[m-rev.] for review: improve standard library error handling

[Tyson Dowd]

On 30-Aug-2001, Simon Taylor <stayl at cs.mu.OZ.AU> wrote:
> 
> Estimated hours taken: 6
> Branches: main
> 
> Generate exceptions rather than program aborts for domain errors
> and out of bounds array accesses.
> 
> Improve the handling of the arithmetic functions.
> 
> library/float.m:
> library/int.m:
> compiler/builtin_ops.m:
> 	Handle division by zero with an exception rather than a
> 	program abort.
> 
> 	Add int__unchecked_quotient and float__unchecked_quotient,
> 	which don't check for division by zero.
> 
> 	Remove reverse modes of the arithmetic functions in float.m.
> 	Richard O'Keefe pointed out a while ago that they don't work
> 	because of rounding errors.
> 
> 	Remove the long obsolete `int__builtin_*' and
> 	`float__builtin_float_*' predicates.
> 
> library/math.m:
> library/array.m:
> 	Generate exceptions rather than program aborts.
> 	The bounds and domain checks are now implemented in
> 	Mercury, so they do not need to be duplicated for each
> 	target language.
> 
> 	As discussed on mercury-users a while ago, math__pow(0.0, 0.0)
> 	should return 1.0. This is also consistent with float__pow.
> 	See <http://www.cs.mu.oz.au/research/mercury/mailing-lists/mercury-users/mercury-users.0104/0130.html>.
> 
> NEWS:
> 	Document the changes.
> 
> tests/general/float_test.m:
> tests/general/string_format_test.m:
> tests/hard_coded/ho_solns.m:
> tests/hard_coded/ho_univ_to_type.m:
> tests/hard_coded/qual_strang.m:
> tests/hard_coded/qual_strung.m:
> 	Rename occurrences of `builtin_*'.
> 
> 
> Index: NEWS
> ===================================================================
> RCS file: /home/mercury1/repository/mercury/NEWS,v
> retrieving revision 1.216
> diff -u -u -r1.216 NEWS
> --- NEWS	2001/08/13 05:49:27	1.216
> +++ NEWS	2001/08/29 16:13:50
> @@ -22,6 +22,16 @@
>    Reference Manual.
>  
>  Changes to the Mercury standard library:
> +* The predicates and functions in int.m, float, math.m and array.m now 
> +  generate exceptions rather than program aborts on domain errors and
> +  out-of-bounds array accesses.
> +
> +* We've removed the buggy reverse modes of the arithmetic functions in
> +  float.m (because of rounding errors the functions aren't actually
> +  reversible).
> +
> +* math__pow(0.0, 0.0) now returns 1.0, not 0.0.
> +  
>  * The exception module has a new predicate `try_store', which is
>    like `try_io', but which works with stores rather than io__states.
>  
> @@ -46,6 +56,11 @@
>    representations of term components as strings.
>  
>  * We've made the outputs of the string concatenation primitives unique.
> +
> +* We've removed the long obsolete `int__builtin_*' and
> +  `float__builtin_float_*' predicates, which were synonyms
> +  for the arithmetic functions dating from when Mercury didn't
> +  have functions.


>  
> Index: library/array.m
> ===================================================================
> RCS file: /home/mercury1/repository/mercury/library/array.m,v
> retrieving revision 1.97
> diff -u -u -r1.97 array.m
> --- library/array.m	2001/08/13 01:19:56	1.97
> +++ library/array.m	2001/08/29 13:52:47
> @@ -91,6 +91,11 @@
>  :- mode array_muo == out(mostly_uniq_array).
>  :- mode array_mui == in(mostly_uniq_array).
>  
> +	% An `array__error' is the exception thrown if any of the
> +	% operations fail. The string is a description of the error.
> +:- type array__error
> +	---> array__error(string).
> +


I feel a specific error for array index out-of-bounds would be nice, but
I'm not adamant about it.

> -#ifdef	ML_ARRAY_THROW_EXCEPTIONS
> -  #include ""exception.h"" 		/* for ML_throw_string */
> -  /* shut up warnings about casting away const */
> -  #define	ML_array_raise(s)	ML_throw_string((char *) (void *) s)
> +:- pragma foreign_proc("C#", bounds_checks,
> +		[thread_safe], "
> +#if ML_OMIT_ARRAY_BOUNDS_CHECKS
> +	SUCCESS_INDICATOR = FALSE;
>  #else
> -  #include ""mercury_misc.h""		/* for MR_fatal_error() */
> -  #define	ML_array_raise(s)	MR_fatal_error(s)
> +	SUCCESS_INDICATOR = TRUE;
>  #endif
> +").		
> +

Unfortunately there is no C# support for SUCCESS_INDICATOR (yet).
Instead please use MC++ for the moment (the code is fine, just change
"C#" to be "MC++".

Same in math.m.

It's also not worth putting in #if ML_OMIT_ARRAY_BOUNDS_CHECKS
because array bounds checking is done by the .NET runtime, and cannot be
switched off.

The issue regarding pow should be resolved before making any changes --
I'll leave that up to you and Fergus and whoever else wants to duke it
out.

Otherwise the change looks good, thanks.

Tyson.
--------------------------------------------------------------------------
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
--------------------------------------------------------------------------