[Truffle] New IO implementation #2866
Conversation
…ading the kernel.
Conflicts: truffle/src/main/java/org/jruby/truffle/runtime/core/CoreLibrary.java
Conflicts: truffle/src/main/java/org/jruby/truffle/translator/BodyTranslator.java
| //if (!state -> check_async(calling_environment)) | ||
| // return NULL; | ||
| //io -> ensure_open(state); | ||
| getContext().getSafepointManager().poll(this); |
There was a problem hiding this comment.
If this actually works on STDIN, this will be awesome!
There was a problem hiding this comment.
STDIN.each_line do |line|
puts line
endWith both tty and piped input works. Is that what you meant?
There was a problem hiding this comment.
I mean, the ability to interrupt reading from STDIN is what interests me.
There was a problem hiding this comment.
No, I'm afraid it doesn't appear to work at the moment. Can you interrupt all reads in a process somehow?
Conflicts: truffle/src/main/ruby/core/rubinius/common/kernel.rb truffle/src/main/ruby/core/shims.rb
[Truffle] New IO implementation
|
Merged, but still open for comments |
| public EachNode(RubyContext context, SourceSection sourceSection) { | ||
| super(context, sourceSection); | ||
| eachSymbol = getContext().getSymbolTable().getSymbol("each"); |
There was a problem hiding this comment.
I think we use getContext().newSymbol() almost exclusively. It might make future refactorings easier to stick with that.
There was a problem hiding this comment.
Didn't know about that. I would expect it to be getContext().getSymbol() since this will likely not create a new Symbol everytime.
There was a problem hiding this comment.
It's perhaps a bit misleadingly named. It just does return symbolTable.getSymbol(name, ASCIIEncoding.INSTANCE);, which will return a cached Symbol if it already exists or creates a new one and populates the cache otherwise.
There was a problem hiding this comment.
Fixed in 5c53e3c
This reimplements Truffle's File and IO classes using Rubinius, on top of a few Rubinius primitives and some POSIX calls. There's now no reference to File or IO in the Java code at all.
This gives patch alone gives us 761 new passing specs (about 5% of the entire core specs), but it also opens the door for the rest of IO and things like Socket.
The primitives use
ruby()a lot, mainly to read and write instance variables, which we might want to remove sooner rather than later.There is also a lot of copying going on here. I think
putswill copy at least four times. The Rubinius primitives also rely on stack allocation of large buffers being efficient, which does not translate well into Java.@eregon @nirvdrum @thomaswue