make for, while, etc. return Nothing instead of ()

[StefanKarpinski]

We discussed this once a long time ago and decided that it would be nice to have a Nothing object that prints as nothing. Then when entering code into the repl, one doesn't have to worry about putting an unseemly ; at the end of a for loop to suppress the annoying (). Also it would make some potential errors more sensible: if you try to do anything with the Nothing value it can throw a fairly specific error, whereas () is a perfectly legitimate value for many expressions to produce.

[JeffBezanson]

BTW, not "if" since that returns the value of whatever branch executes.

[JeffBezanson]

...unless of course no branch executes.
I see you assigned yourself so I'll just make some helpful comments. In the intermediate representation I use (null) (null_sym in C) for "no value" so the interpreter and compiler should return Nothing for this instead. Nothing should be defined in boot.j and handled in init.c and dump.c the same as the other stuff in jl_get_builtin_hooks(). Then replace jl_null with jl_nothing where appropriate in the C code --- this is tricky, as some cases actually need to be 0-tuples. Replacing the cases in j/ should be fairly easy.

[StefanKarpinski]

Excellent. Thanks for the tips. Will attempt this.

On Jun 2, 2011, at 6:32 PM, JeffBezansonreply@reply.github.com wrote:

...unless of course no branch executes.
I see you assigned yourself so I'll just make some helpful comments. In the intermediate representation I use (null) (null_sym in C) for "no value" so the interpreter and compiler should return Nothing for this instead. Nothing should be defined in boot.j and handled in init.c and dump.c the same as the other stuff in jl_get_builtin_hooks(). Then replace jl_null with jl_nothing where appropriate in the C code --- this is tricky, as some cases actually need to be 0-tuples. Replacing the cases in j/ should be fairly easy.

Reply to this email directly or view it on GitHub:
#28 (comment)

[StefanKarpinski]

I'm getting errors like this:

type error: new_struct_type: expected Function, got NothingType /Users/stefan/projects/julia/j/sparse.j:2 /Users/stefan/projects/julia/j/sysimg.j:58

which I think comes from this line:

#3  0x00000001000290c9 in jl_f_new_struct_type (env=0x0, args=0x7fff5fbfc478, nargs=4) at builtins.c:993
992     if (args[3] != (jl_value_t*)jl_null)
993         JL_TYPECHK(new_struct_type, function, args[3]);
(gdb) p args[3] == jl_null   
$6 = 0
(gdb) p args[3] == jl_nothing
$7 = 1

So jl_f_new_struct_type is getting jl_nothing when it should be getting jl_null. I think this is a matter of the codegen/interpreter turning null_sym into jl_nothing where it should be yielding jl_null but I'm not sure what to do about it. Presumably the parser should distinguish between () and Nothing.

[JeffBezanson]

In this case the frontend is passing (null) to new_struct_type just as a flag to indicate nothing is there. Change jl_f_new_struct_type to expect jl_nothing instead.

[JeffBezanson]

If you want you can send me a list of occurrences of jl_null you're not sure about and I can help.

[StefanKarpinski]

Closed by 20e911d.