[ruby-talk:6961] Accessing an non-existing segment of an array

A week ago;

raja> Problem;  a[i..j / i...j / i,cnt] returns nil instead of [] for
raja>           non-existing segments.
raja> 
raja>           Creates a problem for iterators as #each isn't defined for nil.
raja>
raja> Fix:      Don't know of a clean one.
raja>  
raja> Comment:  Either #each needs to be defined for nil [Doesn't feel right]
raja> 
raja>           Or a[i..j / i...j / i,cnt] should return the empty array, [],
raja>           for out out of bounds segments. [This feels right] 

matz> This is interesting comment.  Let's discuss.  If index is out of
matz> bound,what should [] return?  nil or empty array?

dave> For me, the primitive operation here is array element
dave> indexing,a[]. Everything else is defined in terms of it.  
dave> 
dave> a[x..y] should return [ a[x] + a[x.succ] ...a[y] ]
dave> ... 
dave> a[2..5] = [ a[2], a[3], a[4], a[5]] => [ 7, nil, nil, nil ]

The above proposal is very interesting but would it not 
complicate the intended use of the array segment?

for x in a[i..j]
   ...  would need to explicitly handle the fact that 
   ...  x could be nil when the user was expecting otherwise.
   ...  almost reduces back to the original problem of a[i..j] itself
   ...  possibly being nil 
end

My personal experience has been that when I ask for a segment of an array 
my intention is: "give me as much as you've got and truncate the rest".  Of
course, if that portion itself had nils in it then it is my responsibility
to be able to handle them.

     0 1 2 3 4
a = [2,3,5,7,11]

b = a[3..10]  #-> [7,11]

Now, if the user asks for a[i..j] the person might be expecting an array of
length j-i+1  and hence may want to access some a[k] 0<=k<=j-i+1

Given ruby's "non-bounds checking" indexing a[k] would give back nil for
those values of k outside the range of the returned truncated segment!
Which is also what your proposal would do.  Interesting potential "benefit"
of not doing bounds checking in ruby ... ;-)

Raja