initial import

git-svn-id: http://svn.irssi.org/repos/irssi-python@4282 dbcabf3a-b0e7-0310-adc4-f8d773084564

1 files changed, 62 insertions, 0 deletions

diff --git a/README b/README
new file mode 100644
index 0000000..ed4201f
--- /dev/null
+++ b/README
@@ -0,0 +1,62 @@
+Pytest for Irssi
+
+This module embeds Python into an Irssi module. It provides just enough of an 
+interface to port the hello.pl example from the Perl docs to Python. There are
+still a lot of things needed to be done, and some design issues to work out.
+See the scripts/ directory for two small example scripts.
+
+To compile:
+    So far I've compiled it on Debian i386 and on FreeBSD i386. Haven't tried
+    elsewhere. The Makefile will need to be edited for it to be compiled. It
+    requires Python 2.4 installed. Put libpytest.so wherever Irssi recognizes
+    modules.
+
+In pytest so far:
+    -Looks for scripts in ~/.irssi/pyscripts
+    -Created base object wrappers for structs with type and chat_type members.
+    -Wrapped members of most of the core structs: 
+        SERVER_REC 
+        WI_ITEM_REC
+        CHANNEL_REC
+        QUERY_REC
+    -Wrapped print methods for the global module, server, and window item.
+    -Wrapped command methods for server and window item.
+    -Included a signal handler in the base server and window item types that
+     seems to prevent crashes when an object is accessed after a server is
+     disconnected or a window item is removed (channel close, etc).  
+    -Python stdout and stderr are redirected to the Irssi status window.
+    -The command_bind method of the Script object binds commands.
+
+TODO:
+    -Wrap the rest of the *_REC structs.
+    -Wrap the important methods and functions.
+    -Signal handling
+    -Wrapper object factory
+    -Handle settings
+    -Handle timeouts
+    -Bunch of other stuff
+
+Design issues:
+   -Threading. Pytest doesn't consider thread usage by scripts at all. Any
+    thread usage would likely cause a crash, as pytest makes no attempt to
+    manipulate Python's interpreter lock or thread state. It would be easy to
+    use PyGILState_* critical sections around important parts where Irssi 
+    calls into Python, but I don't know about the larger impact on Irssi for
+    allowing threads to run. With the signal system and timeouts, is this
+    really an important issue?
+
+   -Structure. C modules could use some adjustment, especially as more code
+    is added. I also noticed how the Perl wrapper, like the rest of Irssi, 
+    seems to be organized around interface type. The Python wrapper may need 
+    some adjustment for this(?)
+
+   -Also, command_bind currently isn't a function in the main namespace like in
+    the Perl wrapper. Instead, its a method of the Script object, which includes
+    a list of bound signals for record keeping. That way, the context is always 
+    certain. Currently, the Script object is inserted into the user's script 
+    through a startup hook function (irssi_main). There could be a better way 
+    of doing this (perhaps by inserting the script object directly into the 
+    script's globals and using an import hook to distribute it to modules the 
+    script imports).
+  
+