Front page | perl.perl5.porters |
Postings from July 2001
Scott L Miller <Scott.L.Miller@compaq.com> writes:
>In my mind at least, clamping and readonly are 2 different animals.
>A hash that has been "clamped" could still have its values modified,
>but can't have a new key added, but a hash that is readonly wouldn't
>be able to add a key, or have its values modified...
Agreed.
My "straw man" proposal of how this is implemented is an
XS module which turns on _existing_ internal SvREADONLY flag.
Thus:
use SomeModule qw(readonly);
# error messages are not definitve ...
my %hash = (a => 1, b => 2);
my $ref = \$hash;
readonly \%hash; # clamped
$hash{'a'} = 10; # okay
$hash{'c'} = 30; # dies "Attempt to add key to readonly hash"
delete $hash{'b'}; # dies "Attempt to delete key from readonly hash"
$ref->{'a'} = 10; # okay
$ref->{'c'} = 30; # dies "Attempt to add key to readonly hash"
delete $ref->{'b'}; # dies "Attempt to delete key from readonly hash"
readonly(\$_) foreach values %hash; # readonly - fix the values
$hash{'a'} = 15; # dies "Attempt to modify readonly value"
The snag with that is the need to pass reference \%hash to readonly xsub.
So a new "operator" could be invented ....
>
>-Scott L. Miller
>
>> >I can imagine convincing arguments for semantics that do and don't
>> >allow updates via the reference.
>>
>> I can't. If the hash in readonly/clamped then it does not matter how
>> you reach it it is readonly/clamped.
>>
>> --
>> Nick Ing-Simmons
>> http://www.ni-s.u-net.com/
>>
--
Nick Ing-Simmons
http://www.ni-s.u-net.com/
Thread Previous
|
Thread Next