Skip to content


Subversion checkout URL

You can clone with
Download ZIP
Branch: master
Fetching contributors…

Cannot retrieve contributors at this time

104 lines (68 sloc) 3.52 kB
Before 1.0.0
- It seems Python 2 pickles don't work in Python 3 and crash ErlPort session.
- Don't import modules on every call. We probably should catch the exception
and import only if needed
- Raise some special errors if language libraries are called without the
ErlPort connection
- We can add erlang.receive() function which waits for next message with
optional timeout (timeouts can be implemented with select/etc. modules and
probably with msvcrt.kbhit() on Windows). What about Ruby?
- How erlang.receive() should work with threads? Can we just wait for calls in
different threads with erlang.receive() (probably it should be an
erlang.wait_for_calls_here() call which calls erlang._receive() without any
arguments)? What if someone call this function in main thread, does it block
everything? Maybe it's better to create an erlang.start_receive_thread()
function? If we run receive() from a thread main thread can be already in
receive mode so it will be UnexpectedMessage error in this case.
- Should we exit in case of Python's SystemExit exception?
- Do we really need erlang.make_ref() function?
- Split erlport_utils:send_request into send_call_request and send_request?
- Probably we need 'queue' option for set_message_handler() in which case
only receive() call can be used to receive messages?
- Set default timeout to 5 sec?
- Allow not only atoms in call() functions and Atom() classes. Convert strings,
unicode(?) to atoms. Test functions with Unicode names.
- Add more API for middleware scenario (python, ruby). Reraise the language
related exceptions in every module. And allow to start a language instance on
other nodes. Also language instances can be created with a pid. We can also
create factory functions at module level.
- Add custom encode/decode handlers in Erlang? Should it be a callback module?
And tests.
- Add [{timeout, Timeout}] options for function
- Allow calls in selected processes in Erlang. We can create a class
(erlang.Process) which has call and exit methods. If Erlang process will die
we just raise 'noproc' error.
- Split Erlang tests by languages? For example it can be useful if some of the
languages not installed.
- Comments for modules
- Check all TODO/FIXME comments
- We need more robust error identification in erlport:spawn_call?
- erlport:server_name/0 and erlport:server_instances() types should be moved to
some utils module? Maybe move types to a header file because -export_type
seems doesn't work as expected in Erlang R13
- Threaded mode with numbered requests. We can use distinct callback module
in this case.
- Support for other languages (Perl, Lua, ?).
- Faster Erlang terms decoder
- Maybe we can also redirect STDIN?
- Can we allow method calls for objects? For example we can pass a tuple
{object, Object} instead of a module name? It also can be another method for
example call_method()?
- We can try to use 'no_suspend' options for port and try to batch port
- We can create references for unpicklable objects (and all objects through the
explicit API). And we can create proxy objects this way. Need to think about
garbage collection however.
- Add xref checks?
- More methods for Ruby's Tuple class?
- Add a pub/sub framework?
- Add workers_pool module with pluggable workers selection policies?
- Maybe it's possible to create an ErlPort interface which loads dynamic
Jump to Line
Something went wrong with that request. Please try again.