[m-dev.] Why can't we pass a det pred where cc_multi is expected?

[Peter Hawkins]

Hi...
Ralph Becket wrote:

>Consider
>
>This seems odd to me.  Is this deliberate or just an oversight?
>  
>
Incidentally, a not entirely unrelated problem was annoying me recently. 
I was writing some command-line argument parsing code like so:

:- type option ---> propagator ; labelling_opt ; allsolns_opt ; 
no_combine_opt.
:- pred option_default(option::out, option_data::out) is nondet.
option_default(propagator, string("domain")).
option_default(labelling_opt, string("ffmaxout")).
option_default(allsolns_opt, bool(no)).
option_default(no_combine_opt, bool(no)).

In the main predicate, I have:
    getopt.process_options(
        option_ops(short_option, long_option, option_default),
        Args,
        NonOptionArgs,
        Result),

When compiling the program, I receive this warning message:
test_steiner.m:185: In `option_default(out, out)':
test_steiner.m:185:   warning: determinism declaration could be tighter.
test_steiner.m:185:   Declared `nondet', inferred `multi'.

However if I change the determinism of option_default to 'multi', I 
receive the following pearl of wisdom:
test_steiner.m:202: In clause for `main(di, uo)':
test_steiner.m:202:   in argument 1 of call to predicate 
`getopt.process_options/4':
test_steiner.m:202:   mode error: variable `V_20' has instantiatedness 
`unique(getopt.option_ops(/* unique */(pred((ground >> ground), (free >> 
ground)) is semidet), /* unique */(pred((ground >> ground), (free >> 
ground)) is semidet), /* unique */(pred((free >> ground), (free >> 
ground)) is multi)))',
test_steiner.m:202:   expected instantiatedness was 
`bound(getopt.option_ops((pred((ground >> ground), (free >> ground)) is 
semidet), (pred((ground >> ground),
(free >> ground)) is semidet), (pred((free >> ground), (free >> ground)) 
is nondet)) ; getopt.option_ops((pred((ground >> ground), (free >> 
ground)) is semidet), (pred((ground >> ground), (free >> ground)) is 
semidet), (pred((free >> ground), (free >> ground)) is nondet), 
(pred((ground >> ground), (ground >> ground), (ground >> ground), (free 
 >> ground)) is semidet)) ; getopt.option_ops_multi((pred((ground >> 
ground), (free >> ground)) is semidet), (pred((ground >> ground), (free 
 >> ground)) is semidet), (pred((free >> ground), (free >> ground)) is 
multi)) ; getopt.option_ops_multi((pred((ground >> ground), (free >> 
ground)) is semidet), (pred((ground >> ground), (free >> ground)) is 
semidet), (pred((free >> ground), (free >> ground)) is multi), 
(pred((ground >> ground), (ground >> ground), (ground >> ground), (free 
 >> ground)) is semidet)))'.
For more information, try recompiling with `-E'.

Perhaps it should be possible to transparently perform determinism 
'up-casts' in cases such as this? Alternatively can the warning message 
be suppressed?

=)
Peter

--------------------------------------------------------------------------
mercury-developers mailing list
Post messages to:       mercury-developers at cs.mu.oz.au
Administrative Queries: owner-mercury-developers at cs.mu.oz.au
Subscriptions:          mercury-developers-request at cs.mu.oz.au
--------------------------------------------------------------------------