Today I want to present you a utility library, collected and written by Mike Travers. It’s name is “mtlisp” and it consist of different functionality, collected into one package.
I’ve contacted Mike and he suggested two the most interesting features of “mtlisp” system.
First interesting part is “ctrace”. It solves the problem of interleaved output which you’ll have when working with standard trace, builtin into Lisp REPL.
Let’s pretend we have a function we’d like to trace but it prints something to stdout and breaks our nice traces, making them hard to read:
POFTHEDAY> (defun fib (n &optional (a 0) (b 1) (acc ()))
(format t "Producing fibbonachi for ~A~%"
n)
(if (zerop n)
(nreverse acc)
(fib (1- n)
b
(+ a b)
(cons a acc))))
POFTHEDAY> (trace fib)
(FIB)
POFTHEDAY> (fib 5)
0: (POFTHEDAY::FIB 5)
Producing fibbonachi for 5
1: (POFTHEDAY::FIB 4 1 1 (0))
Producing fibbonachi for 4
2: (POFTHEDAY::FIB 3 1 2 (1 0))
Producing fibbonachi for 3
3: (POFTHEDAY::FIB 2 2 3 (1 1 0))
Producing fibbonachi for 2
4: (POFTHEDAY::FIB 1 3 5 (2 1 1 0))
Producing fibbonachi for 1
5: (POFTHEDAY::FIB 0 5 8 (3 2 1 1 0))
Producing fibbonachi for 0
5: FIB returned (0 1 1 2 3)
4: FIB returned (0 1 1 2 3)
3: FIB returned (0 1 1 2 3)
2: FIB returned (0 1 1 2 3)
1: FIB returned (0 1 1 2 3)
0: FIB returned (0 1 1 2 3)
(0 1 1 2 3)
Ctrace collects traces into a separate data structure and prints it after the function execution. Probably it also has other features.
The problem is that it works only in AllegroCL, CCL and ABCL, because uses function advices.
I have problems with running ClozureCL and AllegroCL on latest OSX, and mtlisp depends on “acl-compat” which can’t be loaded into ABCL. Also, I encounter other problems trying to load the library into ABCL :(
However, I found it is possible to outcome standard trace’s behaviour with rebinding trace-output stream:
POFTHEDAY> (with-output-to-string (*trace-output*)
(fib 5))
Producing fibbonachi for 5
Producing fibbonachi for 4
Producing fibbonachi for 3
Producing fibbonachi for 2
Producing fibbonachi for 1
Producing fibbonachi for 0
" 0: (POFTHEDAY::FIB 5)
1: (POFTHEDAY::FIB 4 1 1 (0))
2: (POFTHEDAY::FIB 3 1 2 (1 0))
3: (POFTHEDAY::FIB 2 2 3 (1 1 0))
4: (POFTHEDAY::FIB 1 3 5 (2 1 1 0))
5: (POFTHEDAY::FIB 0 5 8 (3 2 1 1 0))
5: FIB returned (0 1 1 2 3)
4: FIB returned (0 1 1 2 3)
3: FIB returned (0 1 1 2 3)
2: FIB returned (0 1 1 2 3)
1: FIB returned (0 1 1 2 3)
0: FIB returned (0 1 1 2 3)
"
Another interesting feature of “mtlisp” is “closstar” package. It combines features of CLOS with its predecessor - Flavors.
This package is based on previous Mike’s work Artificial Flavors made in 88-89 years. I was only 8 years old, when this man wrote such interesting things!
Defclass* macro allows to define classes in a more concise way, almost like you do with defstruct:
POFTHEDAY> (closstar:defclass* blah ()
(foo
(bar nil)))
#<STANDARD-CLASS POFTHEDAY::BLAH>
POFTHEDAY> (describe *)
#<STANDARD-CLASS POFTHEDAY::BLAH>
[standard-object]
Direct superclasses: STANDARD-OBJECT
No subclasses.
Not yet finalized.
Direct slots:
FOO
BAR
POFTHEDAY> (defvar *v* (make-instance 'blah))
*V*
POFTHEDAY> (slot-boundp *v* 'foo)
NIL
POFTHEDAY> (slot-boundp *v* 'bar)
T
POFTHEDAY> (slot-value *v* 'bar)
NIL
Besides brevity, “defclass*” has other interesting features. Take a look at its definition to learn more.
Closstar has it’s own counterpart of “defmethod” - “defmethod*”:
POFTHEDAY> (closstar:defmethod* describe-this ((self blah))
(format t "A BLAH object with bar=~A~%"
bar))
POFTHEDAY> (describe-this *v*)
A BLAH object with bar=NIL
NIL
What is interesting here is that inside method body you can access all object’s slots without having to call slot-value or accessor explicitly.
That is because “defmethod*” calls “with-slots” under the hood:
(defmethod describe-this ((self blah))
(with-slots (bar foo)
self
(declare (ignorable bar foo))
(format t "A BLAH object with bar=~A~%" bar)))
To conclude, “mtlisp” is unsupported now but might include hidden gems. Perhaps one day there will be a craftsman who will cut them.