RE: membership test problem

["Stone, Joshua I" <joshua dot i dot stone at intel dot com>]

On Thursday, October 26, 2006 1:54 PM, Mike Mason wrote:
> What should the following code print?  I expected TRUE, but got
> FALSE.  Looks like if the array element value is 0, the membership
> test fails.  Is this correct?  
> 
> p = pid()
> t = tid()
> 
> prot_tx[p,t] = 0
> 
> if ([p,t] in prot_tx) {
> 	print("TRUE\n")
> } else {
> 	print("FALSE\n")
> }
> 
> - Mike

Setting a value to zero effectively deletes it from the map.  I believe
this was done to save space in maps, since they have a limited number of
entries.  Thus, 'map[key] = 0' is equivalent to 'delete map[key]'.  The
same goes for setting an empty string.

This mostly does what one would expect, since reading an uninitialized
value also yields 0 or "".  It's only the membership test where the
result may be surprising, since it's exposing an implementation detail.

At one time, the 'delete' keyword didn't work in all situations, so
setting 0 was the only way to clear items from the map.  Now that
'delete' works universally, we could think about persisting 0 values so
the membership test succeeds.


Josh