Your code
We typically read Lisp code by indentation, and don't write in all-caps. Since we read by indentation, we don't need to put closing parens (or any parens, really) on their own line. Your code, properly formatted, then, is:
(defun split-list (l)
(cond
((null l) '())
((= 1 (length l)) (list (car l)))
(t (cons (cons (car l)
(split-list (cddr l)))
(cadr l)))))
Getting the base cases right
Split-list is always supposed to return a list of two lists. We should cover those base cases first. When l is empty, then there's nothing in the left list or the right, so we can simply return '(() ()). Then first condition then becomes:
((null l) '(() ())) ; or ((endp l) '(() ()))
Judging by your second case, I gather that you want the second and third cases to be: (i) if there's only one element left, it must be an odd-numbered one, and belongs in the left result; (ii) otherwise, there are at least two elements left, and we can add one to each. Then the second condition should be
((= 1 (length l)) (list (car l) '()))
It's actually kind of expensive to check the length of l at each step. You only care whether there is only one element left. You already know that l isn't empty (from the first case), so you can just check whether the rest oflis the empty list. I find it more readable to usefirstandrest` when working with cons cells as lists, so I'd write the second clause as:
((endp (rest l)) (list (list (first l)) '()))
Handling the recursive case
Now, your final case is where there are at least two elements. That means that l looks like (x y . zs). What you need to do is call split-list on zs to get some result of the form (odd-zs even-zs), and then take it apart and construct ((x . odd-zs) (y . even-zs)). That would look something like this:
(t (let ((split-rest (split-list (rest (rest l)))))
(list (list* (first l) (first split-rest))
(list* (second l) (second split-rest)))))
There are actually some ways you can clean that up. We can use destructuring-bind to pull odd-zs and even-zs out at the same time. Since this is the last clause of the cond, and a clause returns the value of the test if there are no body forms, we don't need the initial t. The last clause can be:
((destructuring-bind (odd-zs even-zs) ; *
(split-list (rest (rest l)))
(list (list* (first l) odd-zs)
(list* (second l) even-zs))))))
*I omitted the t test because if a cond clause has no body forms, then the value of the test is returned. That works just fine here.
Putting that all together, we've reworked your code into
(defun split-list (l)
(cond
((endp l) '(() ()))
((endp (rest l)) (list (list (first l)) '()))
((destructuring-bind (odd-zs even-zs)
(split-list (rest (rest l)))
(list (list* (first l) odd-zs)
(list* (second l) even-zs))))))
CL-USER> (split-list '(a b c 1 2 3))
((A C 2) (B 1 3))
CL-USER> (split-list '(a b c d 1 2 3))
((A C 1 3) (B D 2))
Other approaches
I think it's worth exploring some approaches that are tail recursive, as an implementation that supports tail call optimization can convert them to loops. Tail recursive functions in Common Lisp are also typically easy to translate into do loops, which are much more likely to be implemented as iteration. In these solutions, we'll build up the result lists in reverse, and then reverse them when it's time to return them.
Recursing one element at a time
If the left and right slices are interchangeable
If it doesn't matter which of the two lists is first, you can use something like this:
(defun split-list (list &optional (odds '()) (evens '()))
(if (endp list)
(list (nreverse odds)
(nreverse evens))
(split-list (rest list)
evens
(list* (first list) odds))))
CL-USER> (split-list '(a b c 1 2 3))
((A C 2) (B 1 3))
CL-USER> (split-list '(a b c d 1 2 3))
((B D 2) (A C 1 3))
This can actually be written very concisely using a do loop, but that's typically seen as iterative, not recursive:
(defun split-list (list)
(do ((list list (rest list))
(odds '() evens)
(evens '() (list* (first list) odds)))
((endp list) (list (nreverse odds) (nreverse evens)))))
If they're not interchangable
If you always need the list containing the first element of the original list to be first, you'll need a little bit more logic. One possibility is:
(defun split-list (list &optional (odds '()) (evens '()) (swap nil))
(if (endp list)
(if swap
(list (nreverse evens)
(nreverse odds))
(list (nreverse odds)
(nreverse evens)))
(split-list (rest list)
evens
(list* (first list) odds)
(not swap))))
CL-USER> (split-list '(a b c 1 2 3))
((A C 2) (B 1 3))
CL-USER> (split-list '(a b c d 1 2 3))
((A C 1 3) (B D 2))
I think that (if swap … …) is actually a bit ugly. We can use cond so that we can get multiple forms (or if and progn), and swap the values of odds and evens before returning. I think this is actually a bit easier to read, but if you're aiming for a pure recursive solution (academic assignment?), then you might be avoiding mutation, too, so rotatef wouldn't be available, and using a when just to get some side effects would probably be frowned upon.
(defun split-list (list &optional (odds '()) (evens '()) (swap nil))
(cond
((endp list)
(when swap (rotatef odds evens))
(list (nreverse odds) (nreverse evens)))
((split-list (rest list)
evens
(list* (first list) odds)
(not swap)))))
This lends itself to do as well:
(defun split-list (list)
(do ((list list (rest list))
(odds '() evens)
(evens '() (list* (first list) odds))
(swap nil (not swap)))
((endp list)
(when swap (rotatef odds evens))
(list (nreverse odds) (nreverse evens)))))
Recursing two elements at a time
Another more direct approach would recurse down the list by cddr (i.e., (rest (rest …))) and add elements to the left and right sublists on each recursion. We need to be a little careful that we don't accidentally add an extra nil to the right list when there are an odd number of elements in the input, though.
(defun split-list (list &optional (left '()) (right '()))
(if (endp list)
(list (nreverse left)
(nreverse right))
(split-list (rest (rest list))
(list* (first list) left)
(if (endp (rest list))
right
(list* (second list) right)))))
CL-USER> (split-list '(a b c 1 2 3))
((A C 2) (B 1 3))
CL-USER> (split-list '(a b c d 1 2 3))
((A C 1 3) (B D 2))
And again, a do version:
(defun split-list (list)
(do ((list list (rest (rest list)))
(left '() (list* (first list) left))
(right '() (if (endp (rest list)) right (list* (second list) right))))
((endp list) (list (nreverse left) (nreverse right)))))
