Il CD di internet

home *** CD-ROM | disk | FTP | other *** search

/ Il CD di internet / CD.iso / SOURCE / D / CLISP / CLISPSRC.TAR / clisp-1995-01-01 / doswatcom / impnotes.txt < prev next >

Wrap

Text File | 1993-12-31 | 49.6 KB | 1,476 lines

Implementation Notes for CLISP ============================== Last modified: 21 December 1993. This implementation is mostly compatible to the standard reference Guy L. Steele Jr.: Common Lisp - The Language (1st ed.). Digital Press 1984, 465 pages. ("CLtL1" for short) and to the older parts of Guy L. Steele Jr.: Common Lisp - The Language (2nd ed.). Digital Press 1990, 1032 pages. ("CLtL2" for short) These notes document the differences of the CLISP implementation of Common Lisp to the standard CLtL1, and some implementation details. The differences between CLtL1 and CLtL2 are made up of X3J13 votes. CLISP's position with respect to these votes is listed in cltl2.txt. CHAPTER 1: Introduction ----------------------- No notes. CHAPTER 2: Data Types --------------------- All the data types are implemented: numbers, characters, symbols, lists, arrays, hash tables, readtables, packages, pathnames, streams, random states, structures and functions. 2.1.3. ------ There are four floating point types: short-float, single-float, double-float and long-float: sign mantissa exponent short-float 1 bit 16+1 bits 8 bits single-float 1 bit 23+1 bits 8 bits CLISP uses IEEE format double-float 1 bit 52+1 bits 11 bits CLISP uses IEEE format long-float 1 bit >=64 bits 32 bits The single and double float formats are those of the IEEE standard (1981), except that CLISP does not support features like +0, -0, +inf, -inf, gradual underflow, NaN, etc. (Common Lisp does not make use of these features.) Long floats have variable mantissa length, which is a multiple of 16 (or 32, depending on the word size of the processor). The default length used when long floats are read is given by the place (LONG-FLOAT-DIGITS). It can be set by (SETF (LONG-FLOAT-DIGITS) nnn), where nnn is a positive integer. 2.1.4. ------ Complex numbers can have a real part and an imaginary part of different types. For example, (SQRT -9.0) evaluates to the number #C(0 3.0), which has a real part of exactly 0, not only 0.0 (which would mean "approximately 0"). The type specifier for this is (COMPLEX INTEGER SINGLE-FLOAT), and (COMPLEX type-of-real-part type-of-imaginary-part) in general. The type specifier (COMPLEX type) is equivalent to (COMPLEX type type). 2.2.1. ------ The characters are ordered according to the ASCII encoding. More precisely, CLISP uses the IBM PC character set (code page 437): $0 $1 $2 $3 $4 $5 $6 $7 $8 $9 $A $B $C $D $E $F $00 ** ** ** ** ** ** ** ** ╢ º $10 ** ** $20 ! " # $ % & ' ( ) * + , - . / $30 0 1 2 3 4 5 6 7 8 9 : ; < = > ? $40 @ A B C D E F G H I J K L M N O $50 P Q R S T U V W X Y Z [ \ ] ^ _ $60 ` a b c d e f g h i j k l m n o $70 p q r s t u v w x y z { | } ~ $80 ╟ ⁿ Θ Γ Σ α σ τ Ω δ Φ ∩ ε ∞ ─ ┼ $90 ╔ µ ╞ ⌠ ÷ ≥ √ ∙ ╓ ▄ ó ú Ñ $A0 ß φ ≤ · ± ╤ ¬ ║ ┐ ¼ ╜ ╝ í ½ ╗ $B0 $C0 $D0 $E0 ▀ ╡ $F0 ▒ ≈ ░ ╖ ▓ Here ** are control characters, not graphic characters. (The characters left blank here cannot be represented in this character set). The following are standard characters: #\Space $20 #\Newline $0A The following are semi-standard characters: #\Backspace $08 #\Tab $09 #\Linefeed $0A #\Page $0C #\Return $0D #\Rubout $08 2.2.2. ------ #\Newline is the delimiter between lines. When writing to a file, #\Newline is converted to CR/LF. (This is the usual convention on ATARI, DOS and VMS.) For example, #\Return #\Newline is written as CR/CR/LF. When reading from a file, CR/LF is converted to #\Newline, and CR not followed by LF is read as #\Return. 2.2.3. ------ There are the following additional characters with names: #\Null $00 #\Bell $07 #\Escape $1B 2.2.4. ------ The code of a character is >=0, <256. CHAR-CODE-LIMIT = 256. There are fonts 0 to 15, and CHAR-FONT-LIMIT = 16. But the system itself uses only font 0. The following bits attributes are implemented: :CONTROL, :META, :SUPER, :HYPER. Therefore CHAR-BITS-LIMIT = 16. The system itself uses these bits only to mention special keys and Control/Alternate/Shift key status on return from (READ-CHAR *KEYBOARD-INPUT*). 2.5. ---- The maximum rank (number of dimensions) of an array is 65535 on 16-bit processors, 4294967295 on 32-bit processors. 2.13. ----- All the functions built by FUNCTION, COMPILE and the like are atoms. There are built-in functions written in C, compiled functions (both of type COMPILED-FUNCTION) and interpreted functions (of type FUNCTION). The possible function names (CLtL1 p. 59) are symbols and lambda expressions. 2.14. ----- This is the list of objects whose external representation can not be meaningfully read in: * all structures lacking a keyword constructor. * all arrays except strings, if *PRINT-ARRAY* = NIL. * #<SYSTEM-FUNCTION name> built-in function written in C * #<SPECIAL-FORM name> special form handler * #<COMPILED-CLOSURE name> compiled function, if *PRINT-CLOSURE* = NIL * #<CLOSURE name ...> interpreted function * #<FRAME-POINTER #x...> pointer to a stack frame * #<DISABLED POINTER> frame pointer which has become invalid on exit from the corresponding BLOCK or TAGBODY * #<...-STREAM ...> stream * #<PACKAGE name> package * #<HASH-TABLE #x...> hash table, if *PRINT-ARRAY* = NIL * #<READTABLE #x...> readtable * #<UNBOUND> "value" of a symbol without value, "value" of an unsupplied optional or keyword argument * #<SPECIAL REFERENCE> environment marker for variables declared SPECIAL * #<DOT> internal READ result for "." * #<END OF FILE> internal READ result, when the end of file is reached * #<READ-LABEL ...> intermediate READ result for #n# * #<ADDRESS #x...> machine address, should not occur * #<SYSTEM-POINTER #x...> should not occur 2.15. ----- The type NUMBER is the disjoint union of the types REAL and COMPLEX. (CLtL wording: "exhaustive partition") The type REAL is the disjoint union of the types RATIONAL and FLOAT. The type RATIONAL is the disjoint union of the types INTEGER and RATIO. The type INTEGER is the disjoint union of the types FIXNUM and BIGNUM. The type FLOAT is the disjoint union of the types SHORT-FLOAT, SINGLE-FLOAT, DOUBLE-FLOAT and LONG-FLOAT. CHAPTER 3: Scope and Extent --------------------------- is implemented as described. CHAPTER 4: Type Specifiers -------------------------- 4.4. ---- The CLtL2 type specifier (EQL object) denotes the singleton set {object}. 4.5. ---- The general form of the COMPLEX type specifier is (COMPLEX type-of-real-part type-of-imaginary-part). The type specifier (COMPLEX type) is equivalent to (COMPLEX type type). 4.6. ---- The CLtL2 type specifier (REAL low high) denotes the real numbers between low and high. 4.7. ---- DEFTYPE lambda lists are subject to destructuring (nested lambda lists are allowed, as in DEFMACRO) and may contain a &WHOLE marker, but no &ENVIRONMENT marker. 4.9. ---- The possible results of TYPE-OF are: CONS SYMBOL NULL FIXNUM BIGNUM RATIO SHORT-FLOAT SINGLE-FLOAT DOUBLE-FLOAT LONG-FLOAT COMPLEX CHARACTER (ARRAY element-type dimensions), (SIMPLE-ARRAY element-type dimensions) (VECTOR T size), (SIMPLE-VECTOR size) (STRING size), (SIMPLE-STRING size) (BIT-VECTOR size), (SIMPLE-BIT-VECTOR size) FUNCTION COMPILED-FUNCTION STREAM PACKAGE HASH-TABLE READTABLE PATHNAME RANDOM-STATE BYTE LOAD-TIME-EVAL SYMBOL-MACRO READ-LABEL FRAME-POINTER SYSTEM-INTERNAL ADDRESS (should not occur) any other symbol (structure types or CLOS classes) a class (CLOS classes without proper name) CHAPTER 5: Program Structure ---------------------------- 5.1.3. ------ In addition to the 24 special forms listed on p. 57 (CLtL2: p. 73), the CLtL2 special forms LOCALLY, SYMBOL-MACROLET, LOAD-TIME-VALUE are implemented, and the macros PSETQ, PROG1, PROG2, WHEN, UNLESS, COND, MULTIPLE-VALUE-LIST, MULTIPLE-VALUE-BIND, MULTIPLE-VALUE-SETQ, AND, OR are implemented as special forms. Constants may not be bound dynamically or lexically. 5.2.2. ------ LAMBDA-LIST-KEYWORDS = (&OPTIONAL &REST &KEY &ALLOW-OTHER-KEYS &AUX &BODY &WHOLE &ENVIRONMENT) LAMBDA-PARAMETERS-LIMIT is 65536 on 16-bit processors, 4294967296 on 32-bit processors. 5.3. ---- DEFUN and DEFMACRO are allowed in non-toplevel positions. As an example, consider the definition of GENSYM: (let ((gensym-prefix "G") (gensym-count 1)) (defun gensym (&optional (x nil s)) (when s (cond ((stringp x) (setq gensym-prefix x)) ((integerp x) (if (minusp x) (error "~S: index ~S is negative" 'gensym x) (setq gensym-count x) )) (t (error "~S: argument ~S of wrong type" 'gensym x)) ) ) (prog1 (make-symbol (concatenate 'string gensym-prefix (write-to-string gensym-count :base 10 :radix nil) ) ) (incf gensym-count) ) ) 5.3.2. ------ (PROCLAIM '(SPECIAL var)) declarations may not be undone. The same holds for DEFVAR, DEFPARAMETER and DEFCONSTANT declarations. It is an error if a DEFCONSTANT variable is bound at the moment the DEFCONSTANT is executed, but DEFCONSTANT does not check this. Constants may not be bound dynamically or lexically. CHAPTER 6: Predicates --------------------- 6.2.2. ------ REALP returns T is its argument is a real number, NIL otherwise. COMPILED-FUNCTION-P returns T on built-in functions written in C, compiled functions and special form handlers. Therefore COMPILED-FUNCTION is not a subtype of FUNCTION. 6.3. ---- EQ compares characters and fixnums as EQL does. No unnecessary copies are made of characters and numbers. Nevertheless, one should use EQL. (let ((x y)) (eq x x)) always returns T, regardless of y. 6.4. ---- AND and OR are implemented as special forms and, as such, rather efficient. CHAPTER 7: Control Structure ---------------------------- 7.1.1. ------ (FUNCTION symbol) returns the local function definition established by FLET or LABELS, if it exists, otherwise the global function definition. The CLtL2 place (FDEFINITION function-name) is implemented. (SPECIAL-FORM-P symbol) returns NIL or T. If it returns T, then (SYMBOL-FUNCTION symbol) returns the (useless) special form handler. 7.1.2. ------ PSETQ is implemented as a special form and, as such, rather efficient. 7.2. ---- (SETF (SYMBOL-FUNCTION symbol) object) requires object to be either a function, a SYMBOL-FUNCTION return value or a lambda expression. A lambda expression is thereby immediately converted to a function. SETF also accepts places yielding multiple values. Additional places: * FUNCALL: (SETF (FUNCALL #'symbol ...) object) and (SETF (FUNCALL 'symbol ...) object) are equivalent to (SETF (symbol ...) object). * GET-DISPATCH-MACRO-CHARACTER: (SETF (GET-DISPATCH-MACRO-CHARACTER ...) ...) performs a SET-DISPATCH-MACRO-CHARACTER. * LONG-FLOAT-DIGITS: (SETF (LONG-FLOAT-DIGITS) digits) sets the default mantissa length of long floats to digits bits. * VALUES: (SETF (VALUES place1 ... placek) form) is approximately equivalent to (MULTIPLE-VALUE-BIND (dummy1 ... dummyk) form (SETF place1 dummy1 ... placek dummyk) (VALUES dummy1 ... dummyk) ) Example: (SETF (VALUES A B) (VALUES B A)) interchanges the values of A and B. * VALUES-LIST: (SETF (VALUES-LIST list) form) is equivalent to (VALUES-LIST (SETF list (MULTIPLE-VALUE-LIST form))) &KEY markers in DEFSETF lambda lists are supported, but the corresponding keywords must appear literally in the program text. (GET-SETF-METHOD form &optional env) and (GET-SETF-METHOD-MULTIPLE-VALUE form &optional env) receives as optional argument the environment necessary for macro expansions. In DEFINE-SETF-METHOD lambda lists, one can specify &ENVIRONMENT and a variable, which will be bound to the environment. This environment should be passed to all calls of GET-SETF-METHOD and GET-SETF-METHOD-MULTIPLE-VALUE. If this is done, even local macros will be interpreted as places correctly. 7.3. ---- CALL-ARGUMENTS-LIMIT is 65536 on 16-bit processors, 4294967296 on 32-bit processors. 7.4. ---- PROG1 and PROG2 are implemented as special forms and, as such, rather efficient. 7.5. ---- The CLtL2 special form SYMBOL-MACROLET is implemented. The macro DEFINE-SYMBOL-MACRO establishes symbol macros with global scope (as opposed to symbol macros defined with SYMBOL-MACROLET, which have local scope): (DEFINE-SYMBOL-MACRO symbol expansion). Calling BOUNDP, SYMBOL-VALUE or MAKUNBOUND on symbols defined as symbol macros is not allowed. If using the optional package MACROS3: The macros LETF and LETF* are like LET and LET*, resp., except that they can bind places, even places with multiple values. Example: (LETF (((VALUES A B) form)) ...) is equivalent to (MULTIPLE-VALUE-BIND (A B) form ...) (LETF (((FIRST L) 7)) ...) is approximately equivalent to (LET* ((#:G1 L) (#:G2 (FIRST #:G1))) (UNWIND-PROTECT (PROGN (SETF (FIRST #:G1) 7) ...) (SETF (FIRST #:G1) #:G2) ) ) 7.6. ---- WHEN, UNLESS, COND are implemented as special forms and, as such, rather efficient. 7.8.4. ------ The function MAPCAP is like MAPCAN, except that it concatenates the resulting lists with APPEND instead of NCONC: (MAPCAP fun x1 ... xn) == (apply #'append (mapcar fun x1 ... xn)) (Actually a bit more efficient that this would be.) The function MAPLAP is like MAPCON, except that it concatenates the resulting lists with APPEND instead of NCONC: (MAPLAP fun x1 ... xn) = (apply #'append (maplist fun x1 ... xn)) (Actually a bit more efficient that this would be.) 7.9.1. ------ MULTIPLE-VALUES-LIMIT = 128 MULTIPLE-VALUE-LIST, MULTIPLE-VALUE-BIND, MULTIPLE-VALUE-SETQ are implemented as special forms and, as such, rather efficient. The macro NTH-VALUE: (NTH-VALUE n form) returns the (n+1)st value (n>=0) of form. CHAPTER 8: Macros ----------------- No notes. CHAPTER 9: Declarations ----------------------- 9.1. ---- The CLtL2 macro DECLAIM is implemented. 9.2. ---- The declarations (TYPE type var ...), (FTYPE type fun ...), (FUNCTION name arglist result-type), (OPTIMIZE (quality value) ...) are ignored by the interpreter and the compiler. The CLtL2 declaration (OPTIMIZE (DEBUG ...)) is legal. Additional declarations: * The declaration (COMPILE) has the effect that the current form is compiled prior to execution. Examples: (LOCALLY (DECLARE (COMPILE)) form) executes a compiled version of form. (let ((x 0)) (flet ((inc () (declare (compile)) (incf x)) (dec () (decf x))) (values #'inc #'dec) ) ) returns two functions. The first is compiled and increments x, the second is interpreted (slower) and decrements the same x. 9.3. ---- The type assertion (THE value-type form) enforces a type check in interpreted code. No type check is done in compiled code. If using the optional package MACROS3: (ETHE value-type form) enforces a type check in both interpreted and compiled code. CHAPTER 10: Symbols ------------------- No notes. CHAPTER 11: Packages -------------------- 11.6. ----- The package SYSTEM has the nicknames "SYS" and, additionally, "COMPILER". The CLtL2 packages * COMMON-LISP with nickname "CL" and * COMMON-LISP-USER with nickname "CL-USER" are implemented. The package COMMON-LISP exports only those symbols from the proposed ANSI CL draft that are actually implemented. 11.7. ----- The CLtL2 macro DEFPACKAGE is implemented. 11.8. ----- The function REQUIRE receives as optional argument either a pathname or a list of pathnames: files to be loaded if the required module is not already in memory. CHAPTER 12: Numbers ------------------- The single and double float formats are those of the IEEE standard (1981), except that CLISP does not support features like +0, -0, +inf, -inf, gradual underflow, NaN, etc. (Common Lisp does not make use of these features.) The default number of mantissa bits in long floats is given by the place (LONG-FLOAT-DIGITS). Example: (SETF (LONG-FLOAT-DIGITS) 3322) sets the default precision of long floats to 1000 decimal digits. 12.1. ----- Complex numbers can have a real part and an imaginary part of different types. If the imaginary part is EQL to 0, the number is automatically converted to a real number. (Cf. CLtL1 p. 195) This has the advantage that (let ((x (sqrt -9.0))) (* x x)) - instead of evaluting to #C(-9.0 0.0), with x = #C(0.0 3.0) - evaluates to #C(-9.0 0) = -9.0, with x = #C(0 3.0). Coercions on operations involving different types: The result of an arithmetic operation whose arguments are of different float types is rounded to the float format of the shortest (least precise) of the arguments. rational -> long float -> double float -> single float -> short float (in contrast to CLtL1 p. 195!) Rationale: See it mathematically. Add intervals: {1.0 +/- 1e-8} + {1.0 +/- 1e-16} = {2.0 +/- 1e-8} So, if we add 1.0s0 and 1.0d0, we should get 2.0s0. Shortly: Do not suggest accuracy of a result by giving it a precision that is greater than its accuracy. Example: (- (+ 1.7 pi) pi) should not return 1.700000726342836417234L0, it should return 1.7f0 (or 1.700001f0 if there were rounding errors). Experience: If in a computation using thousands of short floats, a long float (like pi) happens to be used, the long precision should not propagate throughout all the intermediate values. Otherwise, the long result would look precise, but its accuracy is only that of a short float; furthermore much computation time would be lost by calculating with long floats when only short floats would be needed. When rational numbers are to be converted to floats (due to FLOAT, COERCE, SQRT or a transcendental function), the result type is given by the variable *DEFAULT-FLOAT-FORMAT*. 12.4. ----- (LCM), called without arguments, returns 1, which is the neutral element of composition with LCM. (! n) returns the factorial of n, n a nonnegative integer. (EXQUO x y) returns the quotient x/y of two integers x,y, and checks that it is an integer. (This is more efficient than /.) (XGCD x1 ... xn) returns the values g, c1, ..., cn, where g is the greatest common divisor of the integers x1,...,xn, and c1,...,cn are integer coefficients such that g = (GCD x1 ... xn) = (+ (* c1 x1) ... (* cn xn)) 12.5.1. ------- (EXPT base exponent) is not very precise if exponent has large absolute value. (LOG number base) signals an error if base = 1. 12.5.2. ------- The value of PI is a long float with the precision given by (LONG-FLOAT-DIGITS). When this precision is changed, the value of PI is automatically recomputed. Therefore PI is a variable, not a constant. 12.6. ----- FLOAT-RADIX always returns 2. (FLOAT-DIGITS number digits) coerces `number' (a real number) to a floating point number with at least `digits' mantissa digits. The following holds: (>= (FLOAT-DIGITS (FLOAT-DIGITS number digits)) digits) 12.7. ----- BOOLE-CLR = 0 BOOLE-SET = 15 BOOLE-1 = 10 BOOLE-2 = 12 BOOLE-C1 = 5 BOOLE-C2 = 3 BOOLE-AND = 8 BOOLE-IOR = 14 BOOLE-XOR = 6 BOOLE-EQV = 9 BOOLE-NAND = 7 BOOLE-NOR = 1 BOOLE-ANDC1 = 4 BOOLE-ANDC2 = 2 BOOLE-ORC1 = 13 BOOLE-ORC2 = 11 12.10. ------ MOST-POSITIVE-FIXNUM = 2^24-1 = 16777215 MOST-NEGATIVE-FIXNUM = -2^24 = -16777216 Together with PI, the other long float constants MOST-POSITIVE-LONG-FLOAT, LEAST-POSITIVE-LONG-FLOAT, LEAST-NEGATIVE-LONG-FLOAT, MOST-NEGATIVE-LONG-FLOAT, LONG-FLOAT-EPSILON, LONG-FLOAT-NEGATIVE-EPSILON are recomputed whenever (LONG-FLOAT-DIGITS) is changed. They are variables, not constants. CHAPTER 13: Characters ---------------------- See first above: 2.2. 13.1. ----- CHAR-CODE-LIMIT = 256 CHAR-FONT-LIMIT = 16 CHAR-BITS-LIMIT = 16 13.2. ----- String-chars are those characters with font = 0 and bits = 0. The graphic characters have been described above. The standard characters are #\Newline and those graphic characters with a code between 32 and 126 (inclusive). The alphabetic characters are these string-chars: ABCDEFGHIJKLMNOPQRSTUVWXYZ abcdefghijklmnopqrstuvwxyz and the international alphabetic characters from the character set: ╟ⁿΘΓΣαστΩδΦ∩ε∞─┼╔µ╞⌠÷≥√∙ ╓▄▀ßφ≤·±╤¬║A etc. The functions CHAR-EQUAL, CHAR-NOT-EQUAL, CHAR-LESSP, CHAR-GREATERP, CHAR-NOT-GREATERP, CHAR-NOT-LESSP ignore bits and font attributes of their arguments. 13.4. ----- The string chars that are not graphic chars and the space character have names: (code-char #x00) = #\Null (code-char #x07) = #\Bell (code-char #x08) = #\Backspace = #\Rubout (code-char #x09) = #\Tab (code-char #x0A) = #\Newline = #\Linefeed (code-char #x0B) = #\Code11 (code-char #x0C) = #\Page (code-char #x0D) = #\Return (code-char #x1A) = #\Code26 (code-char #x1B) = #\Escape (code-char #x20) = #\Space 13.5. ----- CHAR-CONTROL-BIT = 1 CHAR-META-BIT = 2 CHAR-SUPER-BIT = 4 CHAR-HYPER-BIT = 8 CHAPTER 14: Sequences --------------------- 14.1. ----- The result of NREVERSE is always EQ to the argument. NREVERSE on a vector swaps pairs of elements. NREVERSE on a list swaps the first and the last element and reverses the list chaining between them. 14.2. ----- For iteration through a sequence, a macro DOSEQ, analogous to DOLIST, may be used instead of MAP : (doseq (var seqform [resultform]) {declaration}* {tag|statement}* ) 14.3. ----- REMOVE, REMOVE-IF, REMOVE-IF-NOT, REMOVE-DUPLICATES return their argument unchanged, if no element has to be removed. DELETE, DELETE-IF, DELETE-IF-NOT, DELETE-DUPLICATES destructively modify their argument: If the argument is a list, the CDR parts are modified. If the argument is a vector with fill pointer, the fill pointer is lowered and the remaining elements are compacted below the new fill pointer. 14.5. ----- SORT and STABLE-SORT have two additional keywords :START and :END : (SORT sequence predicate &key :key :start :end) (STABLE-SORT sequence predicate &key :key :start :end) SORT and STABLE-SORT are identical. They implement the mergesort algorithm. CHAPTER 15: Lists ----------------- 15.4. ----- SUBLIS and NSUBLIS apply the :KEY argument to the nodes of the cons tree and not to the keys of the alist. CHAPTER 16: Hash Tables ----------------------- 16.1. ----- MAKE-HASH-TABLE has an additional keyword :INITIAL-CONTENTS : (MAKE-HASH-TABLE &key :test :initial-contents :size :rehash-size :rehash-threshold) The :INITIAL-CONTENTS argument is an alist that is used to initialize the new hash table. The :REHASH-THRESHOLD argument is ignored. For iteration through a hash table, a macro DOHASH, analogous to DOLIST, can be used instead of MAPHASH : (dohash (key-var value-var hash-table-form [resultform]) {declaration}* {tag|statement}* ) CHAPTER 17: Arrays ------------------ 17.1. ----- ARRAY-RANK-LIMIT is 65536 on 16-bit processors, 4294967296 on 32-bit processors. ARRAY-DIMENSION-LIMIT = 2^24 = 16777216 ARRAY-TOTAL-SIZE-LIMIT = 2^24 = 16777216 17.6. ----- An array to which another array is displaced should not be shrunk (using ADJUST-ARRAY) in such a way that the other array points into void space. This is not checked at the time ADJUST-ARRAY is called! CHAPTER 18: Strings ------------------- 18.2. ----- String comparison is based on the function CHAR<=. Therefore diphtongs do not obey the usual national rules. Example: "o" < "oe" < "z" < "÷". CHAPTER 19: Structures ---------------------- 19.5. ----- The :PRINT-FUNCTION option should contain a lambda expression (lambda (structure stream depth) (declare (ignore depth)) ...) This lambda expression names a function whose task is to output the external representation of structure onto the stream. This may be done by outputting text onto the stream using WRITE-CHAR, WRITE-STRING, WRITE, PRIN1, PRINC, PRINT, PPRINT, FORMAT and the like. The following rules must be obeyed: * The value of *PRINT-ESCAPE* must be respected. * The value of *PRINT-PRETTY* should not and cannot be respected, since the pretty-print mechanism is not accessible from outside. * The value of *PRINT-CIRCLE* need not to be respected. This is managed by the system. (But the print-circle mechanism handles only those objects that are (direct or indirect) components of structure.) * The value of *PRINT-LEVEL* is respected by WRITE, PRIN1, PRINC, PRINT, PPRINT, FORMAT ~A, FORMAT ~S, FORMAT ~W and FORMAT ~D,~B,~O,~X,~R,~F,~E,~G,~$ with not-numerical arguments. Therefore the print-level mechanism works automatically if only these functions are used for outputting objects and if they are not called on objects with nesting level > 1. (The print-level mechanism does not recognize how many parentheses you have output. It only counts how many times it was called recursively.) * The value of *PRINT-LENGTH* must be respected, especially if you are outputting an arbitrary number of components. * The value of *PRINT-READABLY* must be respected. Remember that the values of *PRINT-ESCAPE*, *PRINT-LEVEL*, *PRINT-LENGTH* don't matter if *PRINT-READABLY* is true. The value of *PRINT-READABLY* is respected by PRINT-UNREADABLE-OBJECT, WRITE, PRIN1, PRINC, PRINT, PPRINT, FORMAT ~A, FORMAT ~S, FORMAT ~W and FORMAT ~D,~B,~O,~X,~R,~F,~E,~G,~$ with not-numerical arguments. Therefore *PRINT-READABLY* will be respected automatically if only these functions are used for outputting objects. * You need not bother about the values of *PRINT-BASE*, *PRINT-RADIX*, *PRINT-CASE*, *PRINT-GENSYM*, *PRINT-ARRAY*, *PRINT-CLOSURE*, *PRINT-RPARS*. The :INHERIT option is exactly like :INCLUDE except that it doesn't create new accessors for the inherited slots. Use this option to avoid the problems that occur when using the same :CONC-NAME for the new and the inherited structure. CHAPTER 20: The Evaluator ------------------------- As in Scheme, the Macro (THE-ENVIRONMENT) returns the current lexical environment. This works only in interpreted code and is not compilable! (EVAL-ENV form [env]) evaluates a form in a given lexical environment, just if the form had been part of the program text that environment came from. CHAPTER 21: Streams ------------------- 21.1. ----- *TERMINAL-IO* is not the only stream that communicates directly with the user: During execution of the body of a (WITH-KEYBOARD . body) form, *KEYBOARD-INPUT* is the stream that reads the keystrokes from the keyboard. It returns every keystroke in detail, as character with the following bits: HYPER if a non-standard key. These are: function keys, cursor keypads, numeric keypad. CHAR-CODE the Ascii code for standard keys, for non-standard keys: F1 -> #\F1, ..., F10 -> #\F10, F11 -> #\F11, F12 -> #\F12, Insert -> #\Insert, Delete -> #\Delete, Home -> #\Home, End -> #\End, PgUp -> #\PgUp, PgDn -> #\PgDn, Arrow keys -> #\Up, #\Down, #\Left, #\Right. SUPER if pressed together with Shift key(s) and if the keystroke would have been an other without Shift. CONTROL if pressed together with the Control key. META if pressed together with the Alternate key. This keyboard input is not echoed on the screen. During execution of a (WITH-KEYBOARD . body) form, no input from *TERMINAL-IO* or any synonymous stream should be requested. 21.2. ----- The macro WITH-OUTPUT-TO-PRINTER (with-output-to-printer (var) {declaration}* {form}*) binds the variable var to an output stream that sends its output to the printer. 21.3. ----- CLOSE ignores its :ABORT argument. CHAPTER 22: Input/Output ------------------------ 22.1.2. ------- A "reserved token", i.e. a token that has potential number syntax but cannot be interpreted as a number, is interpreted as symbol when being read. (CLtL1 p. 341) When a token with package markers is read, then (CLtL1 p. 343/344) no checking is done whether the package part and the symbol-name part do not have number syntax. (What's the purpose of this check?) So we consider tokens like USER:: or :1 or LISP::4711 or 21:3 as symbols. 22.1.3. ------- The backquote read macro also works when nested. Example: (eval ``(,#'(lambda () ',a) ,#'(lambda () ',b))) = (eval `(list #'(lambda () ',a) #'(lambda () ',b))) = (eval (list 'list (list 'function (list 'lambda nil (list 'quote a))) (list 'function (list 'lambda nil (list 'quote b))) ) ) Multiple backquote combinations like ,,@ or ,@,@ are not implemented. Their use would be confusing anyway. 22.1.4. ------- #\ allows inputting characters of arbitrary code: #\Code231 yields the character (code-char 231.). Additional read dispatch macros: * #Y is used to read compiled functions. * #" is used to read pathnames: #"test.lsp" is the value of (pathname "test.lsp") As in all strings, backslashes must be written twice here: #"A:\\programs\\test.lsp" 22.1.5. ------- Is it impossible to get the read macro function of a dispatch macro character like #\# using GET-MACRO-CHARACTER. The CLtL2 place READTABLE-CASE is implemented. The possible values of (READTABLE-CASE readtable) are :UPCASE, :DOWNCASE and :PRESERVE. 22.1.6. ------- In absence of SYS::WRITE-FLOAT, floating point numbers are output in radix 2. If *PRINT-READABLY* is true, *READ-DEFAULT-FLOAT-FORMAT* has no influence on the way floating point numbers are printed. Pathnames are written according to the syntax #"namestring" if *PRINT-ESCAPE* /= NIL. If *PRINT-ESCAPE* = NIL, only the namestring is printed. *PRINT-CASE* controls the output not only of symbols, but also of characters and some #<...> objects. *PRINT-PRETTY* is initially = NIL. *PRINT-ARRAY* is initially = T. An additional variable *PRINT-CLOSURE* controls whether compiled and interpreted functions (closures) are output in detailed form. If *PRINT-CLOSURE* /= NIL, compiled closures are output in #Y syntax the reader understands. *PRINT-CLOSURE* is initially = NIL. An additional variable *PRINT-RPARS* controls the output of right (closing) parentheses. If *PRINT-RPARS* /= NIL, closing parentheses which don't fit onto the same line as the the corresponding opening parenthesis are output just below their corresponding opening parenthesis, in the same column. *PRINT-RPARS* is initially = T. 22.3.1. ------- The functions WRITE and WRITE-TO-STRING have an additional keyword :CLOSURE that can be used to bind *PRINT-CLOSURE*. The CLtL2 macro PRINT-UNREADABLE-OBJECT is implemented. 22.3.3. ------- The FORMAT option ~W is analogous to ~A and ~S, but avoids binding of *PRINT-ESCAPE*. (FORMAT stream "~W" object) is equivalent to (WRITE object :stream stream). FORMAT ~R and FORMAT ~:R can output only integers in the range |n| < 10^66. The output is in English, according to the American conventions, and these conventions are identical to the British conventions only in the range |n| < 10^9. FORMAT ~:@C does not output the character itself, only the instruction how to type the character. For FORMAT ~E and FORMAT ~G, the value of *READ-DEFAULT-FLOAT-FORMAT* doesn't matter if *PRINT-READABLY* is true. FORMAT ~T can determine the current column of any stream. CHAPTER 23: File System Interface --------------------------------- 23.1. ----- For most operations, pathnames denoting files and pathnames denoting directories can not be used interchangeably. This is especially important for the functions DIRECTORY, DIR, CD, MAKE-DIR, DELETE-DIR. The minimum filename syntax that may be used portably is: "xxx" for a file with name xxx, "xxx.yy" for a file with name xxx and type yy, ".yy" for a pathname with type yy and no name specified. Hereby xxx denote 1 to 8 characters, and yy denote 1 to 3 characters, each of which being either alphanumerical or the underscore #\_. Other properties of pathname syntax vary between operating systems. 23.1.1. ------- Pathname components: HOST always NIL DEVICE NIL or :WILD or "A"|...|"Z" DIRECTORY (startpoint . subdirs) where startpoint = :RELATIVE | :ABSOLUTE subdirs = () | (subdir . subdirs) subdir = :CURRENT (means ".") or subdir = :PARENT (means "..") or subdir = :WILD (means "...", all subdirectories) or subdir = (name . type) name = :WILD or a simple string with 8 characters maximum type = :WILD or a simple string with 3 characters maximum NAME NIL or :WILD or a simple string with 8 characters maximum TYPE NIL or :WILD or a simple string with 3 characters maximum VERSION always NIL (may also be specified as :WILD or :NEWEST) When a pathname is to be fully specified (no wildcards), that means that no :WILD is allowed, and NAME = NIL may not be allowed either. External notation: A:\sub1.typ\sub2.typ\name.typ using defaults: \sub1.typ\sub2.typ\name.typ or name.typ or *:\sub1.typ\*.*\name.* or similar. Instead of '\' one may use '/', as usual for DOS calls. 23.1.2. ------- External notation of pathnames (cf. PARSE-NAMESTRING and NAMESTRING), of course without spaces, [,],{,}: [ [drivespec] a letter '*'|'A'|...|'Z'|'a'|...|'z' : ] { name [. type] \ } each one a subdirectory, '\' may be replaced by '/' [ name [. type] ] filename with type (extension) Name and type may be character sequences of any length (consisting of alphanumeric characters and '-', '_'). They are shortened to 8 resp. 3 characters and converted to upper case. A single '*' is allowed for :WILD. The function USER-HOMEDIR-PATHNAME is not implemented. 23.2. ----- The file streams returned by OPEN are always buffered. 23.3. ----- FILE-AUTHOR always returns NIL. FILE-POSITION works on any file stream. When a Newline is output to resp. input from a file stream, its file position is increased by 2 since Newline is encoded as CR/LF in the file. 23.4. ----- LOAD has two additional keywords :ECHO and :COMPILING. (LOAD filename &key :verbose :print :echo :if-does-not-exist :compiling) :VERBOSE T causes LOAD to emit a short message that a file is being loaded. The default is *LOAD-VERBOSE*, which is initially = T. :PRINT T causes LOAD to print the value of each form. The default is *LOAD-PRINT*, which is initially = NIL. :ECHO T causes the input from the file to be echoed to *STANDARD-OUTPUT* (normally to the screen). Should there be an error in the file, you can see at one glance where it is. The default is *LOAD-ECHO*, which is initially = NIL. :COMPILING T causes each form read to be compiled on the fly. The compiled code is executed at once and - in contrast to COMPILE-FILE - not written to a file. The CLtL2 variables *LOAD-PATHNAME* and *LOAD-TRUENAME* are implemented. The variable *LOAD-PATHS* contains a list of directories where program files are searched - additionally to the specified or current directory - by LOAD, REQUIRE, COMPILE-FILE. 23.5. ----- (DIRECTORY [pathname [:full] [:circle]]) can run in two modes: * If pathname contains no name or type component, a list of all matching directories is produced. * Otherwise a list of all matching files is returned. If the :FULL argument is /= NIL, this contains additional information: for each matching file you get a list of at least four elements (file-pathname file-truename file-write-date-as-decoded-time file-length). (DIR [pathname]) is like DIRECTORY, but displays the pathnames instead of returning them. (DIR) shows the contents of the current directory. (CD [pathname]) manages the current device and the current directory. (CD pathname) sets it, (CD) returns it. (DEFAULT-DIRECTORY) is equivalent to (CD), (SETF (DEFAULT-DIRECTORY) pathname) is equivalent to (CD pathname). (MAKE-DIR directory-pathname) creates a new subdirectory. (DELETE-DIR directory-pathname) removes an (empty) subdirectory. (EXECUTE programfile arg1 arg2 ...) executes an external program. Its name is programfile. It is given the strings arg1, arg2, ... as arguments. (SHELL [command]) calls the operating system's shell. (SHELL) calls the shell for interactive use. (SHELL command) calls the shell only for execution of the one given command. CHAPTER 24: Errors ------------------ 24.1. ----- When an error occurred, you are in a break loop. You can evaluate forms as usual. The HELP command (or help key if there is one) lists the available debugging commands. CHAPTER 25: Miscellaneous Features ---------------------------------- 25.1. ----- The compiler can be called not only by the functions COMPILE, COMPILE-FILE and DISASSEMBLE, also by the declaration (COMPILE). (COMPILE-FILE input-file [:output-file] [:listing] [:warnings] [:verbose] [:print]) compiles a file to bytecode. input-file should be a pathname/string/symbol. The :output-file argument should be NIL or T or a pathname/string/symbol or an output-stream. The default is T. The :listing argument should be NIL or T or a pathname/string/symbol or an output-stream. The default is NIL. The :warnings argument specifies whether warnings should also appear on the screen. The :verbose argument specifies whether error messages should also appear on the screen. The :print argument specifies whether an indication which forms are being compiled should appear on the screen. The variables *COMPILE-WARNINGS*, *COMPILE-VERBOSE*, *COMPILE-PRINT* provide defaults for the :warnings, :verbose, :print keyword arguments, respectively. The CLtL2 variables *COMPILE-FILE-PATHNAME* and *COMPILE-FILE-TRUENAME* are implemented. The CLtL2 special form LOAD-TIME-VALUE is implemented. (LOAD-TIME-VALUE form) is like (QUOTE #,form) except that the former can be generated by macros. The CLtL2 function FUNCTION-LAMBDA-EXPRESSION is implemented. (FUNCTION-LAMBDA-EXPRESSION function) returns information about the source of an interpreted function: lambda-expression, lexical environment, name. 25.2. ----- No on-line documentation is available for the system functions (yet). 25.3. ----- (TRACE fun ...) makes the functions fun, ... traced. Syntax of fun: Either a symbol: symbol or a list of a symbol and some keywords and arguments (which must come in pairs!): (symbol [:suppress-if form] ; no trace output as long as form is true [:step-if form] ; invokes the stepper as soon as form is true [:pre form] ; evaluates form before calling the function [:post form] ; evaluates form after return from the function [:pre-break-if form] ; goes into the break loop before calling the ; function if form is true [:post-break-if form] ; goes into the break loop after return from ; the function if form is true [:pre-print form] ; prints the values of form before calling the ; function [:post-print form] ; prints the values of form after return from ; the function [:print form] ; prints the values of form both before ; calling and after return from the function ) In all these forms you can access the function itself as *TRACE-FUNCTION*, the arguments to the function as *TRACE-ARGS*, the function/macro call as form as *TRACE-FORM*, and after return from the function the list of return values from the function call as *TRACE-VALUES*, and you can leave the function call with specified values by using RETURN. TRACE and UNTRACE are also applicable to functions (SETF symbol) and to macros, but not to locally defined functions and macros. The function INSPECT is not implemented. The function ROOM can only be called without arguments. It returns two values: the number of bytes currently occupied by Lisp objects, and the number of bytes that can be allocated before the next regular garbage collection occurs. The function ED calls the external editor specified by the variable *EDITOR* (see config.lsp). If using the optional package EDITOR: ED behaves like this only if the variable *USE-ED* is NIL. Otherwise ED uses an Emacs-like screen editor with multiple windows. 25.4.1. ------- The variable *DEFAULT-TIME-ZONE* contains the default time zone used by ENCODE-UNIVERSAL-TIME and DECODE-UNIVERSAL-TIME. It is initially set to -1 (which means 1 hour east of Greenwich, i.e. Mid European Time). The timezone in a decoded time must not necessarily be an integer, but (as float or rational number) it should be a multiple of 1/4. INTERNAL-TIME-UNITS-PER-SECOND = 100. 25.4.2. ------- The functions MACHINE-TYPE, MACHINE-VERSION, MACHINE-INSTANCE and SHORT-SITE-NAME, LONG-SITE-NAME should be defined by every user in his site-specific CONFIG.LSP file. The variable *FEATURES* initially contains the symbols CLISP ; this implementation COMMON-LISP CLTL1 INTERPRETER COMPILER CLOS ATARI ; if hardware = Atari ST/TT and operating system = TOS AMIGA ; if hardware = Amiga and operating system = Exec/AmigaDOS DOS ; if hardware = PC (clone) and operating system = DOS OS/2 ; if hardware = PC (clone) and operating system = OS/2 PC386 ; if hardware = PC (clone) with a 386/486 VMS ; if hardware = VAX and operating system = VMS UNIX ; if operating system = Unix ; (yes, in this case the ; hardware is irrelevant!) language ; same as the value of *LANGUAGE* The constant *LANGUAGE* is a string containing the language in which the system communicates with the user. A symbol of the same name is contained in *FEATURES*. Possible values are (yet): ENGLISH, DEUTSCH, FRANCAIS. CHAPTER 28: Common Lisp Object System ------------------------------------- To use CLOS, do (USE-PACKAGE "CLOS"). The functions SLOT-VALUE, SLOT-BOUNDP, SLOT-MAKUNBOUND, SLOT-EXISTS-P, FIND-CLASS, (SETF FIND-CLASS), CLASS-OF, CALL-NEXT-METHOD, NEXT-METHOD-P, CLASS-NAME, (SETF CLASS-NAME), NO-APPLICABLE-METHOD, NO-NEXT-METHOD, FIND-METHOD, ADD-METHOD, REMOVE-METHOD, COMPUTE-APPLICABLE-METHODS, METHOD-QUALIFIERS, FUNCTION-KEYWORDS, SLOT-MISSING, SLOT-UNBOUND, PRINT-OBJECT, DESCRIBE-OBJECT, MAKE-INSTANCE, INITIALIZE-INSTANCE, REINITIALIZE-INSTANCE, SHARED-INITIALIZE, the macros WITH-SLOTS, WITH-ACCESSORS, DEFCLASS, DEFMETHOD, DEFGENERIC, GENERIC-FUNCTION, GENERIC-FLET, GENERIC-LABELS, the classes STANDARD-CLASS, STRUCTURE-CLASS, BUILT-IN-CLASS, STANDARD-OBJECT, STANDARD-GENERIC-FUNCTION, STANDARD-METHOD and all predefined classes, and the method combination STANDARD are implemented. Deviations from CLtL2 chapter 28: DEFCLASS : It *is* required that the superclasses of a class be defined before the DEFCLASS form for the class is evaluated. The REAL type is added to the predefined classes listed in table 28-1. Only STANDARD method combination is implemented. CALL-NEXT-METHOD cannot be called with arguments. CALL-NEXT-METHOD and NEXT-METHOD-P are local macros, not local functions. Use #'(lambda () (call-next-method)) instead of #'call-next-method if you really need it as a function. There is a generic function NO-PRIMARY-METHOD (analogous to NO-APPLICABLE-METHOD) which is called when a generic function of the class STANDARD-GENERIC-FUNCTION is invoked and no primary method on that generic function is applicable. GENERIC-FLET and GENERIC-LABELS are implemented as macros, not as special forms. The function ENSURE-GENERIC-FUNCTION is not implemented. ADD-METHOD can put methods into other generic functions than the one the method came from. PRINT-OBJECT and DESCRIBE-OBJECT are only called on objects of type STANDARD-OBJECT. DESCRIBE-OBJECT should not call DESCRIBE recursively as this would produce more information than is likely to be useful to a human reader. DOCUMENTATION still has the CLtL1 implementation. User-defined method combination is not supported. The sections 28.1.7.3., 28.1.7.4., the macros DEFINE-METHOD-COMBINATION, CALL-METHOD and the functions INVALID-METHOD-ERROR, METHOD-COMBINATION-ERROR, METHOD-QUALIFIERS are not implemented. The special form WITH-ADDED-METHODS is not implemented. Redefining classes is not supported. The sections 28.1.10., 28.1.10.1., 28.1.10.2., 28.1.10.3., 28.1.10.4. and the function UPDATE-INSTANCE-FOR-REDEFINED-CLASS are not implemented. Changing the class of a given instance is not supported. The sections 28.1.11., 28.1.11.1., 28.1.11.2., 28.1.11.3. and the functions CHANGE-CLASS, UPDATE-INSTANCE-FOR-DIFFERENT-CLASS, MAKE-INSTANCES-OBSOLETE are not implemented. CHAPTER 90: Platform independent Extensions ------------------------------------------- 90.1. Saving an Image --------------------- The function (SAVEINITMEM [filename [:quiet] [:init-function]]) saves the running CLISP's memory to a file. The filename defaults to "lispinit.mem". If the :QUIET argument is not NIL, the startup banner and the good-bye message will be suppressed. The :INIT-FUNCTION argument specifies a function that will be executed at startup of the saved image. 90.2. Quitting Lisp ------------------- The functions (EXIT [errorp]), (QUIT [errorp]) and (BYE [errorp]) - all synonymous - terminate CLISP. If errorp is not NIL, CLISP aborts with error status, i.e. the environment is informed that the CLISP session didn't succeed. CHAPTER 99: Platform specific Extensions ---------------------------------------- 99.2. Random Screen Access -------------------------- (SCREEN:MAKE-WINDOW) returns a "window stream". As long as this stream is open, the terminal is in cbreak/noecho mode. *TERMINAL-IO* shouldn't be used for input or output during this time. (Use WITH-KEYBOARD and *KEYBOARD-INPUT* instead.) (SCREEN:WITH-WINDOW . body) binds SCREEN:*WINDOW* to a window stream and executes body. The stream is guaranteed to be closed when the body is left. During its execution, *TERMINAL-IO* shouldn't be used, as above. (SCREEN:WINDOW-SIZE window-stream) returns the window's size, as two values: height (= Ymax+1) and width (= Xmax+1). (SCREEN:WINDOW-CURSOR-POSITION window-stream) returns the position of the cursor in the window, as two values: line (>=0, <=Ymax, 0 means top), column (>=0, <=Xmax, 0 means left margin). (SCREEN:SET-WINDOW-CURSOR-POSITION window-stream line column) sets the position of the cursor in the window. (SCREEN:CLEAR-WINDOW window-stream) clears the window's contents and puts the cursor in the upper left corner. (SCREEN:CLEAR-WINDOW-TO-EOT window-stream) clears the window's contents from the cursor position to the end of window. (SCREEN:CLEAR-WINDOW-TO-EOL window-stream) clears the window's contents from the cursor position to the end of line. (SCREEN:DELETE-WINDOW-LINE window-stream) removes the cursor's line, moves the lines below it up by one line and clears the window's last line. (SCREEN:INSERT-WINDOW-LINE window-stream) inserts a line at the cursor's line, moving the lines below it down by one line. (SCREEN:HIGHLIGHT-ON window-stream) switches highlighted output on. (SCREEN:HIGHLIGHT-OFF window-stream) switches highlighted output off. (SCREEN:WINDOW-CURSOR-ON window-stream) makes the cursor visible, a cursor block in most implementations. (SCREEN:WINDOW-CURSOR-OFF window-stream) makes the cursor invisible, in implementations where this is possible. Authors: -------- Bruno Haible Michael Stoll Augartenstra▀e 40 Gallierweg 39 D - 76137 Karlsruhe D - 53117 Bonn Germany Germany Email: haible@ma2s2.mathematik.uni-karlsruhe.de