The Logo Language
The following is taken from a April 2003 post by Brian Harvey. Differences with PyLogo are noted in sidebars.
PyLogo doesn't use property lists; lists themselves are mutable (like Python lists), and Python dictionaries can also be created and manipulated. These are first-class objects, unlike Logo property lists.
Name space organization: Logo has separate name spaces for procedures, variables, and property lists. Variables are dynamically scoped; procedures and property lists are all global. (Not all versions of Logo include property lists.)
PyLogo can use and manipulate any Python data type, so the possible number of types is much larger. Also, lists actually arrays (Python's list type), and are mutable (though most of the Logo builtins don't mutate the lists). TRUE and FALSE are special values.
Data organization: I'm not quite sure what you mean by this, but I'm guessing that you're asking about aggregate data types. Logo's main types are word and list. A word can be divided into characters, but "character" is not a separate data type -- they're just one-letter words. A number is a special case of a word, although of course internally they're represented in a form that allows for fast arithmetic. Lists are just like Lisp lists: singly-linked. Most versions of Logo do not allow list mutation (which is why property lists are important; those are mutable but not sharable). Some versions of Logo, but not all, include arrays. A few versions include objects (sometimes in the form of first-class turtles that can have local variables and/or methods). For some purposes, Logo recognizes the special case of a "sentence," which is a (flat) list of words. The words TRUE and FALSE serve as Booleans.
Syntactic structural organization: There's no official BNF description of Logo, because there's too much variation among dialects, and most of us feel that allowing for experimentation is more important than uniformity, in a language that's never going to take over the world in any case!
Basically, everything is done with procedure calls. There are no special forms except for TO, which is used to define procedures. Quotation is similar in spirit to Lisp, but different in notation:
PyLogo doesn't have a string literal representation currently, besides the single-word representation.
|(foo 3 4)||foo 3 4|
|'(a b c)||[a b c]|
|"foo baz"||"|foo baz| (in some dialects) or "foo\ baz|
Parentheses are not generally needed for procedure calls because most procedures take a fixed number of arguments (and all procedures take a default number of arguments if parentheses are omitted). But they are allowed for clarity or to give a non-default number of arguments to those procedures that allow it.
Arithmetic operators come in prefix and infix form:
3+4 sum 3 4
PyLogo does not require spaces.
(Some dialects require spaces around the infix operators: 3 + 4.)
The notation :FOO is an abbreviation for THING "FOO. The primitive procedure THING takes a variable name as argument and returns its value. Note that the colon abbreviates THING-quote, not just THING, so ::FOO doesn't do what you might hope. The notation "FOO isn't an abbreviation for anything; there is no QUOTE special form.
Here's how TO works:
to poly :num :size repeat :num [forward :size right 360/:num] end
This defines a procedure named POLY that has formal parameters NUM and SIZE. I believe every Logo dialect allows this form; a few also allow variants such as:
to poly num size to poly "num "size
PyLogo always allows rest args, and puts them in a special :REST variable. If a variable is omitted, it is always set to None. This will probably change in the future. The trailing integer form is allowed. num and "num are also allowed.
Some dialects allow optional and rest args:
to foo :a :b :c [:d 43] [:e [foo baz]] [:f]
where A, B, and C are required args, D and E are optional args (with the indicated default values if omitted), and F is a rest arg (meaning that any number of arguments starting from the sixth are put into a list and assigned to F). Procedure FOO will have a default number of arguments of 3; this can be changed by putting a number at the end of the title line:
to foo :a :b :c [:d 43] [:e [foo baz]] [:f] 5
By the way, note that REPEAT (a primitive used in the POLY example) is not a special form even though it's a control structure. One of its arguments is a list -- a quoted list, in this example, and in most uses of control structures, but like any argument it is computed from whatever expression is provided. Since Logo is dynamically scoped, a procedure is nothing more than its text, and so a quoted instruction list is just as good as a LAMBDA expression would be if Logo had LAMBDA.
We allow list mutation, and with a richer set of objects there's lots of ways circular constructs can be created. But Python's GC does all the work for us.
Since most versions don't allow list mutation, circular lists are impossible, and so in principle a reference counting storage management system would work. But I think most implementations use standard Lispish garbage collectors.
PyLogo is a simple interpreter, and doesn't do any optimization at this time. Procedures are tokenized when they are defined, but they are not fully parsed into expessions until they are run, and no information is cached.
Most implementations are interpreters, rather than compilers, but most achieve some speedup by parsing instruction lists into Lispish list structure the first time a procedure is called. (It can't be done when the procedure is defined, because a procedure can be defined before its helper procedures are defined, and so Logo doesn't know yet how many arguments each procedure takes.)
Python lists, which PyLogo uses, allow constant time access. Procedures like FPUT do not modify their arguments, so they end up copying the list; lists are not shared, and link lists are not used, so the performance behaves differently than in most Logos.
Although lists are singly-linked, the constructors and selectors treat the two ends symmetrically. For example, both FIRST and LAST are provided, even though FIRST is Theta(1) time and LAST is Theta(N) time.