Re: [TCLCORE] null handling and TIP #185

[<lm...@bi...>]

We've been over a lot of the NULL issues in L.  We do not have an answer 
for non-compiled code, i.e., in the interpreted path there isn't any
accurate way to dectect the difference between the empty string and 
a NULL.  On the other hand, for all compiled code paths we believe we
can support the null concept.

The approach we're taking is to create a tclobj, that starts with a ref
count of 1 so it will never go away, and we compare against the address
of that obj to see if the other obj is the same.

So far as we can see, and we here includes Jeff H, there isn't any
practical way to represent a NULL cleanly in interpreted paths.
But the trend is towards all compiled and for interpreted you are
no worse off than you are today.
-- 
---
Larry McVoy                lm at bitmover.com           http://www.bitkeeper.com