List Types and Internal Structure

This chapter gives some more in-depth details about how Scheme lists are constructed, and what implications this has on their use. Please note that this is a somewhat advanced topic and you will probably not come across it too early in your learning curve with Scheme in LilyPond. However, there are occasions where these characteristics can appear at the surface with nastily confusing situations which can best be handled with a proper understanding of the underlying structures.

Maybe it is a good idea to have at least a cursory glance at the chapter, just to know what it is about. Once you run into a corresponding issue you'll know where to search for further information.

How Scheme Lists Are constructed

In the previous chapter we saw that lists are entered as a sequence of elements separated by spaces. Lists are also displayed that way, but this is not how they are organized internally.

A list in Scheme is a pair whose cdr is the rest of the list. The car of this rest is the next list element and its cdr the further remainder of the list. In fact this structure is also known as linked list, where the elements don't know about the list as a whole but are only linked to their subsequent element. As a consequence you can say that any list is also a pair but that a pair is not necessarily a list:

guile> (define lst (list 1 2 3 4))
guile> (define pr (cons 5 6))
guile> lst
(1 2 3 4)
guile> p
(5 . 6)
guile> (list? lst)
#t
guile> (list? p)
#f
guile> (pair? lst)
#t
guile> (pair? p)
#t

Let's inspect the list lst a little closer:

guile> (car lst)
1
guile> (cdr lst)
(2 3 4)

The car of the list is 1 and the cdr is the list (2 3 4), the car of that remainder is 2 and the cdr the list (3 4). The car of this remainder is 3 and the cdr - the list (4).

This last information may be confusing, but on second thought it is quite natural: The cdr of a list element is a list, and so the last element has to be a list as well. But what is this one-element list? As we defined earlier a list is a pair whose cdr is a list. So let us inspect this one-element list individually:

guile> (define ll (list 4))
guile> ll
(4)
guile> (car ll)
4
guile> (cdr ll)
()

Now we see that the car of the last element is the value 4, while the cdr is an empty list. This kind of list - where the last element's cdr is an empty list - is called a proper list (we'll see improper lists below). One consequence of this can be somewhat confusing: whenever you have to retrieve the last element of a proper list you don't really look for the “last element” but rather “the car of the last element”. We will return to this issue in the next chapter.

Constructing Lists As Chained Pairs

If we take our definition literally that a list is a chain of pairs it should be possible to create a list that way as well:

guile> (define pl
  (cons 1
    (cons 2
      (cons 3
        (cons 4 '())))))
guile> pl
(1 2 3 4)

What happens here is that we define pl to be a pair of 1 and a pair of 2 and a pair of 3 and a pair of 4 and an empty list. The same is possible using the literal notation:

guile> '(1 . (2 . (3 . (4 . ()))))
(1 2 3 4)

Improper Lists

What happens if the last element of a list is not a pair with an empty list as car but a simple value? This is not hypothetical but easily achievable:

guile> '(1 . (2 . (3 . 4)))
(1 2 3 . 4)

This is Scheme's syntax for an improper list, where the last element's cdr is a value instead of a pair. The dot is an indicator of the last element's nature as a pair between the 3 and 4 instead of between the 4 and an invisible trailing element. Such improper lists can also be created using cons and written as literals:

guile> (cons 1 (cons 2 (cons 3 4)))
(1 2 3 . 4)
guile> '(1 2 3 . 4)
(1 2 3 . 4)

Concatenating Lists

Maybe this section is more about working with lists, but we insert it here because it provides more insight on how lists are structured internally.

The procedure append takes a list and appends another list at its end, as can be seen like this:

guile> (define lst (list 1 2 3 4))
guile> lst
(1 2 3 4)
guile> (append lst (list 5 6))
(1 2 3 4 5 6)

From the user's perspective we have a list (1 2 3 4) and add the elements 5 and 6 to it. But we have to understand this from what we have learnt above. What append really does is replace the empty list that is the cdr of the first list's last element with the second argument, here (5 6). So what originally was '(4 . ()) has now become '(4 . (5 . (6 . ()))). And as this seamlessly integrates with the construction of chained pairs the overall result is a coherent list (1 2 3 4 5 6). Let's see what happens if we don't append a list but instead a pair:

guile> (append lst (cons 5 6))
(1 2 3 4 5 . 6)

The result is an improper list, and by now we can easily explain why this is the case: we have replaced the trailing empty list with a pair whose cdr is not a list but instead a plain value.

The same is true if we append a literal value to the list:

guile> (append lst 5)
(1 2 3 4 . 5)

This time we have directly replaced the cdr of the last element with a literal value, with the same effect of turning the list into an improper list.

Finally we check what happens when we try to append anything to the resulting improper list:

guile> (define lst2 (append lst 5))
guile> lst2
(1 2 3 4 . 5)
guile> (append lst2 6)
standard input:10:1: In procedure append in expression (append l2 6):
standard input:10:1: Wrong type argument in position 1 (expecting empty list): 5
ABORT: (wrong-type-arg)

This doesn't work, and also here we are now able to explain why: we said that append replaces the empty list in the last element's cdr with the appended value. But the improper list does not have such a trailing empty list, instead the cdr is 5, and therefore append can't do its work.


Last update: January 31, 2020