[m-rev.] for review: start implementing RTTI in Mercury

[Fergus Henderson]

Quite a bit of this code is very much specific to the --no-high-level-data
case; you should add appropriate XXX comments to document that.

On 17-Aug-2001, Tyson Dowd <trd at cs.mu.OZ.AU> wrote:
> library/rtti_implementation.m:
>         Add a new module to implement RTTI for the .NET (and possibly
>         Java) backend.  This module tries to do most of the RTTI tasks
>         in Mercury rather than in C.  Hopefully it will be easier to
>         port this implementation in future.

You need to modify doc/Mmakefile to exclude this from the
modules that go in the Mercury library reference manual.

> library/std_util.m:
>         Call into rtti.m to compare type_infos.

s/rtti.m/rtti_implementation.m/

> runtime/mercury_il.il:
> 	Add helper functions to get the function pointers for tuple
> 	compare and unify procedures.
>
> runtime/mercury_mcpp.cpp:
> runtime/mercury_mcpp.h:
> 	Add tuple to the type_ctor_rep.
> 
> runtime/mercury_type_info.h:
> 	Mention that changes in the type_ctor_rep might require changes
> 	in rtti_implementation and mercury_mcpp.{h,cpp}.

That part of the log message is wrong; it's from the
other change that you posted.

> Index: library/rtti_implementation.m
...
> +	% In the .NET backend, we don't generally have to collapse equivalences
> +	% because they are already collapsed (this is why il grades require
> +	% intermodule optimization).

The comment there has causality the wrong way around. 

> +:- func type_ctor_rep(type_ctor_info) = type_ctor_rep.
> +:- pragma foreign_proc("C#",
> +	type_ctor_rep(TypeCtorInfo::in) = (TypeCtorRep::out), [], "
> +	int rep;
> +	rep = (int) TypeCtorInfo[4];
> +	TypeCtorRep = mercury.runtime.LowLevelData.make_enum(rep);
> +").

Ouch, horrible magic numbers.

I would prefer it if those were extracted out into some code
at the top

	enum type_ctor_info_field_nums {
		type_ctor_rep = 4,
		type_ctor_module_name = 7,
		type_ctor_name = 8,
		type_layout = 11
	}

> +++ runtime/mercury_il.il	16 Aug 2001 14:10:27 -0000
> @@ -42,6 +42,26 @@
>  // MC++ used to handle this, and then it stopped working, so now it's just
>  // not supported at all.  I hope it will make a comeback.
>  
> +.class public RuntimeHelperFunctions {
> +
> +.method public static default class [mscorlib]System.Object[]
> +get_type_ctor_info(class [mscorlib]System.Object[] TypeInfo) {
> +        ldarg TypeInfo
> +	ldc.i4.0
> +	ldelem.ref
> +	isinst class [mscorlib]System.Object[]
> +	brfalse isnull
> +	ldarg TypeInfo
> +	ldc.i4.0
> +	ldelem.ref
> +	castclass class [mscorlib]System.Object[]
> +	ret
> +isnull:
> +	ldarg TypeInfo
> +	ret
> +}
> +} // end class RuntimeHelperFunctions
> +
>  .class public TempHack {

The comment above this code is now in the wrong place --
it refers to the "TempHack" class.

You should also add comments explaining what your new code does
and why it needs to be written in IL.

But looking at the code, I can't see why you wrote it in IL.
It looks like it would be very easy to translate that into C#:

	object[] get_type_ctor_info(object[] TypeInfo) {
		try {
			return (object[]) TypeInfo[0];
		} catch (System.BadCastException) {
			return TypeInfo;
		}
	}

Much shorter, and easier to read.

Are you worried about efficiency,
i.e. the overhead of exception handling?
If so, then I think this is premature optimization.

Writing in assembler using poor data structures (in this case, `object[]')
is a great way to make things run slow and be unmaintainable.

-- 
Fergus Henderson <fjh at cs.mu.oz.au>  |  "I have always known that the pursuit
The University of Melbourne         |  of excellence is a lethal habit"
WWW: <http://www.cs.mu.oz.au/~fjh>  |     -- the last words of T. S. Garp.
--------------------------------------------------------------------------
mercury-reviews mailing list
post:  mercury-reviews at cs.mu.oz.au
administrative address: owner-mercury-reviews at cs.mu.oz.au
unsubscribe: Address: mercury-reviews-request at cs.mu.oz.au Message: unsubscribe
subscribe:   Address: mercury-reviews-request at cs.mu.oz.au Message: subscribe
--------------------------------------------------------------------------