Skip to content

Latest commit

 

History

History
669 lines (519 loc) · 19.1 KB

intro.md

File metadata and controls

669 lines (519 loc) · 19.1 KB

The l1 Language

Jump to API / list of operators

The REPL

Examples in these docs are shown as if typed at the l1 "REPL" (read-eval-print-loop). To leave the REPL, type Control-D or (exit):

$ l1
> (exit)
$

If you want arrow keys for next/previous command or to move forward or backward on the line, history, etc., wrap l1 with the rlwrap command (installed using your favorite package manager), e.g.:

$ rlwrap l1
>

Expressions

Expressions in l1 are atoms, lists, numbers, or functions:

Atoms

Atoms have a name, such as x, foo, or Eisenhower!, and can be "bound to a value in the current environment," meaning, essentially, assigned a value. For example, we can bind a value to a and retrieve it:

> (def a 1)
> a
1
>

Lists

Lists are collections of zero or more expressions. Examples:

(getting atomic)
(10 9 8 7 6 5 4 3 2 1)
(+ 2 2)
()

In general, lists represent operations whose name is the first element and whose arguments are the remaining elements. For example, (+ 2 2) is a list that evaluates to the number 4. To prevent evaluation of a list, prepend a quote character:

> '(+ 2 2)
(+ 2 2)
> (+ 2 2)
4

As in most Lisps, lists are actually implemented "under the hood" as "cons pairs" (Wikipedia page). The list

(1 2 3)

actually represented internally as

(1 . (2 . (3 . ())))

where

(a . b)

is the same as

(cons a b)

In practice, the dot notation is uncommon in l1 programs, except when used to represent rest arguments, described below.

Numbers

Numbers are integer values and can be of arbitrary magnitude:

0
999
7891349058731409803589073418970341089734958701432789

Numbers evaluate to themselves:

> 0
0
> 7891349058731409803589073418970341089734958701432789
7891349058731409803589073418970341089734958701432789

Numbers and atoms can be turned into lists:

> (split 'atomic)
(a t o m i c)
> (split 1234)
(1 2 3 4)

And lists can be turned into atoms or numbers:

> (fuse '(getting atomic))
gettingatomic
> (fuse '(10 9 8 7 6 5 4 3 2 1))
10987654321

Boolean Logic

In l1, the empty list () is the only logical false value; everything else is logical true. Logical true and false are important when evaluating conditional statements such as if, when, or cond. The default logical true value is t. t and () evaluate to themselves.

The and, or and not operators work like they do in most other languages:

> (and t t)
t
> (or () ())
()
> (or () 135987)
135987
> (and () (launch missiles))
()
> (not t)
()
> (not ())
t
> (not 13987)
()
> (if ()
    (launch missiles)
    555)
555

Special Characters

Unlike many modern languages, l1 doesn't have strings. Instead, atoms and lists are used where strings normally would be:

> (printl '(Hello, world!))
Hello, world!
()

(In this example, the Hello, world! is output to the terminal, and then the return value of printl, namely ().)

Atom names can be arbitrarily long (they are simply Go strings under the hood). When l1 parses your code, it will interpret any UTF-8-encoded unicode characters but the following as the start of an atom:

0123456789+-. \t\n\r()~@#;`'

After the first character, anything is allowed except spaces or

\t\n\r()~@#;`'

Deviations from these constraints need special handling. For example:

> (printl '(@Hello, world!))
(...
 (unexpected lexeme 'unexpected character '@' in input' on line 1))

A workaround is to use syntax quote and unquote, described below, to dynamically create a new atom name using the BANG alias for !:

> (printl `(~(fuse `(~ATSIGN Hello,)) world!))
@Hello, world!
()

This is admittedly awkward, but rare in practice for the kinds of programs l1 was designed for. BANG is one of a small set of atoms helpful for this sort of construction:

BANG
COLON
COMMA
NEWLINE
PERIOD
QMARK
SPACE
TAB
QUOTE
DQUOTE
BQUOTE
UPLINE

These all evaluate to atoms whose names are the unreadable characters, some of which may be helpful for text games and other diversions:

> (dotimes 10
    (println
     (fuse
      (repeatedly 10
                  (lambda ()
                    (randchoice (list COMMA
                                      COLON
                                      PERIOD
                                      BANG
                                      QMARK)))))))

.!!!.???..
,??::?!,.?
?,?!?..:!!
,:.,?.:!!!
!!:?!::.,?
,:!!!:,!!:
,???:?!:!?
.,!!?,!:!?
!:,!!!.:!:
??.,,:.:..
()
>

The UPLINE character, in particular, which moves the cursor to the beginning of the previous line, is useful for making command-line output which repeatedly updates the line in place, rather than printing multiple lines.

Functions

Functions come in two flavors: temporary functions, called "lambda" functions for historical reasons, and functions which are defined with a name and kept around in the environment for later use. For example,

> (defn plus2 (x) (+ x 2))
> (plus2 3)
5
> ((lambda (x) (* 5 x)) 3)
15

Since function names are atoms, their names follow the same rules for atoms given above. The following function definitions are all equally valid:

> (defn increase! (x) (* x 1000))
()
> (defn 增加! (x) (* x 1000))
()
> (defn մեծացնել! (x) (* x 1000))
()

Functions can take a fixed number of arguments plus an extra "rest" argument, separated from the fixed arguments with a "."; the rest argument is then bound to a list of all remaining arguments:

> (defn multiply-then-sum (multiplier . xs)
    (* multiplier (apply + xs)))
()
> (multiply-then-sum 5 1)
5
> (multiply-then-sum 5 1 2 3)
30

A function that has a rest argument but no fixed arguments is specified using the empty list as its fixed argument:

> (defn say-hello (() . friends)
    (list* 'hello friends))
> (say-hello 'John 'Jerry 'Eden)
(hello John Jerry Eden)

In addition to the functions described above, some l1 functions are "built in" (implemented in Go as part of the language core). Examples include car, cdr, cons, etc. The API Docs below specify whether a function is built-in or not.

One special family of predefined functions not shown in the API docs (because they are effectively infinite in number) is extensions of car and cdr:

> (car '(1 2 3))
1
> (cadr '(1 2 3))
2
> (caddr '(1 2 3))
3
> (caar '((one fish) (two fish)))
one
> (caadr '((one fish) (two fish)))
two
> (cadar '((one fish) (two fish)))
fish

(cadr x) should be read as (car (cdr x)), and so on.

Functions may invoke themselves recursively:

> (defn sum-nums (l)
    (if-not l
      0
      (+ (car l) (sum-nums (cdr l)))))
()
> (sum-nums '(0 1 2 3 4 5 6 7 8 9))
45

The above function performs an addition after it invokes itself. A function which invokes itself immediately before returning, without doing any more work, is called "tail recursive." Such functions are turned into iterations automatically by the interpreter ("tail call optimization"). The above function can be rewritten into a tail-recursive version:

> (defn sum-nums-accumulator (l acc)
    (if-not l
      acc
      (sum-nums-accumulator (cdr l) (+ acc (car l)))))
()
> (sum-nums-accumulator '(0 1 2 3 4 5 6 7 8 9) 0)
45

Lambda functions can invoke themselves if given a name directly before the parameters are declared. We can rewrite the above function to hide the acc argument from the user:

> (defn sum-nums (l)
    (let ((inner (lambda inner (l acc)
                   (if-not l
                     acc
                     (inner (cdr l) (+ acc (car l)))))))
      (inner l 0)))
()
> (sum-nums '(0 1 2 3 4 5 6 7 8 9))
45

In this version, inner is tail-recursive, and sum-nums is now as convenient to use as our first, non-tail-recursive version was.

Flow of Control

In addition to the basic conditional statements cond, if, if-not, when, and when-not, flow of control is generally implemented via recursion, as it is in Scheme, and inspection of its core library l1.l1 will show several examples of recursive functions being used as the primary recurrence method. A few other control flow methods are also available: while, which loops so long as a condition is true; dotimes, which executes a body of statements a given number of times; foreach, which executes a body of statements for each element in a loop; and loop, which loops forever. Macros can be used to create new control abstractions; however, possibilities are somewhat restricted compared to some languages, due to the inability of l1 to handle branches or "goto" statements.

Assertions and Error Handling

is

There is currently one kind of assertion expression in l1, namely the is macro:

> (is (= 4 (+ 2 2)))
> (is ())
ERROR:
((assertion failed: ()))
>

If the argument to is is logical false (()), then an error is printed and the program exits (if running a program) or the REPL returns to the top level prompt.

If is is checking an equality of two items which fails, the macro is smart enough to print out a more detailed error report showing the two expressions and their values:

> (is (= 5 (+ 1 1)))
ERROR:
((expression 5 ==> 5 is not equal to expression (+ 1 1) ==> 2))
>

error

If desired, an error can be caused deliberately with the error function:

> (defn checking-len (x)
    (if (list? x)
     (len x)
     (error '(argument must be a list))))
()
> (checking-len 3)
ERROR:
((argument must be a list))
>

errors

In some situations, such as during automated tests, it may be desirable to ensure that a specific error is raised. The errors special form takes a list argument and a body, then checks to ensure that the body raises an error which contains the supplied list:

> (errors '(division by zero) (/ 1 0))
()
> (errors '(division) (/ 1 0))
()
> (errors '(rocket crashed) (/ 1 0))
ERROR:
((error 'rocket crashed' not found in '((builtin function /) (division by zero))'))
> (errors '(division by zero) (* 1 0))
ERROR:
((error not found in ((quote (division by zero)) (* 1 0))))
>

try ... catch

Most contemporary languages will print a stacktrace when an error occurs. l1 stacktraces are somewhat rudimentary: in keeping with the rest of the language, they are simply lists. To capture an error occurring within a body of code, wrap the body in a try statement and add a catch clause, as follows:

> (try
    (printl '(got here))
    (+ 3
       (/ 1 0))
    (printl '(did not get here))
  (catch e
    (cons '(oh boy, another error)
          e)))
got here
((oh boy, another error) (builtin function /) (division by zero))
>

The exception e is a list of lists to which items are added (in the front) as the error returns up the call chain. As an ordinary list, it can be manipulated like any other, as shown above using cons.

An important caveat is that, since tail recursion is optimized away, many "stack frames" (or their equivalent) are optimized away - there is no way to track the entire history in detail without losing the space-saving power of the optimization. Nevertheless, the generated exception can be helpful for troubleshooting.

There is, currently, no equivalent of the finally clause one sees in Java or Clojure.

swallow

Rarely, one may wish to swallow any errors and continue execution. swallow will execute all the statements in its body and return t if and only if any of them causes an (uncaught) error:

> (swallow (/ 1 0))
t
> (swallow (+ 1 1))
()
>

swallow is used mainly in the fuzzing tests for l1 (see the examples directory).

Subprocesses

The shell function executes a subprocess command, which should be a list of atoms and numbers, and returns the result in the following form:

((... stdout lines...)
 (... stderr lines...)
 ...exit code..)

Examples (output reformatted for clarity):

> (shell '(pwd))
(((/Users/jacobsen/Programming/go/l1))
 (())
 0)
> (shell '(ls))
(((Dockerfile) (LICENSE) (Makefile) (README.md) (api.md)
  (atom.go) (builtin.go) (builtin_test.go) (bumpver)
  (cons.go) (core.go) (doc.go) (env.go) (env_test.go)
  (error.go) (error_test.go) (eval_test.go) (example.l1)
  (examples) (examples.txt) (go.mod) (go.sum) (intro.md)
  (l1) (l1.html) (l1.jpg) (l1.l1) (l1.md) (lambda.go)
  (lex.go) (lex_test.go) (lisp.go) (main.go) (math.go)
  (math_test.go) (parse.go) (parse_test.go) (sexpr_test.go)
  (shell.go) (state.go) (term.go) (tests.l1) (updatereadme.py)
  (util.go) (version.go))
 (())
 0)
> (shell '(ls -al l1.l1))
(((-rw-r--r-- 1 jacobsen staff 16636 Sep 3 11:28 l1.l1)) (()) 0)
> (shell '(ls /watermelon))
((()) ((ls: /watermelon: No such file or directory)) 1)

Macros

For those familiar with macros (I recommend Paul Graham's On Lisp for those who are not), l1 macros are non-hygienic by default. Gensyms, unique atom names useful for writing safe macros, are available via the gensym built-in function:

> (gensym)
<gensym-0>
> (gensym 'foo)
<gensym-foo-1>

The traditional syntax-quote, unquote, and splicing-unquote are available, and have sugared equivalents:

(let ((name 'larry)
      (names '(moe curly)))
  (syntax-quote (hello, (unquote name) as well as (splicing-unquote names))))
;;=>
(hello, larry as well as moe curly)

is the same as

(let ((name 'larry)
      (names '(moe curly)))
  `(hello, ~name as well as ~@names))

In addition to the quote ('), syntax-quote (`), unquote (~), and splicing unquote (~@) shortcuts, the shortcut #_ is available, which is equivalent to (comment ...), e.g.,

#_(this is a commented form)

is equivalent to

(comment (this is a commented form))

Text User Interfaces

l1 has a few built-in functions for creating simple text UIs:

  • screen-clear: Clear the screen
  • screen-get-key: Get a keystroke
  • screen-write: Write a list, without parentheses, to an x and y position on the screen.
  • with-screen (macro): Enter/exit "screen" (UI) mode

The screen-... functions must occur within a with-screen expression. An example program shows these functions in action.

screen-get-key will return a single-character atom whose name matches the exact key pressed, except for the following keys:

  • INTR: interrupt (control-C)
  • EOF: end-of-file (control-D)
  • CLEAR: clear screen (control-L)
  • BSP: backspace
  • DEL: delete
  • DOWNARROW: down arrow
  • END: end key
  • LEFTARROW: left arrow
  • RIGHTARROW: right arrow
  • UPARROW: up arrow
  • ENTER: enter/return
  • ESC: escape key

An example use of the special keys could be to exit a loop if the user types control-C or control-D:

(with-screen
  (let ((continue t))
    (while continue
      ;; Do something
      (let ((k (screen-get-key)))
        (cond
          ;; Exit if control-C or control-D:
          ((or (= k 'INTR)
               (= k 'EOF))
           (set! continue ()))
          ;; Handle other keys...
          )))))

Loading Source Files

There are four ways of executing a source file, e.g. main.l1:

  1. (Unix/Linux/Mac only) Add a shebang at the beginning of the script;
  2. Pass main.l1 as an argument to l1: l1 main.l1, again at the shell prompt.
  3. Call the load function from the l1 REPL or from within another file: (load main.l1).
  4. "Compile" the source into an executable binary using l1c.

Option 3. is currently the only way of combining multiple source files into a single program. There is currently no packaging or namespacing functionality in l1.

Running l1 Programs as Command Line Scripts

Programs can be run by giving the program name as an argument to l1:

l1 hello.l1

However, if you add a "shebang" line #!/usr/bin/env l1 at the beginning of an l1 file:

#!/usr/bin/env l1
;; hello.l1
(printl '(hello world))

and set the execute bit on the file permissions:

chmod +x hello.l1

then you can run hello.l1 "by itself," without explicitly invoking l1:

$ ./hello.l1
hello world
$

Making Binary Executables

A script l1c is provided which allows one to build a stand-alone binary from an l1 program. It requires a working Go installation and probably only works on Linux / Mac / Un*x machines. Example:

$ cat hello.l1
(printl '(Hello, World!))

$ l1c hello.l1 -o hello
/var/folders/v2/yxn9b9h93lzgjsz645mblbt80000gt/T/tmp.QecBxTkp ~/Somedir/l1
go: creating new go.mod: module myprog
go: to add module requirements and sums:
    go mod tidy
go: finding module for package github.com/eigenhombre/l1/lisp
go: found github.com/eigenhombre/l1/lisp in github.com/eigenhombre/l1 v0.0.56
~/Somedir/l1
...

$ ls -l hello
-rwxr-xr-x  1 jacobsen  staff  3750418 Feb 14 10:46 hello
$ ./hello
Hello, World!
$

Because l1c is a Bash script tied to the Go language build system, you should install l1 via the go tool rather than via a downloaded binary if you wish to use l1c.

Emacs Integration

If you are using Emacs, you can set it up to work with l1 as an "inferior lisp" process as described in the Emacs manual. I currently derive a new major mode from the base lisp-mode and bind a few keys for convenience as follows:

(define-derived-mode l1-mode
  lisp-mode "L1 Mode"
  "Major mode for L1 Lisp code"
  (setq inferior-lisp-program (executable-find "l1")
  (paredit-mode 1)
  (put 'test 'lisp-indent-function 1)
  (put 'testing 'lisp-indent-function 1)
  (put 'errors 'lisp-indent-function 1)
  (put 'if 'lisp-indent-function 1)
  (put 'if-not 'lisp-indent-function 1)
  (put 'foreach 'lisp-indent-function 2)
  (put 'when-not 'lisp-indent-function 1)
  (define-key l1-mode-map (kbd "s-i") 'lisp-eval-last-sexp)
  (define-key l1-mode-map (kbd "s-I") 'lisp-eval-form-and-next)
  (define-key l1-mode-map (kbd "C-o j") 'run-lisp))

(add-to-list 'auto-mode-alist '("\\.l1" . l1-mode))

If l1 has been installed on your path, M-x run-lisp or using the appropriate keybinding should be enough to start a REPL within Emacs and start sending expressions to it.