Re: Do Pandoc's conversion functions which exist in both directions guarantee/aim for bijectivity?

[John MacFarlane <jgm-TVLZxgkOlNX2fBVCVOL8/A@public.gmane.org>]

This is correct.  We aspire to semantic equivalence but
definitely not syntactic.  For example, if you convert

hi # there

to Markdown you may get

hi \# there

Ideally, we'd have the following invariant:

  doc == readF (writeF doc)

where readF/writeF are the reader and writer for a format,
with the same extensions set, and doc is an arbitrary
Pandoc AST.

I wouldn't say we have this now with most formats. Running
this sort of test can indicate places where the reader or
writer needs to be improved.

Daniel Staal <DStaal-Jdbf3xiKgS8@public.gmane.org> writes:

> I don't believe there's any particular aim for it -- notably the 
> Markdown format is known to not always return the input when run on 
> itself.  The aim is for the output to be the semantic equivalent, not 
> that it will be the same.
>
> Daniel T. Staal