Finding Headings and Link Text in MARS files

Question

Hi,
I've been looking into the MARS package and have been able to
extract some great information for my application to read from the
underlying XML, but I'm stuck on a few key items. Does anyone know
how to reliably and programmatically find the following items
throughout a MARS package?
- Headings
- Hyperlink TEXT (I can get the Rect & Dest from the
/page/#/pg.can file but not the text itself)
- Tables & Graphics (where they start and end in the
svg?)
What I'm looking to do is basically build a big inventory of
all these common objects from a .mars file and report them to the
user. I'm just using DOM to read the XML/SVG files.
Thanks!

Answer 1

Hi,
Sadly, there is probably no easy way to do the things you are
asking. Mars files generated from PDF are reliant of the
funadmental properties of PDF. Below I will discuss each of your
points:
Headings:
Depending on the producing application and the settings used
to convert to PDF, there might be some slightly more intelligent
way to do this. A PDF can be "Tagged", such that it has a logical
structure tree. If your sources are very similar in nature (i.e.
same producer and same settings) and come from a producer that
creates Tagged PDF, then it is possible you can use the logical
structure to locate headings. However, if the PDF does not have
this information, the only indication you have that a given block
of text is a heading is that it is bigger than the rest of the text
and perhaps begins with a number. If the PDF is Tagged, then look
for a struct.xml file in the same location as the page. This
references a marked content section in the SVG, which is a named
group.
Hyperlinks:
Sadly, there is very little that can be done here. PDF
represents links as rectangles on the page. There is no direct
association with the text that the user sees as a link. In HTML
links are described as anchors linking to another location with a
specific piece of text representing them. In PDF, it is just a
rectangular overlay on the page. Any change in color (or an
underline perhaps) is a text property and not part of the link.
This is for historical reasons in PDF and Mars does not currently
try to reconcile these pieces of information.
Tables & Graphics:
Well, tables are hard. There is no "table" operator in PDF,
so tables are created using paths to draw the lines and text
operators to place the text inside the table. Again, if you have a
Tagged PDF, it is quite possible that the table has been marked in
the struct.xml and the content identified, but as above, this is
highly dependent on the nature of the source that was used to
generate the PDF. The same is true for graphics. Graphics created
with paths and other drawing operators do not necessarily get
grouped as such, but a Tagged PDF (and therefore a Tagged Mars
file) might have them.
Conclusion:
So, to summarize the above, there are possibilities and if
you have control over the sources, there is a chance that you can
easily obtain the information you require. However, if you have a
number of disparate sources, then I suspect there is no easy
solution for you.
If you have any more specific questions or more details of
the nature of your sources, perhaps there might be something you
can look for. Also, let me know if you need more information
regarding our logical structure implementation in Mars. As always
with PDF, you are restricted by the quality of your sources and
third party PDF creation tools do not necessarily add all the
information that you would require for what you need.
Matthew