Pandoc extension to insert the metadata of a file as headings of levels 1, 2 and 3. This is so that if you have (say) each chapter of a thesis in its own file with its own title block, you can still compile the lot together to the one final thesis. See the file for more details.

chapter.hs

#!/usr/bin/env runhaskell
-- Filter that adds the metadata block as h1, h2, h3 (title, author, date).
--
-- Usage::
-- chmod +x ./chapter.hs
--
-- # (where each chapter is in its own md file and I want to compile them
-- #  into one document preserving the title/author/date for each chapter):
-- 
-- for f in Chapter*.md; do \
--     pandoc --filter ./chapter.hs $f; \
--     echo ""; \
-- done | pandoc --standalone
--
-- As a one-liner for easy copy/paste::
--
-- for f in Chapter*.md; do pandoc --filter ./chapter.hs $f; echo ""; done | pandoc --standalone
--
-- Add your pandoc switches as necessary. For example I'd have a file
-- title.txt with the title block for the entire document and add that in
-- to the last pandoc command to set the metadata for the entire document.
--
-- See [1] for writing pandoc extensions.
-- This script is hosted on gist [2].
-- [1]: http://johnmacfarlane.net/pandoc/scripting.html
-- [2]: https://gist.github.com/mathematicalcoffee/e4f25350449e6004014f
-- Amy Chan, May 2014.
import Text.Pandoc.JSON
import Data.List ( intercalate )
 
main :: IO ()
main = toJSONFilter insertMeta
    where insertMeta (Pandoc meta xs) = Pandoc meta xs'
            where xs' = [ Header 1 ("",["title"],[]) (docTitle meta)
                        , Header 2 ("",["author"],[]) (intercalate [Str ";", Space] (docAuthors meta))       
                        , Header 3 ("",["date"],[]) (docDate meta) ] ++ xs

When you write this:

For example I'd have a file title.txt with the title block for the entire document and add that in to the last pandoc command to set the metadata for the entire document.

Can you explain how that works, exactly? I can't seem to achieve your result. If I add a title.txt to the last pandoc command, pandoc completely ignores input passed to it via the pipe and only parses the title.txt file. (More info on approaches I've tried here)