Activity (RSS)

The JSClass for JSObjects representing JSFunctions (js_FunctionClass) just uses JS_FinalizeStub, which means we aren't cleaning Ruby procs out of @gcthings or ...

There's some annoying copy/paste between js_proxy and js_function_proxy. See if there's a way to share it without sucking.

Fixed, but there are problems with finalization. I'll make some new tickets.

Not quite. It at least needs to be Foo.new(), which is reasonable.

This should work, but doesn't:

ctx = Johnson::Context
ctx[:Foo] = Foo

assert_not_nil(ctx.evaluate("Foo.new"))

When Ruby objects head in to JSland, they should be rooted on the Ruby side. The JSland proxy finalizer should unroot them, and also remove them from the rbids...

Not sure whether these should be supported by the current JSProxy stuff, or whether there's a better way to back a JSArray with a native impl.

They're probably covered just fine by the existing JSProxy stuff, since method_missing is pretty helpful about looking for [] and key?.

Make this work:

ctx[:alert] = lambda { |m| puts m }
ctx.evaluate("alert('hai!')")

The sticky wicket here is that the ContextPrivate slot can't be used to s...

Make sure that weird strings will go in and come back out of the parser.

I cleaned up as much as I could. Only 233 lines. :-)

Fixed warning.

[119]

/Users/jbarnette/work/johnson/lib/johnson/visitors/ecma_visitor.rb:14: warning: regexp has `}' without escape

Make sure they roundtrip in both directions.

• Change "MutableTreeVisitor" to "MutableTreeConstructingVisitor".
• Like a million methods in ImmutableNode shouldn't be public
• Make ImmutableNode's constructor...

Should be able to run "johnson" at the command line and execute JS. Should take a bunch of Ruby's options, if possible, incl. -r, etc.

Make sure that johnson.rubyforge.org has something reasonable on it.

Make sure the API has some okay docs (and RDoc's generating them reasonably well).

Make sure the README doesn't suck for 1.0.

The current implementation of length/size on Proxy masks properties and function calls of the same name on the proxied JS object.

call, each, and call_function_property all run without marshaling any JS exceptions back over to Ruby.

each needs to make sure to free the JSIdArray and marsh...

Make sure we can deal with strings or symbols in all the places we provide property indexers (Context, Proxy, etc).

Make sure max, min etc work on proxies of arrays.

Make sure we figure out how to do a binary Windows gem for 1.0.

We're a little bit mono-platform at the moment. Make sure we work on a few linuxes. Try out a PPC platform too.

The way the SpiderMonkey extension (and vendor/spidermonkey) get built right now is really hacky, because I don't know extconf/mkmf very well. Clean it up.

The JSRuntime heap and JSContext stack sizes are hardcoded in Johnson::Spidermonkey::Context's allocate() method right now. There should be sane defaults, but ...

We pretty much only ever throw Johnson::Error. Compilation error, exception from JS land, whatever. This is lame.