abu@software-lab.de

Perfection is attained
not when there is nothing left to add
but when there is nothing left to take away
(Antoine de Saint-Exupéry)

The Pico Lisp Reference

(c) Software Lab. Alexander Burger

This document describes the concepts, data types, and kernel functions of the Pico Lisp system.

This is not a Lisp tutorial. For an introduction to Lisp, a traditional Lisp book like "Lisp" by Winston/Horn (Addison-Wesley 1981) is recommended. Note, however, that there are significant differences between Pico Lisp and Maclisp (and even greater differences to Common Lisp).

Please take a look at the Pico Lisp Tutorial for an explanation of some aspects of Pico Lisp, and scan through the list of Frequently Asked Questions (FAQ).


Introduction

Pico Lisp is the result of a language design study, trying to answer the question "What is a minimal but useful architecture for a virtual machine?". Because opinions differ about what is meant by "minimal" and "useful", there are many answers to that question, and people might consider other solutions more "minimal" or more "useful". But from a practical point of view, Pico Lisp has proven to be a valuable answer to that question.

First of all, Pico Lisp is a virtual machine architecture, and then a programming language. It was designed in a "bottom up" way, and "bottom up" is also the most natural way to understand and to use it: Form Follows Function.

Pico Lisp has been used in several commercial and research programming projects since 1988. Its internal structures are simple enough, allowing an experienced programmer always to fully understand what's going on under the hood, and its language features, efficiency and extensibility make it suitable for almost any practical programming task.

In a nutshell, emphasis was put on four design objectives. The Pico Lisp system should be

Simple
The internal data structure should be as simple as possible. Only one single data structure is used to build all higher level constructs.
Unlimited
There are no limits imposed upon the language due to limitations of the virtual machine architecture. That is, there is no upper bound in symbol name length, number digit counts, stack depth, or data structure and buffer sizes, except for the total memory size of the host machine.
Dynamic
Behavior should be as dynamic as possible ("run"-time vs. "compile"-time). All decisions are delayed till runtime where possible. This involves matters like memory management, dynamic symbol binding, and late method binding.
Practical
Pico Lisp is not just a toy of theoretical value. It is in use since 1988 in actual application development, research and production.


The Pico Lisp Machine

An important point in the Pico Lisp philosophy is the knowledge about the architecture and data structures of the internal machinery. The high-level constructs of the programming language directly map to that machinery, making the whole system both understandable and predictable.

This is similar to assembly language programming, where the programmer has complete control over the machine.


The Cell

The Pico Lisp virtual machine is both simpler and more powerful than most current (hardware) processors. At the lowest level, it is constructed from a single data structure called "cell":


         +-----+-----+
         | CAR | CDR |
         +-----+-----+

A cell is a pair of machine words, which traditionally are called CAR and CDR in the Lisp terminology. These words can represent either a numeric value (scalar) or the address of another cell (pointer). All higher level data structures are built out of cells.

The type information of higher level data is contained in the pointers to these data. Assuming the implementation on a byte-addressed physical machine, and a pointer size of typically 4 bytes, each cell has a size of 8 bytes. Therefore, the pointer to a cell must point to an 8-byte boundary, and its bit-representation will look like:


      xxxxxxxxxxxxxxxxxxxxxxxxxxxxx000

(the 'x' means "don't care"). For the individual data types, the pointer is adjusted to point to other parts of a cell, in effect setting some of the lower three bits to non-zero values. These bits are then used by the interpreter to determine the data type.

In any case, bit(0) - the least significant of these bits - is reserved as a mark bit for garbage collection.

Initially, all cells in the memory are unused (free), and linked together to form a "free list". To create higher level data types at runtime, cells are taken from that free list, and returned by the garbage collector when they are no longer needed. All memory management is done via that free list; there are no additional buffers, string spaces or special memory areas (With two exceptions: A certain fixed area of memory is set aside to contain the executable code and global variables of the interpreter itself, and a standard push down stack for return addresses and temporary storage. Both are not directly accessible by the programmer).


Data Types

On the virtual machine level, Pico Lisp supports

They are all built from the single cell data structure, and all runtime data cannot consist of any other types than these three.

The following diagram shows the complete data type hierarchy, consisting of the three base types and the symbol variations:


                    cell
                     |
            +--------+--------+
            |        |        |
         Number    Symbol    List
                     |
                     |
   +--------+--------+--------+
   |        |        |        |
  NIL   Internal Transient External


Numbers

A number can represent a signed integral value of arbitrary size. The CARs of one or more cells hold the number's "digits" (each in the machine's word size), to store the number's binary representation.


         Number
         |
         V
      +-----+-----+     +-----+-----+     +-----+-----+
      |'DIG'|  ---+---> |'DIG'|  ---+---> |'DIG'|  /  |
      +-----+-----+     +-----+-----+     +-----+-----+

The first cell holds the least significant digit. The least significant bit of that digit represents the sign.

The pointer to a number points into the middle of the CAR, with an offset of 2 from the cell's start address. Therefore, the bit pattern of a number will be:


      xxxxxxxxxxxxxxxxxxxxxxxxxxxxx010

Thus, a number is recognized by the interpreter when bit(1) is non-zero.


Symbols

A symbol is more complex than a number. Each symbol has a value, and optionally a name and an arbitrary number of properties. The CDR of a symbol cell is also called VAL, and the CAR points to the symbol's tail. As a minimum, a symbol consists of a single cell, and has no name or properties:


            Symbol
            |
            V
      +-----+-----+
      |  /  | VAL |
      +-----+-----+

That is, the symbol's tail is empty (points to NIL, as indicated by the '/' character).

The pointer to a symbol points to the CDR of the cell, with an offset of 4 from the cell's start address. Therefore, the bit pattern of a symbol will be:


      xxxxxxxxxxxxxxxxxxxxxxxxxxxxx100

Thus, a symbol is recognized by the interpreter when bit(2) is non-zero.

A property is a key-value-pair, represented as a cell in the symbol's tail. This is called a "property list". The property list may be terminated by a number representing the symbol's name. In the following example, a symbol with the name "abc" has three properties:


            Symbol
            |
            V
      +-----+-----+
      |  |  | VAL |
      +--+--+-----+
         | tail
         |
         V                                                      name
         +-----+-----+     +-----+-----+     +-----+-----+     +-----+-----+
         |  |  |  ---+---> | KEY |  ---+---> |  |  |  ---+---> |'cba'|  /  |
         +--+--+-----+     +-----+-----+     +--+--+-----+     +-----+-----+
            |                                   |
            V                                   V
            +-----+-----+                       +-----+-----+
            | VAL | KEY |                       | VAL | KEY |
            +-----+-----+                       +-----+-----+

Each property in a symbol's tail is either a symbol (then it represents a boolean value), or a cell with the property key in its CDR and the property value in its CAR. In both cases, the key should be a symbol, because searches in the property list are performed using pointer comparisons.

The name of a symbol is stored as a number at the end of the tail. It contains the characters of the name in UTF-8 encoding, using between one and three 8-bit-bytes per character. The first byte of the first character is stored in the lowest 8 bits of the number.

All symbols have the above structure, but depending on scope and accessibility there are actually four types of symbols: NIL, internal, transient and external symbols.


NIL

NIL is a special symbol which exists exactly once in the whole system. It is used

For that, NIL has a special structure:


      NIL:  /
            |
            V
      +-----+-----+-----+-----+
      |  /  |  /  |  /  |  /  |
      +-----+--+--+-----+-----+

The reason for that structure is NIL's dual nature both as a symbol and as a list:

These requirements are fulfilled by the above structure.


Internal Symbols

Internal Symbols are all those "normal" symbols, as they are used for function definitions and variable names. They are "interned" into an index structure, so that it is possible to find an internal symbol by searching for its name.

There cannot be two different internal symbols with the same name.

Initially, a new internal symbol's VAL is NIL.


Transient Symbols

Transient symbols are only interned into a index structure for a certain time (e.g. while reading the current source file), and are released after that. That means, a transient symbol cannot be accessed then by its name, and there may be several transient symbols in the system having the same name.

Transient symbols are used

Initially, a new transient symbol's VAL is that symbol itself.

A transient symbol without a name can be created with the box or new functions.


External Symbols

External symbols reside in a database file (or a similar resources (see *Ext)), and are loaded into memory - and written back to the file - dynamically as needed, and transparent to the programmer.

The interpreter recognizes external symbols internally by an additional tag bit in the tail structure.

There cannot be two different external symbols with the same name. External symbols are maintained in index structures while they are loaded into memory, and have their external location (disk file and block offset) directly coded into their names.

Initially, a new external symbol's VAL is NIL, unless otherwise specified at creation time.


Lists

A list is a sequence of one or more cells, holding numbers, symbols, or lists. Lists are used in Pico Lisp to emulate composite data structures like arrays, trees, stacks or queues.

In contrast to lists, numbers and symbols are collectively called "Atoms".

Typically, the CDR of each cell in a list points to the following cell, except for the last cell which points NIL. If, however, the CDR of the last cell points to an atom, that cell is called a "dotted pair" (because of its I/O syntax with a dot '.' between the two values).


Memory Management

The Pico Lisp interpreter has complete knowledge of all data in the system, due to the type information associated with every pointer. Therefore, an efficient garbage collector mechanism can easily be implemented. Pico Lisp employs a simple but fast mark-and-sweep garbage collector.

As the collection process is very fast (in the order of milliseconds per megabyte), it was not necessary to develop more complicated, time-consuming and error-prone garbage collection algorithms (e.g. incremental collection). A compacting garbage collector is also not necessary, because the single cell data type cannot cause heap fragmentation.


Programming Environment

Lisp was chosen as the programming language, because of its clear and simple structure.

In some previous versions, a Forth-like syntax was also implemented on top of a similar virtual machine (Lifo). Though that language was more flexible and expressive, the traditional Lisp syntax proved easier to handle, and the virtual machine can be kept considerably simpler. Pico Lisp inherits the major advantages of classical Lisp systems like

In the following, some concepts and peculiarities of the Pico Lisp language and environment are described.


Invocation

When Pico Lisp is invoked from the command line, an arbitrary number of arguments may follow the command name.

By default, each argument is the name of a file to be executed by the interpreter. If, however, the argument's first character is a hyphen '-', then the rest of that argument is taken as a Lisp function call (without the surrounding parentheses). A hyphen by itself as an argument stops evaluation of the rest of the command line (it may be processed later using the argv and opt functions). This mechanism corresponds to calling (load T).

As a convention, Pico Lisp source files have the extension ".l".

Note that the Pico Lisp executable itself does not expect or accept any command line flags or options. They are reserved for application programs.

The simplest and shortest invocation of Pico Lisp does nothing, and exits immediately by calling bye:


$ bin/picolisp -bye
$

In interactive mode, the Pico Lisp interpreter (see load) will also exit when an empty line is entered:


$ bin/picolisp
:                       # Typed ENTER
$

To start up the standard Pico Lisp environment, several files should be loaded. The most commonly used things are in "lib.l" and in a bunch of other files, which are in turn loaded by "ext.l". Thus, a typical call would be:


$ bin/picolisp lib.l ext.l

The recommended way, however, is to call the "p" shell script, which includes "lib.l" and "ext.l". Given that your current project is loaded by some file "myProject.l" and your startup function is main, your invocation would look like:


$ ./p myProject.l -main

For interactive development and debugging it is recommended also to load "dbg.l" (or use './dbg' instead of './p'), to get the vi-style command line editor, single-stepping, tracing and other debugging utilities.


$ ./dbg myProject.l -main

In any case, the directory part of the first file name supplied on the command line (normally, the path to "lib.l") is remembered internally as the Pico Lisp Home Directory. This path is later automatically substituted for any leading "@" character in file name arguments to I/O functions (see path).


Input/Output

In Lisp, each internal data structure has a well-defined external representation in human-readable format. All kinds of data can be written to a file, and restored later to their original form by reading that file.

In normal operation, the Pico Lisp interpreter continuously executes an infinite "read-eval-print loop". It reads one expression at a time, evaluates it, and prints the result to the console. Any input into the system, like data structures and function definitions, is done in a consistent way no matter whether it is entered at the console or read from a file.

Comments can be embedded in the input stream with the hash # character. Everything up to the end of that line will be ignored by the reader.


: (* 1 2 3)  # This is a comment
-> 6

A comment spanning several lines may be enclosed between #{ and }#.

Here is the I/O syntax for the individual Pico Lisp data types:


Numbers

A number consists of an arbitrary number of digits ('0' through '9'), optionally preceded by a sign character ('+' or '-'). Legal number input is:


: 7
-> 7
: -12345678901245678901234567890
-> -12345678901245678901234567890

Fixed-point numbers can be input by embedding a decimal point '.', and setting the global variable *Scl appropriately:


: *Scl
-> 0

: 123.45
-> 123
: 456.78
-> 457

: (setq *Scl 3)
-> 3
: 123.45
-> 123450
: 456.78
-> 456780

Thus, fixed-point input simply scales the number to an integer value corresponding to the number of digits in *Scl.

Formatted output of scaled fixed-point values can be done with the format function:


: (format 1234567890 2)
-> "12345678.90"
: (format 1234567890 2 "." ",")
-> "12,345,678.90"


Symbols

The reader is able to recognize the individual symbol types from their syntactic form. A symbol name should - of course - not look like a legal number (see above).

In general, symbol names are case-sensitive. car is not the same as CAR.


NIL

Besides for standard normal form, NIL is also recognized as (), [] or "".


: NIL
-> NIL
: ()
-> NIL
: ""
-> NIL

Output will always appear as NIL.


Internal Symbols

Internal symbol names can consist of any printable (non-whitespace) character, except for the following meta characters:


   "  '  (  )  ,  [  ]  `  ~ { }

It is possible, though, to include these special characters into symbol names by escaping them with a backslash '\'.

The dot '.' has a dual nature. It is a meta character when standing alone, denoting a dotted pair, but can otherwise be used in symbol names.

As a rule, anything not recognized by the reader as another data type will be returned as an internal symbol.


Transient Symbols

A transient symbol is anything surrounded by double quotes '"'. With that, it looks - and can be used - like a string constant in other languages. However, it is a real symbol, and may be assigned a value or a function definition, and properties.

Initially, a transient symbol's value is that symbol itself, so that it does not need to be quoted for evaluation:


: "This is a string"
-> "This is a string"

However, care must be taken when assigning a value to a transient symbol. This may cause unexpected behavior:


: (setq "This is a string" 12345)
-> 12345
: "This is a string"
-> 12345

The name of a transient symbol can contain any character except zero. A double quote character can be escaped with a backslash '\', and a backslash itself has to be escaped with another backslash. Control characters can be written with a preceding hat '^' character.


: "We^Ird\\Str\"ing"
-> "We^Ird\\Str\"ing"
: (chop @)
-> ("W" "e" "^I" "r" "d" "\\" "S" "t" "r" "\"" "i" "n" "g")

The index for transient symbols is cleared automatically before and after loading a source file, or it can be reset explicitly with the ==== function. With that mechanism, it is possible to create symbols with a local access scope, not accessible from other parts of the program.

A special case of transient symbols are anonymous symbols. These are symbols without name (see box, box? or new). They print as a dollar sign ($) followed by a decimal digit string (actually their machine address).


External Symbols

External symbol names are surrounded by braces ('{' and '}'). The characters of the symbol's name itself identify the physical location of the external object. This is currently the number of the database file, and - separated by a hyphen - the starting block in the database file. Both numbers are encoded in base-64 notation (characters '0' through '9', ':' through ';', 'A' through 'Z' and 'a' through 'z').


Lists

Lists are surrounded by parentheses ('(' and ')').

(A) is a list consisting of a single cell, with the symbol A in its CAR, and NIL in its CDR.

(A B C) is a list consisting of three cells, with the symbols A, B and C respectively in their CAR, and NIL in the last cell's CDR.

(A . B) is a "dotted pair", a list consisting of a single cell, with the symbol A in its CAR, and B in its CDR.

Pico Lisp has built-in support for reading and printing simple circular lists. If the dot in a dotted-pair notation is immediately followed by a closing parenthesis, it indicates that the CDR of the last cell points back to the beginning of that list.


: (let L '(a b c) (conc L L))
-> (a b c .)
: (cdr '(a b c .))
-> (b c a .)
: (cddddr '(a b c .))
-> (b c a .)

A similar result can be achieved with the function circ. Such lists must be used with care, because many functions won't terminate or will crash when given such a list.


Read-Macros

Read-macros in Pico Lisp are special forms that are recognized by the reader, and modify its behavior. Note that they take effect immediately while reading an expression, and are not seen by the eval in the main loop.

The most prominent read-macro in Lisp is the single quote character ', which expands to a call of the quote function. Note that the single quote character is also printed instead of the full function name.


: '(a b c)
-> (a b c)
: '(quote . a)
-> 'a
: (cons 'quote 'a)   # (quote . a)
-> 'a
: (list 'quote 'a)   # (quote a)
-> '(a)

A comma (,) will cause the reader to collect the following data item into an idx tree in the global variable *Uni, and to return a previously inserted equal item if present. This makes it possible to create a unique list of references to data which do normally not follow the rules of pointer equality.

A single backquote character ` will cause the reader to evaluate the following expression, and return the result.


: '(a `(+ 1 2 3) z)
-> (a 6 z)

A tilde character ~ inside a list will cause the reader to evaluate the following expression, and splice the result into the list.


: '(a b c ~(list 'd 'e 'f) g h i)
-> (a b c d e f g h i)

Brackets ('[' and ']') can be used as super parentheses. A closing bracket will match the innermost opening bracket, or all currently open parentheses.


: '(a (b (c (d]
-> (a (b (c (d))))
: '(a (b [c (d]))
-> (a (b (c (d))))

Finally, reading the sequence '{}' will result in a new anonymous symbol with value NIL, equivalent to a call to box without arguments.


: '({} {} {})
-> ($134599965 $134599967 $134599969)
: (mapcar val @)
-> (NIL NIL NIL)


Evaluation

Pico Lisp tries to evaluate any expression encountered in the read-eval-print loop. Basically, it does so by applying the following three rules:


: 1234
-> 1234        # Number evaluates to itself
: *Pid
-> 22972       # Symbol evaluates to its VAL
: (+ 1 2 3)
-> 6           # List is evaluated as a function call

For the third rule, however, things get a bit more involved. First - as a special case - if the CAR of the list is a number, the whole list is returned as it is:


: (1 2 3 4 5 6)
-> (1 2 3 4 5 6)

This is not really a function call but just a convenience to avoid having to quote simple data lists.

Otherwise, if the CAR is a symbol or a list, Pico Lisp tries to obtain an executable function from that, by either using the symbol's value, or by evaluating the list.

What is an executable function? Or, said in another way, what can be applied to a list of arguments, to result in a function call? A legal function in Pico Lisp is

either
a number. When a number is used as a function, it is simply taken as a pointer to executable code that will be called with the list of (unevaluated) arguments as its single parameter. It is up to that code to evaluate the arguments, or not. Some functions do not evaluate their arguments (e.g. quote) or evaluate only some of their arguments (e.g. setq).
or
a lambda expression. A lambda expression is a list, whose CAR is either a symbol or a list of symbols, and whose CDR is a list of expressions. Note: In contrast to other Lisp implementations, the symbol LAMBDA itself does not exist in Pico Lisp but is implied from context.

A few examples should help to understand the practical consequences of these rules. In the most common case, the CAR will be a symbol defined as a function, like the * in:


: (* 1 2 3)    # Call the function '*'
-> 6

Inspecting the VAL of *, however, gives


: *            # Get the VAL of the symbol '*'
-> 67291944

The VAL of * is a number. In fact, it is the numeric representation of a C-function pointer, i.e. a pointer to executable code. This is the case for all built-in functions of Pico Lisp.

Other functions in turn are written as Lisp expressions:


: (de foo (X Y)            # Define the function 'foo'
   (* (+ X Y) (+ X Y)) )
-> foo
: (foo 2 3)                # Call the function 'foo'
-> 25
: foo                      # Get the VAL of the symbol 'foo'
-> ((X Y) (* (+ X Y) (+ X Y)))

The VAL of foo is a list. It is the list that was assigned to foo with the de function. It would be perfectly legal to use setq instead of de:


: (setq foo '((X Y) (* (+ X Y) (+ X Y))))
-> ((X Y) (* (+ X Y) (+ X Y)))
: (foo 2 3)
-> 25

If the VAL of foo were another symbol, that symbol's VAL would be used instead to search for an executable function.

As we said above, if the CAR of the evaluated expression is not a symbol but a list, that list is evaluated to obtain an executable function.


: ((intern (pack "c" "a" "r")) (1 2 3))
-> 1

Here, the intern function returns the symbol car whose VAL is used then. It is also legal, though quite dangerous, to use the code-pointer directly:


: car
-> 67306152
: ((* 2 33653076) (1 2 3))
-> 1

When an executable function is defined in Lisp itself, we call it a lambda expression. A lambda expression always has a list of executable expressions as its CDR. The CAR, however, must be a either a list of symbols, or a single symbol, and it controls the evaluation of the arguments to the executable function according to the following rules:

When the CAR is a list of symbols
For each of these symbols an argument is evaluated, then the symbols are bound simultaneously to the results. The body of the lambda expression is executed, then the VAL's of the symbols are restored to their original values. This is the most common case, a fixed number of arguments is passed to the function.
Otherwise, when the CAR is the symbol @
All arguments are evaluated and the results kept internally in a list. The body of the lambda expression is executed, and the evaluated arguments can be accessed sequentially with the args, next, arg and rest functions. This allows to define functions with a variable number of evaluated arguments.
Otherwise, when the CAR is a single symbol
The symbol is bound to the whole unevaluated argument list. The body of the lambda expression is executed, then the symbol is restored to its original value. This allows to define functions with unevaluated arguments. Any kind of interpretation and evaluation of the argument list can be done inside the expression body.

In all cases, the return value is the result of the last expression in the body.


: (de foo (X Y Z)                   # CAR is a list of symbols
   (list X Y Z) )                   # Return a list of all arguments
-> foo
: (foo (+ 1 2) (+ 3 4) (+ 5 6))
-> (3 7 11)                         # all arguments are evaluated


: (de foo X                         # CAR is a single symbol
   X )                              # Return the argument
-> foo
: (foo (+ 1 2) (+ 3 4) (+ 5 6))
-> ((+ 1 2) (+ 3 4) (+ 5 6))        # the whole unevaluated list is returned


: (de foo @                         # CAR is the symbol '@'
   (list (next) (next) (next)) )    # Return the first three arguments
-> foo
: (foo (+ 1 2) (+ 3 4) (+ 5 6))
-> (3 7 11)                         # all arguments are evaluated

Note that these forms can also be combined. For example, to evaluate only the first two arguments, bind the results to X and Y, and bind all other arguments (unevaluated) to Z:


: (de foo (X Y . Z)                 # CAR is a list with a dotted-pair tail
   (list X Y Z) )                   # Return a list of all arguments
-> foo
: (foo (+ 1 2) (+ 3 4) (+ 5 6))
-> (3 7 ((+ 5 6)))                  # two arguments are evaluated

Or, a single argument followed by a variable number of arguments:


: (de foo (X . @)                   # CAR is a dotted-pair with '@'
   (println X)                      # print the first evaluated argument
   (while (args)                    # while there are more arguments
      (println (next)) ) )          # print the next one
-> foo
: (foo (+ 1 2) (+ 3 4) (+ 5 6))
3                                   # X
7                                   # Next arg
11
-> 11

In general, if more than the expected number of arguments is supplied to a function, these extra arguments will be ignored. Missing arguments default to NIL.


Interrupt

During the evaluation of an expression, the Pico Lisp interpreter can be interrupted at any time by hitting Ctrl-C. It will then enter the breakpoint routine, as if ! were called.

Hitting ENTER at that point will continue evaluation, while (quit) will abort evaluation and return the interpreter to the top level. See also debug, e, ^ and *Dbg


Error Handling

When a runtime error occurs, execution is stopped and an error handler is entered.

The error handler resets the I/O channels to the console, and displays the location (if possible) and the reason of the error, followed by an error message. That message is also stored in the global *Msg, and the location of the error in ^. If the VAL of the global *Err is non-NIL it is executed as a prg body. If the standard input is from a terminal, a read-eval-print loop (with a question mark "?" as prompt) is entered (the loop is exited when an empty line is input). Then all pending finally expressions are executed, all variable bindings restored, and all files closed. If the standard input is not from a terminal, the interpreter terminates. Otherwise it is reset to its top-level state.


: (de foo (A B) (badFoo A B))       # 'foo' calls an undefined symbol
-> foo
: (foo 3 4)                         # Call 'foo'
!? (badFoo A B)                     # Error handler entered
badFoo -- Undefined
? A                                 # Inspect 'A'
-> 3
? B                                 # Inspect 'B'
-> 4
?                                   # Empty line: Exit
:

Errors can be caught with catch, if a list of substrings of possible error messages is supplied for the first argument. In such a case, the matching substring is returned.


@ Result

In certain situations, the result of the last evaluation is stored in the VAL of the symbol @. This can be very convenient, because it often makes the assignment to temporary variables unnecessary.

load
In read-eval loops, the last three results which were printed at the console are available in @@@, @@ and @, in that order (i.e the latest result is in @).


: (+ 1 2 3)
-> 6
: (/ 128 4)
-> 32
: (- @ @@)        # Subtract the last two results
-> 26

Flow functions
Flow- and logic-functions store the result of their controlling expression - respectively non-NIL results of their conditional expression - in @.


: (while (read) (println 'got: @))
abc            # User input
got: abc       # print result
123            # User input
got: 123       # print result
NIL
-> 123

: (setq L (1 2 3 4 5 1 2 3 4 5))
-> (1 2 3 4 5 1 2 3 4 5)
: (and (member 3 L) (member 3 (cdr @)) (set @ 999))
-> 999
: L
-> (1 2 3 4 5 1 2 999 4 5)

Functions with controlling expressions are case, prog1, prog2, and the bodies of *Run tasks.

Functions with conditional expressions are and, cond, do, for, if, if2, ifn, loop, nand, nond, nor, or, state, unless, until, when and while.

@ is generally local to functions and methods, its value is automatically saved upon function entry and restored at exit.


Comparing

In Pico Lisp, it is legal to compare data items of arbitrary type. Any two items are either

Identical
They are the same memory object (pointer equality). For example, two internal symbols with the same name are identical.
Equal
They are equal in every respect (structure equality), but need not to be identical. Examples are numbers with the same value, transient symbols with the same name or lists with equal elements.
Or they have a well-defined ordinal relationship
Numbers are comparable by their numeric value, strings by their name, and lists recursively by their elements (if the CAR's are equal, their CDR's are compared). For differing types, the following rule applies: Numbers are less than symbols, and symbols are less than lists. As special cases, NIL is always less than anything else, and T is always greater than anything else.

To demonstrate this, sort a list of mixed data types:


: (sort '("abc" T (d e f) NIL 123 DEF))
-> (NIL 123 DEF "abc" (d e f) T)

See also max, min, rank, <, =, > etc.


OO Concepts

Pico Lisp comes with built-in object oriented extensions. There seems to be a common agreement upon three criteria for object orientation:

Encapsulation
Code and data are encapsulated into objects, giving them both a behavior and a state. Objects communicate by sending and receiving messages.
Inheritance
Objects are organized into classes. The behavior of an object is inherited from its class(es) and superclass(es).
Polymorphism
Objects of different classes may behave differently in response to the same message. For that, classes may define different methods for each message.

Pico Lisp implements both objects and classes with symbols. Object-local data are stored in the symbol's property list, while the code (methods) and links to the superclasses are stored in the symbol's VAL (encapsulation).

In fact, there is no formal difference between objects and classes (except that objects usually are anonymous symbols containing mostly local data, while classes are named internal symbols with an emphasis on method definitions). At any time, a class may be assigned its own local data (class variables), and any object can receive individual method definitions in addition to (or overriding) those inherited from its (super)classes.

Pico Lisp supports multiple inheritance. The VAL of each object is a (possibly empty) association list of message symbols and method bodies, concatenated with a list of classes. When a message is sent to an object, it is searched in the object's own method list, and then (with a left-to-right depth-first search) in the tree of its classes and superclasses. The first method found is executed and the search stops. The search may be explicitly continued with the extra and super functions.

Thus, which method is actually executed when a message is sent to an object depends on the classes that the object is currently linked to (polymorphism). As the method search is fully dynamic (late binding), an object's type (i.e. its classes and method definitions) can be changed even at runtime!

While a method body is being executed, the global variable This is set to the current object, allowing the use of the short-cut property functions =:, : and ::.


Database

On the lowest level, a Pico Lisp database is simply a collection of external symbols. They reside in a database file, and are dynamically swapped in and out of memory. Only one database can be open at a time (pool).

In addition, further external symbols can be specified to originate from arbitrary sources via the *Ext mechanism.

The symbols in the database can be used to store arbitrary information structures. In typical use, some symbols represent nodes of search trees, by holding keys, values, and links to subtrees in their VAL's. Such a search tree in the database is called index.

For the most part, other symbols in the database are objects derived from the +Entity class.

Entities depend on objects of the +Relation class hierarchy. Relation-objects manage the property values of entities, they define the application database model and are responsible for the integrity of mutual object references and index trees.

Relations are stored as properties in the entity classes, their methods are invoked as daemons whenever property values in an entity are changed. When defining an +Entity class, relations are defined - in addition to the method definitions of a normal class - with the rel function. Predefined relation classes include


Pilog (Pico Lisp Prolog)

A declarative language is built on top of Pico Lisp, that has the semantics of Prolog, but uses the syntax of Lisp.

For an explanation of Prolog's declarative programming style, an introduction like "Programming in Prolog" by Clocksin/Mellish (Springer-Verlag 1981) is recommended.

Facts and rules can be declared with the be function. For example, a Prolog fact 'likes(john,mary).' is written in Pilog as:


(be likes (John Mary))

and a rule 'likes(john,X) :- likes(X,wine), likes(X,food).' is in Pilog:


(be likes (John @X) (likes @X wine) (likes @X food))

As in Prolog, the difference between facts and rules is that the latter ones have conditions, and usually contain variables.

A variable in Pilog is any symbol starting with an at-mark character ("@"). The symbol @ itself can be used as an anonymous variable: It will match during unification, but will not be bound to the matched values.

The cut operator of Prolog (usually written as an exclamation mark (!)) is the symbol T in Pilog.

Pilog can be called from Lisp and vice versa:

An interactive query can be done with the ? function:


(? (likes John @X))

This will print all solutions, waiting for user input after each line. If a non-empty line (not just a ENTER key, but for example a dot (.) followed by ENTER) is typed, it will terminate.


Naming Conventions

It was necessary to introduce - and adhere to - a set of conventions for Pico Lisp symbol names. Because all (internal) symbols have a global scope (there are no packages or name spaces), and each symbol can only have either a value or function definition, it would otherwise be very easy to introduce name conflicts. Besides this, source code readability is increased when the scope of a symbol is indicated by its name.

These conventions are not hard-coded into the language, but should be so into the head of the programmer. Here are the most commonly used ones:

For historical reasons, the global constant symbols T and NIL do not obey these rules, and are written in upper case.

For example, a local variable could easily overshadow a function definition:


: (de max-speed (car)
   (.. (get car 'speeds) ..) )
-> max-speed

Inside the body of max-speed (and all other functions called during that execution) the kernel function car is redefined to some other value, and will surely crash if something like (car Lst) is executed. Instead, it is safe to write:


: (de max-speed (Car)            # 'Car' with upper case first letter
   (.. (get Car 'speeds) ..) )
-> max-speed

Note that there are also some strict naming rules (as opposed to the voluntary conventions) that are required by the corresponding kernel functionalities, like:

With that, the last of the above conventions (local functions start with an underscore) is not really necessary, because true local scope can be enforced with transient symbols.


Breaking Traditions

Pico Lisp does not try very hard to be compatible with traditional Lisp systems. If you are used to some other Lisp dialects, you may notice the following differences:

Case Sensitivity
Pico Lisp distinguishes between upper case and lower case characters in symbol names. Thus, CAR and car are different symbols, which was not the case in traditional Lisp systems.
QUOTE
In traditional Lisp, the QUOTE function returns its first unevaluated argument. In Pico Lisp, on the other hand, quote returns all (unevaluated) argument(s).
LAMBDA
The LAMBDA function, in some way at the heart of traditional Lisp, is completely missing (and quote is used instead).
PROG
The PROG function of traditional Lisp, with its GOTO and ENTER functionality, is also missing. Pico Lisp's prog function is just a simple sequencer (as PROGN in some Lisps).
Function/Value
In Pico Lisp, a symbol cannot have a value and a function definition at the same time. Though this is a disadvantage at first sight, it allows a completely uniform handling of functional data.


Bugs

The names of the symbols T and NIL violate the naming conventions. They are global symbols, and should therefore start with an asterisk "*". It is too easy to bind them to some other value by mistake:


(de foo (R S T)
   ...

However, lint will issue a warning in such a case.


Function Reference

This section provides a reference manual for the kernel functions, and some extensions. See the thematically grouped list of indexes below.

Though Pico Lisp is a dynamically typed language (resolved at runtime, as opposed to statically (compile-time) typed languages), many functions can only accept and/or return a certain set of data types. For each function, the expected argument types and return values are described with the following abbreviations:

The primary data types:

Other (derived) data types

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z Other

Symbol Functions
new sym str char name sp? pat? fun? all intern extern ==== qsym loc box? str? ext? touch zap length size format chop pack glue pad align center text wrap pre? sub? low? upp? lowc uppc fold val getd set setq def de dm recur undef redef daemon patch xchg on off onOff zero one default expr subr let let? use accu push push1 pop cut del queue fifo idx lup cache locale dirname
Property Access
put get prop ; =: : :: putl getl wipe meta
Predicates
atom pair lst? num? sym? flg? sp? pat? fun? box? str? ext? bool not == n== = <> =0 =T n0 nT < <= > >= match
Arithmetics
+ - * / % */ ** inc dec >> lt0 ge0 gt0 abs bit? & | x| sqrt seed rand max min length size accu format pad oct hex fmt64 money
List Processing
car cdr caar cadr cdar cddr caaar caadr cadar caddr cdaar cdadr cddar cdddr cadddr cddddr nth con cons conc circ rot list need full make made chain link yoke copy mix append delete delq replace insert remove place strip split reverse flip trim clip head tail stem fin last member memq mmeq sect diff index offset assoc asoq rank sort uniq group length size val set xchg push push1 pop cut queue fifo idx balance get fill apply
Control Flow
load args next arg rest pass quote as pid lit eval run macro curry def de dm recur recurse undef box new type isa method meth send try super extra with bind job let let? use and or nand nor xor bool not nil t prog prog1 prog2 if if2 ifn when unless cond nond case state while until loop do at for catch throw finally ! e $ sys call tick ipid opid kill quit task fork pipe later timeout abort bye
Mapping
apply pass maps map mapc maplist mapcar mapcon mapcan filter extract seek find pick cnt sum maxi mini fish by
Input/Output
path in ipid out opid pipe ctl any sym str load hear tell key poll peek char skip eol eof from till line format scl read print println printsp prin prinl msg space beep tab flush rewind rd pr wr rpc wait sync echo info file dir lines open close port listen accept host connect nagle udp script once rc pretty pp show view here prEval mail
Object Orientation
*Class class dm rel var var: new type isa method meth send try object extend super extra with This
Database
pool journal id seq lieu lock commit rollback mark free dbck rel dbs dbs+ db: fmt64 tree root fetch store count leaf minKey maxKey genKey useKey init step scan iter prune zapTree chkTree db aux collect
Pilog
be goal prove -> unify ?
Debugging
pretty pp show loc debug vi ld trace lint lintAll fmt64
System Functions
cmd argv opt gc raw alarm protect heap env up date time usec stamp dat$ $dat datSym datStr strDat expDat day week ultimo tim$ $tim telStr expTel locale allowed allow pwd cd chdir ctty info dir dirname call tick kill quit task fork pipe timeout mail test bye
Globals
NIL *OS *DB T *Solo *PPid *Pid @ @@ @@@ This *Dbg *Zap *Scl *Class *Dbs *Run *Hup *Sig1 *Sig2 ^ *Err *Msg *Uni *Led *Adr *Allow *Fork *Bye


Download

The Pico Lisp system can be downloaded from the Pico Lisp Download page.