31 KiB
slope
slope is an s-expression based language loosely modeled after scheme R4RS. It does, however, diverge quite a bit in some areas (especially IO and filesystem related constructs). slope also does not include many standard constructs of scheme (such as let, *let, case, etc.) in favor of using other existing constructs (this often involves slightly longer code that uses a narrower range of generalized constructs).
The interpreter is implemented in golang and many of the library differences between slope and scheme are based around generalizing the go standard library into a usable base system for a dynamically typed language. A great example of this generalization is that in slope you can use the function write
to easily write to a file (including stdout/err), a tcp connection, or a buffer object. slope knows what you are trying to do based on what you pass it, so separate file-write
, tcp-conn-write
, string-buf-write
functions are not necessary. Even with a runtime panic, slope will also attempt (I will stop short of guarantee) to close any open files or net connections. Other niceties are planned.
Why?
Like many of my projects: to learn and have fun. This interpreter is based on some existing code. I rewrote sections of the lexer and parser (areas I enjoy working in), left the apply and environment setup alone, and made a few small changes to the eval function. As such, this could be considered a fork of that code with some added core language constructs as well as a significantly expanded library (another area I enjoy working in). I have definitely learned different things than my previous language project while working on this codebase, and I think have already (early days as I am writing this) ended up with a much more realistically usable language (though that is certainly contextual).
What is implemented?
As mentioned above, this is not a scheme implementation in any proper sense and it will go its own way when it sees fit.
Types
slope recognizes a few types, that are generally created and used dynamically:
number
- implemented as
float64
under the hood - hexidecimal can be written with a leading
0x
, as in0xFFF
- octal can be written with a leading
0
, as in077
- in strings, values can be escaped to characters as decimal, octal, or hex:
"\27[1mI am bold text\033[0m, I am not bold. I am a hex based char: \0x84"
- implemented as
symbol
- strings that represent language objects/constructs/vars
- can be created using
(quote my-symbol)
or'my-symbol
bool
- as
#t
and#f
- anything can be cast to bool with a
#t
value, except#f
(the only truly falsy value without using~bool
, see below)
- as
string
- anything in "quotes"
- allows for backslash escapes within the quotes (for chars by number or shortcuts to tab, newline, etc)
list
/pair
- includes the empty list (null)
- lists are the primary data structure
exception
- an error response
- can be created using
(! "My exception text")
- can be tested for with
(exception? some-var)
- two functions exist to set behavior for exceptions:
(exception-mode-panic)
, will panic and halt execution when an exception is encountered (default)(exception-mode-pass)
, will return the exception like any other value and it can be handled- the two modes can be switched on and off at various points in the code
There is also the concept/type IOHandle
. This type cannot be cast to another (except string, but the output will likely not be useful for much beyond debugging), nor can it be entered manually. It is returned by things like net-conn
and file-open-read
. It is the general concept of something that can be read from and/or written to.
Special Forms
There are a number of special forms in the language that will allow, for example, un-initialized symbols to be passed (for example, in define
), among other things less available from regular lambdas (which are themselves a special form).
quote
- Can be shorthanded so that
'(1 2 3)
is equal to(quote (1 2 3))
- Can be shorthanded so that
if
and
or
cond
set!
define
- Lexically scoped, such that its use from within a lambda scopes any defined variables to the scope of that lambda and its child scopes
lambda
- The body of a lambda has an implied
(begin ...)
surrounding it. As such, you may place non-nested procedure calls or values inside of a lambda body and the last will be returned - Variadic lambdas can be created by using the special param name
args-list
(which will be a list of all of the remaining arguments when accessed rom the lambda body). This param should be the last param in the lambda definition, as it will eat all arguments that come after its position. Any params defined afterargs-list
have undefined behavior (they will generally not be created)
- The body of a lambda has an implied
begin
- Can be shorthanded so that
{(+ 1 2) (display "Hi")}
is equal to(begin (+ 1 2) (display "Hi"))
- Can be shorthanded so that
begin0
- Like begin, but returns the value resulting from the evaluation of the first expression. No shorthand exists for
begin0
- Like begin, but returns the value resulting from the evaluation of the first expression. No shorthand exists for
filter
- Takes a procedure that, in turn, takes a single argument. Each value from the second argument, a list, will be passed to the procedure. Anything truthy results are returned in a new list
load
- Loads and executes a slope file. Absolute path to file must be used
load-mod
- Loads a module. Module name should be used
load-mod-file
- Used within modules to load additional files
Special Forms API Description
Anything inside `[]` is a placeholder representing an argument. Anything inside `` is a placeholder representing an _optional_ argument. `...` indicates a variadic value.
What the procedure returns is indicated, though any procedure can return an exception in addition to the given type.
(quote [expression...])
:bool
(if [condition check: expression] [true branch: expression] [[false branch: expression]])
:value
(and [expression...])
:bool
(or [expression...])
:bool
(cond [([condition check: expression] [expression])...])
:value
(set! [variable name: symbol] [value|expression])
:value/expression
(define [variable name: symbol] [value|expression|procedure])
:value/expression
(lambda [([[argument: symbol...]])] [[expression...]])
:procedure
(begin [expression...])
:value
(begin0 [expression...])
:value
(filter [test: procedure] [list])
:list
(load [filepath: string...])
:symbol
(load-mod [filepath: string...])
:symbol
(load-mod-file [filepath: string...])
:symbol
Library
Values
PI
, E
Number related
Implemented:
positive?
, negative?
, zero?
, abs
, floor
, ceil
, round
, +
, -
, *
, /
, min
, max
, number->string
Not implemented, but on my radar:
modulo
, remainder
, quotient
, expt
, sin
, cos
, tan
, atan
, sqrt
Number API Description
Anything inside `[]` is a placeholder representing an argument. Anything inside `` is a placeholder representing an _optional_ argument. `...` indicates a variadic value.
What the procedure returns is indicated, though any procedure can return an exception in addition to the given type.
(positive? [number])
:bool
(negative? [number])
:bool
(zero? [number])
:bool
(abs [number])
:number
(floor [number])
:number
(ceil [number])
:number
(round [number] [[number: decimal places to round to]])
:number
(+ [number...])
:number
(- [number...])
:number
(* [number...])
:number
(/ [number...])
:number
(min [number...])
:number
(max [number...])
:number
(number-> string [number])
:number
Conditional
Implemented:
>
, >=
, <
, <=
, not
, equal?
, number?
, string?
, bool?
, symbol?
, pair?
, null?
, list?
, procedure?
, atom?
, ~bool
, assoc?
, string-buf?
, io-handle?
~bool
is non-standard and will cast any value as a boolean value using a looser set of rules. Under normal conditional rules in slope anythong other than #f
is considered truthy (#t
). When using ~bool
the number 0
is false, the list empty list ()
is false, and empty string ""
is false, and a closed IOHandle is false. Everything else, except the literal value #f
, is true (#t
).
So...
> (if (~bool 0) "Truthy" "Falsy")
#1=> Falsy
> (if 0 "Truthy" "Falsy")
#2=> Truthy
Conditional API Description
Anything inside `[]` is a placeholder representing an argument. Anything inside `` is a placeholder representing an _optional_ argument. `...` indicates a variadic value.
What the procedure returns is indicated, though any procedure can return an exception in addition to the given type.
(> [number] [number])
:bool
(>= [number] [number])
:bool
(< [number] [number])
:bool
(<= [number] [number])
:bool
(not [value])
:bool
(equal? [value] [value])
:bool
(number? [value])
:bool
(string? [value])
:bool
(bool? [value])
:bool
(symbol? [value])
:bool
(pair? [value])
:bool
(list? [value])
:bool
(null? [value])
:bool
(procedure? [value])
:bool
(atom? [value])
:bool
(assoc? [value])
:bool
(io-handle? [value])
:bool
(string-buf? [value])
:bool
(~bool [value])
:bool
List related
Implemented:
length
*, cons
, car
, cdr
, append
, list
, map
, for-each
, list->string
, list-ref
, list-sort
, reverse
, assoc-get
, assoc-set
, member?
length
andreverse
will take a string or a list.list-sort
will sort a list in ascending order by casting all elements within the list as strings and doing a compare. The exception to this is numbers, which will be compared among themselves as numbers. In a mixed list numbers will come after other values. To sort descending, pass the results oflist-sort
toreverse
List API Description
Anything inside `[]` is a placeholder representing an argument. Anything inside `` is a placeholder representing an _optional_ argument. `...` indicates a variadic value.
What the procedure returns is indicated, though any procedure can return an exception in addition to the given type.
(length [list|string|exception|IOHandle:string-buf])
:number
(list [value...])
:list
(cons [value] [list])
:list
(car [list])
:list
(cdr [list])
:list
(append [list] [value...])
:list
(map [procedure] [list...])
:list
(for-each [procedure] [list...])
:list
(list->string [list])
:string
(list-ref [list] [number])
:value
(list-sort [list])
:list
(reverse [list|string])
:list
(assoc-get [assoc-list] [value])
:value
(assoc-set [assoc-list] [value] [value])
:value
(member? [list] [value])
:bool
String related
Implemented:
length
*, substring
, string-append
, string->number
, string-format
, string->list
, string-trim-space
, string-index-of
, string-ref
, string-upper
, string-lower
, string-make-buf
, string-buf-clear
, regex-match?
, regex-find
, regex-replace
, string-fields
length
will take a string, an IOHandle representing a string-buf, or a list. In the case of a string it will return the number of runes as thought of in golang. Loosely equivalent to characters, rather than bytes.- Standard read and write procedures work with the IOHandle returned from
string-make-buf
length
andreverse
will take a string or a list.string-format
works like a simplified sprintf. All values are represented by%v
(for value). You can pad values with spaces on the right by adding the desired field width after the percent:%10v
with the string"hello"
would produce" hello"
and you can left justify the text with a-
after the percent:%-10v
for the same variable would produce"hello "
. There are no other modifiers.
String API Description
Anything inside `[]` is a placeholder representing an argument. Anything inside `` is a placeholder representing an _optional_ argument. `...` indicates a variadic value.
What the procedure returns is indicated, though any procedure can return an exception in addition to the given type.
(length [list|string|exception|IOHandle:string-buf])
:number
(list [value...])
:list
(substring [string] [number] [number])
:string
(string-append [string...])
:string
(string-format [format: string] [value...])
:string
(replaces `{}`'s in string with successive values)(string->number [string])
:number
(string->list [string] [[value]])
:list
(splits the first string at each instance of value, cast as a string)(string-fields [string])
:list
(string-trim-space [string])
:string
(string-index-of [string] [string])
:number
(`-1` if the string could not be found)(string-ref [string] [number])
:string
(string-upper [string])
:string
(string-lower [string])
:string
(string-make-buf)
:IOHandle:string-buf
(string-buf-clear [IOHandle:string-buf])
:()
(regex-match? [string (pattern)] [string])
:bool
(regex-find [string (pattern)] [string])
:list
(regex-replace [string (pattern)] [string])
:string
IO Related
Implemented:
newline
, display
, display-lines
, write
, write-raw
, read-line
, read-char
, read-all
, close
file-create
, file-open-read
, file-open-write
, file-append-to
, read-all-lines
Not implemented, but on my radar:
read-bytes
, file-seek
IO API Description
Anything inside `[]` is a placeholder representing an argument. Anything inside `` is a placeholder representing an _optional_ argument. `...` indicates a variadic value.
What the procedure returns is indicated, though any procedure can return an exception in addition to the given type.
(newline)
:()
(display [value...])
:()
(display-lines [value...])
:()
(write [string] [[IOHandle]])
:()|IOHandle
(write-raw [string] [[IOHandle]])
:()|IOHandle
(read-all [[IOHandle]])
:string
(read-line [[IOHandle]])
:string|symbol:EOF
(read-char [[IOHandle]])
:string
(read-all-lines [[IOHandle]])
:list
(close [IOHandle])
:()
(file-create [filepath: string])
:IOHandle
(file-open-read [filepath: string])
:IOHandle
(file-open-write [filepath: string])
:IOHandle
(file-append-to [filepath: string] [string...])
:()
System Related
Implemented:
exit
, license
, apply
, !
, exception-mode-panic
, exception-mode-pass
, sys-args
, term-size
, term-restore
, term-char-mode
, term-raw-mode
, term-cooked-mode
, term-sane-mode
(license)
will print the license terms for slope, this is mostly input as a convenience in the repl.- The various term modes are an experimental feature at the moment, but should work as described
System API Description
Anything inside `[]` is a placeholder representing an argument. Anything inside `` is a placeholder representing an _optional_ argument. `...` indicates a variadic value.
What the procedure returns is indicated, though any procedure can return an exception in addition to the given type.
(exit [[number]])
:()
(license)
:()
(apply [procedure] [arguments: list])
:value
(! [string])
:exception
(exception-mode-panic)
:()
(exception-mode-pass)
:()
(sys-args)
:list
(term-size)
:list
(term-restore)
:()
(term-char-mode)
:()
(term-raw-mode)
:()
(term-cooked-mode)
:()
(term-sane-mode)
:()
OS Related
path-exists?
, path-is-dir?
, path-abs
, path-join
, path-ext
, path-glob
, chmod
, chdir
, env
, subprocess
Not implemented, but on my radar:
path-dir-contents
, path-dir
, path-base
, chown
, mkdir
OS API Description
Anything inside `[]` is a placeholder representing an argument. Anything inside `` is a placeholder representing an _optional_ argument. `...` indicates a variadic value.
What the procedure returns is indicated, though any procedure can return an exception in addition to the given type.
(path-exists? [filepath: string])
:bool
(path-is-dir? [filepath: string])
:bool
(path-abs [filepath: string])
:string
(path-join [string])
:string
(path-extension [filepath: string] [[replacement-extension: string]])
:string
(path-glob [filepath: string])
:list
(chmod [filepath: string] [file-mode: number])
:()
(chdir [filepath: string])
:()
(env [[env-key: string]] [[env-value: string]])
:list|string|()
(subprocess [list] [[output-redirection: IOHandle|#f]] [[error-redirection: IOHandle|#f]])
:number
(Passing `#f` to output redirection allows you to do stdout while still redirecting stderr. The `#f` for stderr only exists for symetry)
Net Related
Implemented:
net-conn
For the time being, net-conn
can be used to make most requests. At the moment the tls settings are not particularly secure and I am looking into ways to allow for more granular customization of both the timeout and the tls without weighing down the function or introducing more types.
Net API Description
Anything inside `[]` is a placeholder representing an argument. Anything inside `` is a placeholder representing an _optional_ argument. `...` indicates a variadic value.
What the procedure returns is indicated, though any procedure can return an exception in addition to the given type.
(net-conn [host: string] [port: string|number] [[use-tls: bool]] [[timeout-seconds: number]])
:IOHandle
Date/Time Related
Implemented:
timestamp
, date
, timestamp->date
, date->timestamp
, date-format
, sleep
, date-default-format
Date/Time API Description
Anything inside `[]` is a placeholder representing an argument. Anything inside `` is a placeholder representing an _optional_ argument. `...` indicates a variadic value.
What the procedure returns is indicated, though any procedure can return an exception in addition to the given type.
(timestamp)
:timestamp: number
(date)
:date: string
(timestamp->date [timestamp: string|number] [[format: string]])
:string
(date->timestamp [date: string] [format: string])
:timestamp: number
(date-format [start-format: string] [date: string] [end-format: string])
:string
(sleep [[milliseconds: number]])
:()
(date-default-format)
:string
(returns a format string for the default format produced bydate
)
In the above procedures there are strings listed as "format". The format string is similar to the format string used in PHP, Python, etc.
Certain characters are escaped by %
symbols followed by a single character. Here is the list of the conversions:
%a
- 12 hour segment, lowercase:
pm
%A
- 12 hour segment, uppercase:
PM
%d
- Day without leading zero:
4
%D
- Day with leading zero:
04
%e
- Day without leading zero, but with space padding:
4
%f
- Month name, short:
Jan
%F
- Month name, long:
January
%g
or%G
- Hour, 24 hour format:
15
%h
- Hour, 12 hour format without leading zero:
2
%H
- Hour, 12 hour format with leading zero:
02
%i
- Minutes without leading zero:
4
%I
- Minutes with leading zero:
04
%m
- Month number, without leading zero:
6
%M
- Month number, with leading zero:
06
%o
- Timezone offset, without minutes:
-07
%O
- Timezone offset, with minutes:
-0700
%s
- Seconds, without leading zero:
5
%S
- Seconds, with leading zero:
05
%w
- Short weekeday:
Wed
%W
- Long weekday:
Wednesday
%y
year: %Y
- Four digit year:
2021
%Z
- Time zone, as three chars:
UTC
,PST
, etc. %
- Will yield a literal
%
anything else
- A percent followed by an unrecognized char will yield a
?
, as will a hanging % as the last char in the string
21
So the string: "%F %e, %Y %h:%I%a"
would be equivalend to something like: "August 8, 2021 4:03pm"
When a string is converted to a date and no time zone information is present, slope will assume the user's local time.
*the general write, read, and close operations in the IO section apply to net objects as well as files and string buffers
Preload
slope allows for files in a specific directory to be automatically loaded. For this to occur slope should be invoked with the -L
flag. When that flag is passed slope will do its normal setup routine then preload all files in the preload folder then go on to whatever its main task is (repl, run file, run single line from command).
slope will use the variable $SLOPE_PRELOAD_DIR
if it is defined and not empty. Otherwise, if $XDG_DATA_HOME
is defined and not empty it will use $XDG_DATA_HOME/slope/preload/
. Lastly it will use ~/.local/share/slope/preload/
.
Unlike modules, arbitrary code execution will occur with preloads (not just define
expressions). The rationalle is that you, the user, have put this code there and as such it is considered safe and in your control. Preloads are a really easy way to make procedures and variables outside of the standard library, but that you regularly use, available at a repl session without having to manually load them.
Modules
slope has a very basic module system. When you use the load
procedure you are loading a single file from a given path. Anything in that file will be run when it is loaded. Using load-mod
allows you to load a module from a designated location on your system by simply passing the module's name (as represented by the directory name it is contained within).
load-mod
will load modules from the first non-empty-string value it finds out of these options:
$SLOPE_MOD_PATH
$XDG_DATA_HOME/slope/modules
~/.local/share/slope/modules
So, a hypethetical module named test
would be found at ~/.local/share/slope/modules/test
.
A separate tool is coming soon to help manage acquiring/removing/viewing modules as well as dealing with subdependencies. Until then it is a more manual process.
How Modules are Processed
When load-mod
is called the module will be located and main.slo
will be parsed. All expressions are ignored except define
, load-mod
, and load-mod-file
. Arbitrary code cannot be run (there are no side effects to load-mod
or load-mod-file
other than populating the global environment with the defined values from the given module).
A define
expression will still be fully evaluated. As such, something like: (define twenty (+ 15 5))
is perfectly fine. But something like (display "I am in a module")
will not be evaluated.
Calling load-mod
from within a module allows modules to have their own dependencies. Calling load-mod-file
allows modules to be made up of more files than just main.slo
. More info about files can be found in the next section.
Module Files
main.slo
Taking our above example of a module named test
, the bare minimum that the folder ~/.local/share/slope/modules/test
should contain is a file named main.slo
. When a module is loaded, this is the file that is looked for. If it is not present, loading will fail and a runtime panic will result.
Other slope source files
You may, optionally, have other code files present in the module and can load them by calling load-mod-file
. This procedure should be given a path relative to the module root directory. So within our main.slo
file we might call something like: (load-mod-file "./submodules/unit-tests.slo")
. Something to note is that a call to load-mod-file
cannot jump out of the module's root directory (doing something like ../../../some-other-code.py
or /etc/something
will result in a runtime panic).
module.info
Another special file that is not yet in use, but is planned, is module.info
. This file will contain details about the module: name, dependencies, repo url, description, version number, commit-hash, license, etc. This will be used by an eventual tool that will manage modules and dependencies.
Building
A Makefile is available. Running make
will build in the local directory. Running sudo make install
will install the software (sudo may or may not be needed depending on your setup, or the functionality may be provided by a different tool such as doas). slope was written with Go 1.16, but will likely build on older versions with limited modifications.
Running
All of these assume slosh is on your $PATH
:
To run the repl:
slope
From the repl you can execute (license)
to view the license.
To run a one liner:
slope -run "(display (+ 1 3))"
To run a file:
slope ./myfile.slo
To see command line options:
slope -h
To see version information:
slope -v
TODO
- Add termios, net-serve, and other procedures
- Maybe add some form of character primitive type (will make reading by either byte or char from a file easier)
- Build cool things with slope
- Think out a basic package system with imports