} = $el->getCDATA;

   return $frame;
} ## end sub parse_textish

As we’re finally heading to PDF-land, we’re doing Y axis transformation right on the spot. SVG has Y coordinates start from top and increase going down, while PDF starts from the bottom and increases going up. Nothing too difficult to takle, anyway.

The style parsing is conveniently already done by SVG::DOM, so we just have to look into the corrensponding hash reference provided by $el->getAttribute('style'). We just take the font family, the size and its alignment, which we will eventually use to drive PDF::Collage.

Stay safe!