Re: Updates to ob-racket mode, code review request

[Tom Gillespie <tgbugs@gmail.com>]

Hi George,
  I took a glance at your additions, nothing jumps out as being problematic. I
have been using wallyqs ob-racket (https://github.com/wallyqs/ob-racket) for
about 4 years because it has support for specifying the #lang in the header,
and I seem to recall that there were some other features that were lacking in
all the (2? 3?) other ob-racket's that I tried. I do not think that any of the
existing ob-racket implementations should be put on melpa since they do not
interoperate with the recent work that Greg has done on racket-mode to
support comint and multiple repls (not to mention racket-xp-mode). Which is
to say, that I think that proper session support is a prerequisite for any
ob-racket implementation become "the" ob-racket on melpa or elsewhere. There
are also significant issues with how racket's "the top level is hopeless"
attitude interacts with things like org-babel's concept of a session (which
one might be able to make equivalent to a racket-mode repl). There are likely
ways around this, but I suspect they will involve something like
dynamic-rerequire and fiddling with namespaces to make it possible to
rerequire a module into an existing session. Alternately, compiled racket
modules load fast enough that running raco make on racket source blocks
behind the scenes might be able to fake a session-like experience. One place
to look for inspiration on this might be ob-clojure-literate which has to
deal with some similar issues. Another though related to sessions is around
the possibility of running racket blocks pointing to remote sessions. With
lisp blocks it is possible to switch to a session running on a remote server
and send the code to it that way. With the work that Greg has done for racket
mode it should be possible to spin up a remote racket session or connect to
one and then run the block in that session. Don't let this stop you though :)
Best!
Tom