In the olden days of computing, data was stored in patterns of holes on stiff pieces of paper called punched cards. Putting just the right holes in just the right places on these cards was fiddly and so the task was mediated by a special machine called a keypunch.
Though widely considered at best a lateral move, today in most applications the punched card has been replaced by the PDF document. And much like its predecessor the raw PDF document format requires a fastidious hand to put every bit in its proper place, lest even the simplest documents be rendered (or rather not rendered) unreadable.
keypunch is a low-level Haskell library for constructing well-formed PDF documents. Here low-level means we are mainly concerned with exposing the smallest possible semantic building blocks of PDFs while hiding the grittiest details involved in producing a sequence of bytes usable by PDF readers. Text layout algorithms, for instance, are beyond the scope, although it should be possible to implement one (or more!) using this library as a foundation. A key design goal is to produce PDFs which can be efficiently streamed, such as by a web server.
Current status: not useful for any purpose.
TODO:
- Better serialization for e.g. draw commands. cache length computation?
- Type safe names; e.g. font name
- when a font name is expected, make it unpossible to use a different kind of name
- addFont returns a tagged name, and setFont consumes it
- API for name and number trees
- insert things into the name tree
- convert the name tree to Objs
- render the name tree with the catalog