Orchid/orchid
1
0
Fork 1
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

backup commit

Browse Source
This commit is contained in:
Lawrence Bethlenfalvy
2022-10-24 03:16:04 +01:00
parent fbbd6ed256
commit 778c87db77
43 changed files with 1156 additions and 174 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
Cargo.lock generated
View File

@@ -160,6 +160,7 @@ dependencies = [
"itertools",
"mappable-rc",
"ordered-float",
"smallvec",
"thiserror",
]
@@ -196,6 +197,12 @@ dependencies = [
"proc-macro2",
]
[[package]]
name = "smallvec"
version = "1.10.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a507befe795404456341dfab10cef66ead4c041f62b8b11bbb92bffe5d0953e0"
[[package]]
name = "syn"
version = "1.0.95"

1
Cargo.toml
View File

@@ -13,3 +13,4 @@ hashbrown = "0.12"
mappable-rc = "0.1"
ordered-float = "3.0"
itertools = "0.10"
smallvec = "1.10.0"

61
examples/dummy_project/main.orc
View File

@@ -1,23 +1,46 @@
-- import std::io::(println, out) -- imports
opaque := \T. T
-- single word rule (alias)
greet =1=> (\name. printf out "Hello {}!\n" [name])
-- multi-word exported rule
export ;> $a =200=> (greet $a)
reeee := \$a.b
-- single-word exported rule
export main := (
print "What is your name?" >>
readln >>= \name.
greet name
--[ Typeclass definition (also just a type) ]--
define Add $L:type $R:type $O:type as $L -> $R -> $O
-- HKTC
define Mappable $C:(type -> type) as @T. @U. (T -> U) -> $C T -> $C U
-- Dependency on existing typeclass
define Zippable $C:(type -> type) as @:Mappable $C. (
@T. @U. @V. (T -> U -> V) -> $C T -> $C U -> $C V
)
define Default $T:type as $T
-- Is the intersection of typeclasses an operation we need?
export < $a ...$rest /> := (createElement (tok_to_str $a) [(props_carriage ...$rest)])
export (props_carriage $key = $value) := (tok_to_str $key) => $value
--[ Type definition ]--
define Cons $elem:type as loop \r. Option (Pair T $elem)
nil := @T. from @(Cons T) none
cons := @T. \el:T. (
generalise @(Cons T)
|> (\list. some t[el, into list])
|> categorise @(Cons T)
)
export map := @T. @U. \f:T -> U. (
generalise @(Cons T)
|> loop ( \recurse. \option.
map option \pair. t[f (fst pair), recurse (snd pair)]
)
|> categorise @(Cons U)
)
-- Universal typeclass implementation; no parameters, no overrides, no name for overriding
impl Mappable Cons via map
-- Blanket typeclass implementation; parametric, may override, must have name for overriding
impl (@T. Add (Cons T) (Cons T) (Cons T)) by concatenation over elementwiseAddition via concat
-- The broadest trait definition in existence
Foo := (Bar Baz)
-- default anyFoo = @T. @impl:(T (Bar Baz)). impl:(T Foo)
-- Scratchpad
filterBadWords := @C:type -> type. @:Mappable C. \strings:C String. (
map strings \s. if intersects badWords (slice " " s) then none else some s
):(C (Option String))
-- /Scratchpad
main := \x. foo @bar x
foo := @util. \x. util x
export opaque := \T. atom

18
examples/rule_demo/main.orc
View File

@@ -1,18 +0,0 @@
export ::(main, foo)
main := [foo, bar, baz, quz]
foo := steamed hams
[...$data] := (cons_start ...$data cons_carriage(none))
[] := none
...$prefix:1 , ...$item cons_carriage(
$tail
) := ...$prefix cons_carriage(
(some (cons (...$item) $tail))
)
cons_start ...$item cons_carriage($tail) := some (cons (...$item) $tail)

18
notes.md
View File

@@ -1,18 +0,0 @@
# Anatomy of a code file
```orchid
import std::io::(println, out) -- imports
-- single word substitution (alias)
greet == \name. printf out "Hello {}!\n" [name]
-- multi-word exported substitution with nonzero priority
export (...$pre ;) $a ...$post) =200=> (...$pre (greet $a) ...$post)
-- single-word exported substitution
export main == (
print "What is your name? >>
readln >>= \name.
greet name
)
```

35
notes/type_system/definitions.md Normal file
View File

@@ -0,0 +1,35 @@
## Type definitions
```orc
define Cons as \T:type. loop \r. Option (Pair T r)
```
Results in
- (Cons Int) is not assignable to @T. Option T
- An instance of (Cons Int) can be constructed with `categorise @(Cons Int) (some (pair 1 none))`
but the type parameter can also be inferred from the expected return type
- An instance of (Cons Int) can be deconstructed with `generalise @(Cons Int) numbers`
but the type parameter can also be inferred from the argument
These inference rules are never reversible
```orc
categorise :: @T:type. (definition T) -> T
generalise :: @T:type. T -> (definition T)
definition :: type -> type -- opaque function
```
## Unification
The following must unify:
```orc
@T. @:Add T T T. Mult Int T T
Mult Int (Cons Int) (Cons Int)
```
### Impls for types
Impls for types are generally not a good idea as autos with types like Int can
often be used in dependent typing to represent eg. an index into a type-level conslist to be
deduced by the compiler, and impls take precedence over resolution by unification.

0
notes/type_system/impls.md Normal file
View File

27
notes/type_system/unification.md Normal file
View File

@@ -0,0 +1,27 @@
# Steps of validating typed lambda
- Identify all expressions that describe the type of the same expression
- enqueue evaluation steps for each of them and put them in a unification group
- evaluation step refers to previous step, complete expression tree
- unification **succeeds** if either
- the trees are syntactically identical in any two steps between the targets
- unification succeeds for all substeps:
- try to find an ancestor step that provably produces the same value as any lambda in this
step (for example, by syntactic equality)
- if found, substitute it with the recursive normal form of the lambda
- recursive normal form is `Apply(Y, \r.[body referencing r on point of recursion])`
- find all `Apply(\x.##, ##)` nodes in the tree and execute them
- unification **fails** if a member of the concrete tree differs (only outermost steps add to
the concrete tree so it belongs to the group and not the resolution) or no substeps are found
for a resolution step _(failure: unresolved higher kinded type)_
- if neither of these conclusions is reached within a set number of steps, unification is
**indeterminate** which is also a failure but suggests that the same value-level operations
may be unifiable with better types.
The time complexity of this operation is O(h no) >= O(2^n). For this reason, a two-stage limit
is recommended: one for the recursion depth which is replicable and static, and another,
configurable, time-based limit enforced by a separate thread.
How does this interact with impls?
Idea: excluding value-universe code from type-universe execution.
Digression: Is it possible to recurse across universes?

102
src/executor/foreign.rs Normal file
View File

@@ -0,0 +1,102 @@
use std::any::Any;
use std::fmt::{Display, Debug};
use std::hash::Hash;
use mappable_rc::Mrc;
use crate::representations::typed::{Expr, Clause};
pub trait ExternError: Display {}
/// Represents an externally defined function from the perspective of the executor
/// Since Orchid lacks basic numerical operations, these are also external functions.
#[derive(Eq)]
pub struct ExternFn {
name: String, param: Mrc<Expr>, rttype: Mrc<Expr>,
function: Mrc<dyn Fn(Clause) -> Result<Clause, Mrc<dyn ExternError>>>
}
impl ExternFn {
pub fn new<F: 'static + Fn(Clause) -> Result<Clause, Mrc<dyn ExternError>>>(
name: String, param: Mrc<Expr>, rttype: Mrc<Expr>, f: F
) -> Self {
Self {
name, param, rttype,
function: Mrc::map(Mrc::new(f), |f| {
f as &dyn Fn(Clause) -> Result<Clause, Mrc<dyn ExternError>>
})
}
}
fn name(&self) -> &str {&self.name}
fn apply(&self, arg: Clause) -> Result<Clause, Mrc<dyn ExternError>> {(self.function)(arg)}
}
impl Clone for ExternFn { fn clone(&self) -> Self { Self {
name: self.name.clone(),
param: Mrc::clone(&self.param),
rttype: Mrc::clone(&self.rttype),
function: Mrc::clone(&self.function)
}}}
impl PartialEq for ExternFn { fn eq(&self, other: &Self) -> bool { self.name() == other.name() }}
impl Hash for ExternFn {
fn hash<H: std::hash::Hasher>(&self, state: &mut H) { self.name.hash(state) }
}
impl Debug for ExternFn {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
write!(f, "##EXTERN[{}]:{:?} -> {:?}##", self.name(), self.param, self.rttype)
}
}
pub trait Atomic: Any + Debug where Self: 'static {
fn as_any(&self) -> &dyn Any;
fn definitely_eq(&self, _other: &dyn Any) -> bool;
fn hash(&self, hasher: &mut dyn std::hash::Hasher);
}
/// Represents a unit of information from the perspective of the executor. This may be
/// something like a file descriptor which functions can operate on, but it can also be
/// information in the universe of types or kinds such as the type of signed integers or
/// the kind of types. Ad absurdum it can also be just a number, although Literal is
/// preferable for types it's defined on.
#[derive(Eq)]
pub struct Atom {
typ: Mrc<Expr>,
data: Mrc<dyn Atomic>
}
impl Atom {
pub fn new<T: 'static + Atomic>(data: T, typ: Mrc<Expr>) -> Self { Self{
typ,
data: Mrc::map(Mrc::new(data), |d| d as &dyn Atomic)
} }
pub fn data(&self) -> &dyn Atomic { self.data.as_ref() as &dyn Atomic }
pub fn try_cast<T: Atomic>(&self) -> Result<&T, ()> {
self.data().as_any().downcast_ref().ok_or(())
}
pub fn is<T: 'static>(&self) -> bool { self.data().as_any().is::<T>() }
pub fn cast<T: 'static>(&self) -> &T {
self.data().as_any().downcast_ref().expect("Type mismatch on Atom::cast")
}
}
impl Clone for Atom {
fn clone(&self) -> Self { Self {
typ: Mrc::clone(&self.typ),
data: Mrc::clone(&self.data)
} }
}
impl Hash for Atom {
fn hash<H: std::hash::Hasher>(&self, state: &mut H) {
self.data.hash(state);
self.typ.hash(state)
}
}
impl Debug for Atom {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
write!(f, "##ATOM[{:?}]:{:?}##", self.data(), self.typ)
}
}
impl PartialEq for Atom {
fn eq(&self, other: &Self) -> bool {
self.data().definitely_eq(other.data().as_any())
}
}

3
src/executor/mod.rs Normal file
View File

@@ -0,0 +1,3 @@
mod foreign;
pub use foreign::ExternFn;
pub use foreign::Atom;

70
src/main.rs
View File

@@ -1,13 +1,18 @@
#![feature(specialization)]
use std::{env::current_dir, process::exit};
use std::env::current_dir;
mod executor;
mod parse;
mod project;
mod utils;
mod expression;
mod representations;
mod rule;
use expression::{Expr, Clause};
mod types;
use file_loader::LoadingError;
pub use representations::ast;
use ast::{Expr, Clause};
use representations::typed as t;
use mappable_rc::Mrc;
use project::{rule_collector, Loaded, file_loader};
use rule::Repository;
@@ -34,37 +39,60 @@ export (match_sequence $lhs) >>= (match_sequence $rhs) =100=> (bind ($lhs) ($rhs
fn initial_tree() -> Mrc<[Expr]> {
to_mrc_slice(vec![Expr(Clause::Name {
local: None,
qualified: to_mrc_slice(vec!["main".to_string(), "main".to_string()])
}, None)])
qualified: literal(&["main", "main"])
}, to_mrc_slice(vec![]))])
}
fn main() {
#[allow(unused)]
fn typed_notation_debug() {
let t = t::Clause::Auto(None,
t::Clause::Lambda(Some(Mrc::new(t::Clause::Argument(0))),
t::Clause::Lambda(Some(Mrc::new(t::Clause::Argument(1))),
t::Clause::Argument(1).wrap_t(t::Clause::Argument(2))
).wrap()
).wrap()
).wrap();
let f = t::Clause::Auto(None,
t::Clause::Lambda(Some(Mrc::new(t::Clause::Argument(0))),
t::Clause::Lambda(Some(Mrc::new(t::Clause::Argument(1))),
t::Clause::Argument(0).wrap_t(t::Clause::Argument(2))
).wrap()
).wrap()
).wrap();
println!("{:?}", t::Clause::Apply(t::Clause::Apply(Mrc::clone(&t), t).wrap(), f))
}
#[allow(unused)]
fn load_project() {
let cwd = current_dir().unwrap();
let collect_rules = rule_collector(move |n| {
let collect_rules = rule_collector(move |n| -> Result<Loaded, LoadingError> {
if n == literal(&["prelude"]) { Ok(Loaded::Module(PRELUDE.to_string())) }
else { file_loader(cwd.clone())(n) }
}, vliteral(&["...", ">>", ">>=", "[", "]", ",", "=", "=>"]));
match collect_rules.try_find(&literal(&["main"])) {
Ok(rules) => {
let rules = match collect_rules.try_find(&literal(&["main"])) {
Ok(rules) => rules,
Err(err) => panic!("{:#?}", err)
};
let mut tree = initial_tree();
println!("Start processing {tree:?}");
let repo = Repository::new(rules.as_ref().to_owned());
println!("Ruleset: {repo:?}");
let mut i = 0; loop {
if 10 <= i {break} else {i += 1}
xloop!(let mut i = 0; i < 10; i += 1; {
match repo.step(Mrc::clone(&tree)) {
Ok(Some(phase)) => {
println!("Step {i}: {phase:?}");
tree = phase;
println!("Step {i}: {tree:?}")
},
Ok(None) => exit(0),
Err(e) => {
eprintln!("Rule error: {e:?}");
exit(0)
}
}
}
}
Err(err) => println!("{:#?}", err)
Ok(None) => {
println!("Execution complete");
break
},
Err(e) => panic!("Rule error: {e:?}")
}
}; println!("Macro execution didn't halt"));
}
fn main() {
// lambda_notation_debug();
load_project();
}

6
src/parse/enum_parser.rs
View File

@@ -1,3 +1,9 @@
/// Produces parsers for tokenized sequences of enum types:
/// ```rs
/// enum_parser!(Foo::Bar | "Some error!") // Parses Foo::Bar(T) into T
/// enum_parser!(Foo::Bar) // same as above but with the default error "Expected Foo::Bar"
/// enum_parser!(Foo >> Quz; Bar, Baz) // Parses Foo::Bar(T) into Quz::Bar(T) and Foo::Baz(U) into Quz::Baz(U)
/// ```
#[macro_export]
macro_rules! enum_parser {
($p:path | $m:tt) => {

30
src/parse/expression.rs
View File

@@ -1,11 +1,11 @@
use chumsky::{self, prelude::*, Parser};
use mappable_rc::Mrc;
use crate::enum_parser;
use crate::expression::{Clause, Expr, Literal};
use crate::utils::to_mrc_slice;
use crate::representations::{Literal, ast::{Clause, Expr}};
use crate::utils::{to_mrc_slice, one_mrc_slice};
use super::lexer::Lexeme;
/// Parses any number of expr wrapped in (), [] or {}
fn sexpr_parser<P>(
expr: P
) -> impl Parser<Lexeme, Clause, Error = Simple<Lexeme>> + Clone
@@ -13,6 +13,8 @@ where P: Parser<Lexeme, Expr, Error = Simple<Lexeme>> + Clone {
Lexeme::paren_parser(expr.repeated()).map(|(del, b)| Clause::S(del, to_mrc_slice(b)))
}
/// Parses `\name.body` or `\name:type.body` where name is any valid name and type and body are
/// both expressions. Comments are allowed and ignored everywhere in between the tokens
fn lambda_parser<P>(
expr: P
) -> impl Parser<Lexeme, Clause, Error = Simple<Lexeme>> + Clone
@@ -37,6 +39,7 @@ where P: Parser<Lexeme, Expr, Error = Simple<Lexeme>> + Clone {
})
}
/// see [lambda_parser] but `@` instead of `\` and the name is optional
fn auto_parser<P>(
expr: P
) -> impl Parser<Lexeme, Clause, Error = Simple<Lexeme>> + Clone
@@ -50,17 +53,22 @@ where P: Parser<Lexeme, Expr, Error = Simple<Lexeme>> + Clone {
.then_ignore(enum_parser!(Lexeme::Comment).repeated())
.ignore_then(expr.clone().repeated())
.then_ignore(enum_parser!(Lexeme::Comment).repeated())
.or_not().map(Option::unwrap_or_default)
)
.then_ignore(just(Lexeme::name(".")))
.then_ignore(enum_parser!(Lexeme::Comment).repeated())
.then(expr.repeated().at_least(1))
.try_map(|((name, typ), body), s| if name.is_none() && typ.is_empty() {
.try_map(|((name, typ), body): ((Option<String>, Vec<Expr>), Vec<Expr>), s| {
if name.is_none() && typ.is_empty() {
Err(Simple::custom(s, "Auto without name or type has no effect"))
} else {
Ok(Clause::Auto(name, to_mrc_slice(typ), to_mrc_slice(body)))
}
})
}
/// Parses a sequence of names separated by :: <br/>
/// Comments are allowed and ignored in between
fn name_parser() -> impl Parser<Lexeme, Vec<String>, Error = Simple<Lexeme>> + Clone {
enum_parser!(Lexeme::Name).separated_by(
enum_parser!(Lexeme::Comment).repeated()
@@ -69,6 +77,7 @@ fn name_parser() -> impl Parser<Lexeme, Vec<String>, Error = Simple<Lexeme>> + C
).at_least(1)
}
/// Parse any legal argument name starting with a `$`
fn placeholder_parser() -> impl Parser<Lexeme, String, Error = Simple<Lexeme>> + Clone {
enum_parser!(Lexeme::Name).try_map(|name, span| {
name.strip_prefix('$').map(&str::to_string)
@@ -76,7 +85,7 @@ fn placeholder_parser() -> impl Parser<Lexeme, String, Error = Simple<Lexeme>> +
})
}
/// Parse an expression without a type annotation
/// Parse an expression
pub fn xpr_parser() -> impl Parser<Lexeme, Expr, Error = Simple<Lexeme>> {
recursive(|expr| {
let clause =
@@ -102,12 +111,17 @@ pub fn xpr_parser() -> impl Parser<Lexeme, Expr, Error = Simple<Lexeme>> {
}),
sexpr_parser(expr.clone()),
lambda_parser(expr.clone()),
auto_parser(expr.clone())
auto_parser(expr.clone()),
just(Lexeme::At).to(Clause::Name {
local: Some("@".to_string()),
qualified: one_mrc_slice("@".to_string())
})
))).then_ignore(enum_parser!(Lexeme::Comment).repeated());
clause.clone().then(
just(Lexeme::Type)
.ignore_then(expr.clone()).or_not()
.ignore_then(clause.clone())
.repeated()
)
.map(|(val, typ)| Expr(val, typ.map(Mrc::new)))
.map(|(val, typ)| Expr(val, to_mrc_slice(typ)))
}).labelled("Expression")
}

1
src/parse/import.rs
View File

@@ -10,6 +10,7 @@ use super::lexer::Lexeme;
#[derive(Debug, Clone)]
pub struct Import {
pub path: Mrc<[String]>,
/// If name is None, this is a wildcard import
pub name: Option<String>
}

1
src/parse/name.rs
View File

@@ -19,6 +19,7 @@ fn op_parser<'a, T: AsRef<str> + Clone>(ops: &[T]) -> BoxedParser<'a, char, Stri
/// - `"` and `'` are read as primitives and would never match.
/// - `(` and `)` are strictly balanced and this must remain the case for automation and streaming.
/// - `.` is the discriminator for parametrics.
/// - ',' is always a standalone single operator, so it can never be part of a name
///
/// FIXME: `@name` without a dot should be parsed correctly for overrides. Could be an operator but
/// then parametrics should take precedence, which might break stuff. investigate.

2
src/parse/parse.rs
View File

@@ -4,7 +4,7 @@ use chumsky::{prelude::{Simple, end}, Stream, Parser};
use itertools::Itertools;
use thiserror::Error;
use crate::{expression::Rule, parse::lexer::LexedText};
use crate::{ast::Rule, parse::lexer::LexedText};
use super::{Lexeme, FileEntry, lexer, line_parser, LexerEntry};

6
src/parse/sourcefile.rs
View File

@@ -1,7 +1,7 @@
use std::collections::HashSet;
use crate::{enum_parser, box_chain};
use crate::expression::{Expr, Clause, Rule};
use crate::ast::{Expr, Clause, Rule};
use crate::utils::to_mrc_slice;
use crate::utils::Stackframe;
use crate::utils::iter::box_empty;
@@ -74,8 +74,8 @@ fn visit_all_names_expr_recur<'a, F>(
) where F: FnMut(&'a [String]) {
let Expr(val, typ) = expr;
visit_all_names_clause_recur(val, binds.clone(), cb);
if let Some(t) = typ {
visit_all_names_expr_recur(t, binds, cb)
for typ in typ.as_ref() {
visit_all_names_clause_recur(typ, binds.clone(), cb);
}
}

4
src/project/mod.rs
View File

@@ -5,6 +5,6 @@ mod name_resolver;
mod loaded;
pub use loaded::Loaded;
mod module_error;
mod file_loader;
pub mod file_loader;
pub use file_loader::file_loader;
use crate::expression::Rule;
use crate::ast::Rule;

10
src/project/name_resolver.rs
View File

@@ -4,7 +4,7 @@ use thiserror::Error;
use crate::utils::{Stackframe, to_mrc_slice};
use crate::expression::{Expr, Clause};
use crate::ast::{Expr, Clause};
type ImportMap = HashMap<String, Mrc<[String]>>;
@@ -50,14 +50,16 @@ where
symbol: Mrc<[String]>,
import_path: Stackframe<Mrc<[String]>>
) -> Result<Mrc<[String]>, ResolutionError<E>> {
if let Some(cached) = self.cache.get(&symbol) { return cached.as_ref().map_err(|e| e.clone()).map(Mrc::clone) }
if let Some(cached) = self.cache.get(&symbol) {
return cached.as_ref().map_err(|e| e.clone()).map(Mrc::clone)
}
// The imports and path of the referenced file and the local name
let path = (self.get_modname)(Mrc::clone(&symbol)).ok_or_else(|| {
ResolutionError::NoModule(Mrc::clone(&symbol))
})?;
let name = &symbol[path.len()..];
if name.is_empty() {
panic!("Something's really broken\n{:?}", import_path)
panic!("get_modname matched all to module and nothing to name in {:?}", import_path)
}
let imports = (self.get_imports)(Mrc::clone(&path))?;
let result = if let Some(source) = imports.get(&name[0]) {
@@ -110,7 +112,7 @@ where
fn process_expression_rec(&mut self, Expr(token, typ): &Expr) -> Result<Expr, ResolutionError<E>> {
Ok(Expr(
self.process_clause_rec(token)?,
self.process_exprmrcopt_rec(typ)?
typ.iter().map(|t| self.process_clause_rec(t)).collect::<Result<_, _>>()?
))
}

4
src/project/prefix.rs
View File

@@ -1,6 +1,6 @@
use mappable_rc::Mrc;
use crate::{expression::{Expr, Clause}, utils::collect_to_mrc};
use crate::{ast::{Expr, Clause}, utils::{collect_to_mrc, to_mrc_slice}};
/// Replaces the first element of a name with the matching prefix from a prefix map
@@ -36,6 +36,6 @@ fn prefix_clause(
pub fn prefix_expr(Expr(clause, typ): &Expr, namespace: Mrc<[String]>) -> Expr {
Expr(
prefix_clause(clause, Mrc::clone(&namespace)),
typ.as_ref().map(|e| Mrc::new(prefix_expr(e, namespace)))
to_mrc_slice(typ.iter().map(|e| prefix_clause(e, Mrc::clone(&namespace))).collect())
)
}

5
src/project/rule_collector.rs
View File

@@ -5,7 +5,7 @@ use std::rc::Rc;
use mappable_rc::Mrc;
use crate::expression::Rule;
use crate::ast::Rule;
use crate::parse::{self, FileEntry};
use crate::utils::{Cache, mrc_derive, to_mrc_slice};
@@ -40,6 +40,7 @@ where
(load_mod_rc.borrow_mut())(path).map_err(ModuleError::Load)
}));
// Map names to the longest prefix that points to a valid module
// At least one segment must be in the prefix, and the prefix must not be the whole name
let modname = Rc::new(Cache::new({
let loaded = Rc::clone(&loaded);
move |symbol: Mrc<[String]>, _| -> Result<Mrc<[String]>, Vec<ModuleError<ELoad>>> {
@@ -50,7 +51,7 @@ where
else { Ok(()) }
};
loop {
let path = mrc_derive(&symbol, |s| &s[..s.len() - errv.len()]);
let path = mrc_derive(&symbol, |s| &s[..s.len() - errv.len() - 1]);
match loaded.try_find(&path) {
Ok(imports) => match imports.as_ref() {
Loaded::Module(_) => break Ok(path),

47
src/expression.rs → src/representations/ast.rs
View File

@@ -1,35 +1,19 @@
use mappable_rc::Mrc;
use itertools::Itertools;
use ordered_float::NotNan;
use std::hash::Hash;
use std::fmt::Debug;
use crate::executor::{ExternFn, Atom};
/// An exact value
#[derive(Clone, PartialEq, Eq, Hash)]
pub enum Literal {
Num(NotNan<f64>),
Int(u64),
Char(char),
Str(String),
}
impl Debug for Literal {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
match self {
Self::Num(arg0) => write!(f, "{:?}", arg0),
Self::Int(arg0) => write!(f, "{:?}", arg0),
Self::Char(arg0) => write!(f, "{:?}", arg0),
Self::Str(arg0) => write!(f, "{:?}", arg0),
}
}
}
use super::Literal;
/// An S-expression with a type
#[derive(PartialEq, Eq, Hash)]
pub struct Expr(pub Clause, pub Option<Mrc<Expr>>);
pub struct Expr(pub Clause, pub Mrc<[Clause]>);
impl Clone for Expr {
fn clone(&self) -> Self {
Self(self.0.clone(), self.1.as_ref().map(Mrc::clone))
Self(self.0.clone(), Mrc::clone(&self.1))
}
}
@@ -37,8 +21,10 @@ impl Debug for Expr {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
let Expr(val, typ) = self;
write!(f, "{:?}", val)?;
if let Some(typ) = typ { write!(f, "{:?}", typ) }
else { Ok(()) }
for typ in typ.as_ref() {
write!(f, ":{:?}", typ)?
}
Ok(())
}
}
@@ -53,13 +39,14 @@ pub enum Clause {
S(char, Mrc<[Expr]>),
Lambda(String, Mrc<[Expr]>, Mrc<[Expr]>),
Auto(Option<String>, Mrc<[Expr]>, Mrc<[Expr]>),
/// Second parameter:
ExternFn(ExternFn),
Atom(Atom),
Placeh{
key: String,
/// None => matches one token
/// Some((prio, nonzero)) =>
/// prio is the sizing priority for the vectorial (higher prio grows first)
/// nonzero is whether the vectorial matches 1..n or 0..n tokens
Placeh{
key: String,
vec: Option<(usize, bool)>
},
}
@@ -94,7 +81,9 @@ impl Clone for Clause {
n.clone(), Mrc::clone(t), Mrc::clone(b)
),
Clause::Placeh{key, vec} => Clause::Placeh{key: key.clone(), vec: *vec},
Clause::Literal(l) => Clause::Literal(l.clone())
Clause::Literal(l) => Clause::Literal(l.clone()),
Clause::ExternFn(nc) => Clause::ExternFn(nc.clone()),
Clause::Atom(a) => Clause::Atom(a.clone())
}
}
}
@@ -136,7 +125,9 @@ impl Debug for Clause {
},
Self::Placeh{key, vec: None} => write!(f, "${key}"),
Self::Placeh{key, vec: Some((prio, true))} => write!(f, "...${key}:{prio}"),
Self::Placeh{key, vec: Some((prio, false))} => write!(f, "..${key}:{prio}")
Self::Placeh{key, vec: Some((prio, false))} => write!(f, "..${key}:{prio}"),
Self::ExternFn(nc) => write!(f, "{nc:?}"),
Self::Atom(a) => write!(f, "{a:?}")
}
}
}

120
src/representations/ast_to_typed.rs Normal file
View File

@@ -0,0 +1,120 @@
use mappable_rc::Mrc;
use crate::utils::{Stackframe, to_mrc_slice};
use super::{ast, typed};
#[derive(Clone)]
pub enum Error {
/// `()` as a clause is meaningless in lambda calculus
EmptyS,
/// Only `(...)` may be converted to typed lambdas. `[...]` and `{...}` left in the code are
/// signs of incomplete macro execution
BadGroup(char),
/// `foo:bar:baz` will be parsed as `(foo:bar):baz`, explicitly specifying `foo:(bar:baz)`
/// is forbidden and it's also meaningless since `baz` can only ever be the kind of types
ExplicitBottomKind,
/// Name never bound in an enclosing scope - indicates incomplete macro substitution
Unbound(String),
/// Namespaced names can never occur in the code, these are signs of incomplete macro execution
Symbol,
/// Placeholders shouldn't even occur in the code during macro execution. Something is clearly
/// terribly wrong
Placeholder,
/// It's possible to try and transform the clause `(foo:bar)` into a typed clause,
/// however the correct value of this ast clause is a typed expression (included in the error)
///
/// [expr] handles this case, so it's only really possible to get this
/// error if you're calling [clause] directly
ExprToClause(typed::Expr)
}
/// Try to convert an expression from AST format to typed lambda
pub fn expr(expr: &ast::Expr) -> Result<typed::Expr, Error> {
expr_rec(expr, Stackframe::new(None))
}
/// Try and convert a single clause from AST format to typed lambda
pub fn clause(clause: &ast::Clause) -> Result<typed::Clause, Error> {
clause_rec(clause, Stackframe::new(None))
}
/// Try and convert a sequence of expressions from AST format to typed lambda
pub fn exprv(exprv: &[ast::Expr]) -> Result<typed::Expr, Error> {
exprv_rec(exprv, Stackframe::new(None))
}
/// Recursive state of [exprv]
fn exprv_rec(v: &[ast::Expr], names: Stackframe<Option<&str>>) -> Result<typed::Expr, Error> {
if v.len() == 0 {return Err(Error::EmptyS)}
if v.len() == 1 {return expr_rec(&v[0], names)}
let (head, tail) = v.split_at(2);
let f = expr_rec(&head[0], names)?;
let x = expr_rec(&head[1], names)?;
// TODO this could probably be normalized, it's a third copy.
tail.iter().map(|e| expr_rec(e, names)).fold(
Ok(typed::Clause::Apply(Mrc::new(f), Mrc::new(x))),
|acc, e| Ok(typed::Clause::Apply(
Mrc::new(typed::Expr(acc?, to_mrc_slice(vec![]))),
Mrc::new(e?)
))
).map(|cls| typed::Expr(cls, to_mrc_slice(vec![])))
}
/// Recursive state of [expr]
fn expr_rec(ast::Expr(val, typ): &ast::Expr, names: Stackframe<Option<&str>>)
-> Result<typed::Expr, Error> {
let typ: Vec<typed::Clause> = typ.iter()
.map(|c| clause_rec(c, names))
.collect::<Result<_, _>>()?;
if let ast::Clause::S(paren, body) = val {
if *paren != '(' {return Err(Error::BadGroup(*paren))}
let typed::Expr(inner, inner_t) = exprv_rec(body.as_ref(), names)?;
let new_t = if typ.len() == 0 { inner_t } else {
to_mrc_slice(if inner_t.len() == 0 { typ } else {
inner_t.iter().chain(typ.iter()).cloned().collect()
})
};
Ok(typed::Expr(inner, new_t))
} else {
Ok(typed::Expr(clause_rec(&val, names)?, to_mrc_slice(typ)))
}
}
/// Recursive state of [clause]
fn clause_rec(cls: &ast::Clause, names: Stackframe<Option<&str>>)
-> Result<typed::Clause, Error> {
match cls {
ast::Clause::ExternFn(e) => Ok(typed::Clause::ExternFn(e.clone())),
ast::Clause::Atom(a) => Ok(typed::Clause::Atom(a.clone())),
ast::Clause::Auto(no, t, b) => Ok(typed::Clause::Auto(
if t.len() == 0 {None} else {
let typed::Expr(c, t) = exprv_rec(t.as_ref(), names)?;
if t.len() > 0 {return Err(Error::ExplicitBottomKind)}
else {Some(Mrc::new(c))}
},
Mrc::new(exprv_rec(b.as_ref(), names.push(no.as_ref().map(|n| &**n)))?)
)),
ast::Clause::Lambda(n, t, b) => Ok(typed::Clause::Lambda(
if t.len() == 0 {None} else {
let typed::Expr(c, t) = exprv_rec(t.as_ref(), names)?;
if t.len() > 0 {return Err(Error::ExplicitBottomKind)}
else {Some(Mrc::new(c))}
},
Mrc::new(exprv_rec(b.as_ref(), names.push(Some(&**n)))?)
)),
ast::Clause::Literal(l) => Ok(typed::Clause::Literal(l.clone())),
ast::Clause::Name { local: Some(arg), .. } => Ok(typed::Clause::Argument(
names.iter().position(|no| no == &Some(&**arg))
.ok_or_else(|| Error::Unbound(arg.clone()))?
)),
ast::Clause::S(paren, entries) => {
if *paren != '(' {return Err(Error::BadGroup(*paren))}
let typed::Expr(val, typ) = exprv_rec(entries.as_ref(), names)?;
if typ.len() == 0 {Ok(val)}
else {Err(Error::ExprToClause(typed::Expr(val, typ)))}
},
ast::Clause::Name { local: None, .. } => Err(Error::Symbol),
ast::Clause::Placeh { .. } => Err(Error::Placeholder)
}
}

22
src/representations/literal.rs Normal file
View File

@@ -0,0 +1,22 @@
use ordered_float::NotNan;
use std::fmt::Debug;
/// An exact value, read from the AST and unmodified in shape until compilation
#[derive(Clone, PartialEq, Eq, Hash)]
pub enum Literal {
Num(NotNan<f64>),
Int(u64),
Char(char),
Str(String),
}
impl Debug for Literal {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
match self {
Self::Num(arg0) => write!(f, "{:?}", arg0),
Self::Int(arg0) => write!(f, "{:?}", arg0),
Self::Char(arg0) => write!(f, "{:?}", arg0),
Self::Str(arg0) => write!(f, "{:?}", arg0),
}
}
}

5
src/representations/mod.rs Normal file
View File

@@ -0,0 +1,5 @@
pub mod ast;
pub mod typed;
pub mod literal;
pub mod ast_to_typed;
pub use literal::Literal;

147
src/representations/typed.rs Normal file
View File

@@ -0,0 +1,147 @@
use mappable_rc::Mrc;
use crate::executor::Atom;
use crate::utils::{to_mrc_slice, one_mrc_slice};
use crate::{executor::ExternFn, utils::string_from_charset};
use super::{Literal, ast_to_typed};
use super::ast;
use std::fmt::{Debug, Write};
/// Indicates whether either side needs to be wrapped. Syntax whose end is ambiguous on that side
/// must use parentheses, or forward the flag
#[derive(PartialEq, Eq)]
struct Wrap(bool, bool);
#[derive(PartialEq, Eq, Hash)]
pub struct Expr(pub Clause, pub Mrc<[Clause]>);
impl Expr {
fn deep_fmt(&self, f: &mut std::fmt::Formatter<'_>, depth: usize, tr: Wrap) -> std::fmt::Result {
let Expr(val, typ) = self;
if typ.len() > 0 {
val.deep_fmt(f, depth, Wrap(true, true))?;
for typ in typ.as_ref() {
f.write_char(':')?;
typ.deep_fmt(f, depth, Wrap(true, true))?;
}
} else {
val.deep_fmt(f, depth, tr)?;
}
Ok(())
}
}
impl Clone for Expr {
fn clone(&self) -> Self {
Self(self.0.clone(), Mrc::clone(&self.1))
}
}
impl Debug for Expr {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
self.deep_fmt(f, 0, Wrap(false, false))
}
}
#[derive(PartialEq, Eq, Hash)]
pub enum Clause {
Literal(Literal),
Apply(Mrc<Expr>, Mrc<Expr>),
/// Explicit specification of an Auto value
Explicit(Mrc<Expr>, Mrc<Expr>),
Lambda(Option<Mrc<Clause>>, Mrc<Expr>),
Auto(Option<Mrc<Clause>>, Mrc<Expr>),
Argument(usize),
ExternFn(ExternFn),
Atom(Atom)
}
const ARGNAME_CHARSET: &str = "abcdefghijklmnopqrstuvwxyz";
fn parametric_fmt(
f: &mut std::fmt::Formatter<'_>,
prefix: &str, argtyp: Option<Mrc<Clause>>, body: Mrc<Expr>, depth: usize, wrap_right: bool
) -> std::fmt::Result {
if wrap_right { f.write_char('(')?; }
f.write_str(prefix)?;
f.write_str(&string_from_charset(depth, ARGNAME_CHARSET))?;
if let Some(typ) = argtyp {
f.write_str(":")?;
typ.deep_fmt(f, depth, Wrap(false, false))?;
}
f.write_str(".")?;
body.deep_fmt(f, depth + 1, Wrap(false, false))?;
if wrap_right { f.write_char(')')?; }
Ok(())
}
impl Clause {
fn deep_fmt(&self, f: &mut std::fmt::Formatter<'_>, depth: usize, Wrap(wl, wr): Wrap)
-> std::fmt::Result {
match self {
Self::Literal(arg0) => write!(f, "{arg0:?}"),
Self::ExternFn(nc) => write!(f, "{nc:?}"),
Self::Atom(a) => write!(f, "{a:?}"),
Self::Lambda(argtyp, body) => parametric_fmt(f,
"\\", argtyp.as_ref().map(Mrc::clone), Mrc::clone(body), depth, wr
),
Self::Auto(argtyp, body) => parametric_fmt(f,
"@", argtyp.as_ref().map(Mrc::clone), Mrc::clone(body), depth, wr
),
Self::Argument(up) => f.write_str(&string_from_charset(depth - up - 1, ARGNAME_CHARSET)),
Self::Explicit(expr, param) => {
if wl { f.write_char('(')?; }
expr.deep_fmt(f, depth, Wrap(false, true))?;
f.write_str(" @")?;
param.deep_fmt(f, depth, Wrap(true, wr && !wl))?;
if wl { f.write_char(')')?; }
Ok(())
}
Self::Apply(func, x) => {
if wl { f.write_char('(')?; }
func.deep_fmt(f, depth, Wrap(false, true) )?;
f.write_char(' ')?;
x.deep_fmt(f, depth, Wrap(true, wr && !wl) )?;
if wl { f.write_char(')')?; }
Ok(())
}
}
}
pub fn wrap(self) -> Mrc<Expr> { Mrc::new(Expr(self, to_mrc_slice(vec![]))) }
pub fn wrap_t(self, t: Clause) -> Mrc<Expr> { Mrc::new(Expr(self, one_mrc_slice(t))) }
}
impl Clone for Clause {
fn clone(&self) -> Self {
match self {
Clause::Auto(t, b) => Clause::Auto(t.as_ref().map(Mrc::clone), Mrc::clone(b)),
Clause::Lambda(t, b) => Clause::Lambda(t.as_ref().map(Mrc::clone), Mrc::clone(b)),
Clause::Literal(l) => Clause::Literal(l.clone()),
Clause::ExternFn(nc) => Clause::ExternFn(nc.clone()),
Clause::Atom(a) => Clause::Atom(a.clone()),
Clause::Apply(f, x) => Clause::Apply(Mrc::clone(f), Mrc::clone(x)),
Clause::Explicit(f, x) => Clause::Explicit(Mrc::clone(f), Mrc::clone(x)),
Clause::Argument(lvl) => Clause::Argument(*lvl)
}
}
}
impl Debug for Clause {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
self.deep_fmt(f, 0, Wrap(false, false))
}
}
impl TryFrom<&ast::Expr> for Expr {
type Error = ast_to_typed::Error;
fn try_from(value: &ast::Expr) -> Result<Self, Self::Error> {
ast_to_typed::expr(value)
}
}
impl TryFrom<&ast::Clause> for Clause {
type Error = ast_to_typed::Error;
fn try_from(value: &ast::Clause) -> Result<Self, Self::Error> {
ast_to_typed::clause(value)
}
}

18
src/rule/executor/execute.rs
View File

@@ -1,7 +1,7 @@
use hashbrown::HashMap;
use mappable_rc::Mrc;
use crate::{expression::{Expr, Clause}, utils::{iter::{box_once, into_boxed_iter}, to_mrc_slice, one_mrc_slice}};
use crate::{ast::{Expr, Clause}, utils::{iter::{box_once, into_boxed_iter}, to_mrc_slice, one_mrc_slice}, unwrap_or};
use super::{super::RuleError, state::{State, Entry}, slice_matcher::SliceMatcherDnC};
fn verify_scalar_vec(pattern: &Expr, is_vec: &mut HashMap<String, bool>)
@@ -36,18 +36,18 @@ fn verify_scalar_vec(pattern: &Expr, is_vec: &mut HashMap<String, bool>)
};
Ok(())
};
let Expr(val, typ_opt) = pattern;
let Expr(val, typ) = pattern;
verify_clause(val, is_vec)?;
if let Some(typ) = typ_opt {
verify_scalar_vec(typ, is_vec)?;
for typ in typ.as_ref() {
verify_clause(typ, is_vec)?;
}
Ok(())
}
fn slice_to_vec(src: &mut Mrc<[Expr]>, tgt: &mut Mrc<[Expr]>) {
let prefix_expr = Expr(Clause::Placeh{key: "::prefix".to_string(), vec: Some((0, false))}, None);
let postfix_expr = Expr(Clause::Placeh{key: "::postfix".to_string(), vec: Some((0, false))}, None);
let prefix_expr = Expr(Clause::Placeh{key: "::prefix".to_string(), vec: Some((0, false))}, to_mrc_slice(vec![]));
let postfix_expr = Expr(Clause::Placeh{key: "::postfix".to_string(), vec: Some((0, false))}, to_mrc_slice(vec![]));
// Prefix or postfix to match the full vector
let head_multi = matches!(src.first().expect("Src can never be empty!").0, Clause::Placeh{vec: Some(_), ..});
let tail_multi = matches!(src.last().expect("Impossible branch!").0, Clause::Placeh{vec: Some(_), ..});
@@ -121,13 +121,13 @@ fn write_slice(state: &State, tpl: &Mrc<[Expr]>) -> Mrc<[Expr]> {
write_slice(state, body)
), xpr_typ.to_owned())),
Clause::Placeh{key, vec: None} => {
let real_key = if let Some(real_key) = key.strip_prefix('_') {real_key} else {key};
let real_key = unwrap_or!(key.strip_prefix('_'); key);
match &state[real_key] {
Entry::Scalar(x) => box_once(x.as_ref().to_owned()),
Entry::Name(n) => box_once(Expr(Clause::Name {
local: Some(n.as_ref().to_owned()),
qualified: one_mrc_slice(n.as_ref().to_owned())
}, None)),
}, to_mrc_slice(vec![]))),
_ => panic!("Scalar template may only be derived from scalar placeholder"),
}
},
@@ -135,7 +135,7 @@ fn write_slice(state: &State, tpl: &Mrc<[Expr]>) -> Mrc<[Expr]> {
into_boxed_iter(v.as_ref().to_owned())
} else {panic!("Vectorial template may only be derived from vectorial placeholder")},
// Explicit base case so that we get an error if Clause gets new values
c@Clause::Literal(_) | c@Clause::Name { .. } =>
c@Clause::Literal(_) | c@Clause::Name { .. } | c@Clause::ExternFn(_) | c@Clause::Atom(_) =>
box_once(Expr(c.to_owned(), xpr_typ.to_owned()))
}).collect()
}

14
src/rule/executor/slice_matcher.rs
View File

@@ -2,8 +2,8 @@ use std::fmt::Debug;
use mappable_rc::Mrc;
use crate::expression::{Expr, Clause};
use crate::unwrap_or_continue;
use crate::ast::{Expr, Clause};
use crate::unwrap_or;
use crate::utils::iter::box_empty;
use crate::utils::{Side, Cache, mrc_derive, mrc_try_derive, to_mrc_slice};
@@ -92,10 +92,7 @@ impl SliceMatcherDnC {
pub fn valid_subdivisions(&self,
range: Mrc<[Expr]>
) -> impl Iterator<Item = (Mrc<[Expr]>, Mrc<[Expr]>, Mrc<[Expr]>)> {
let own_max = {
if let Some(x) = self.own_max_size(range.len()) {x}
else {return box_empty()}
};
let own_max = unwrap_or!(self.own_max_size(range.len()); return box_empty());
let own_min = self.own_min_size();
let lmin = self.min(Side::Left);
let _lmax = self.max(Side::Left, range.len());
@@ -261,10 +258,11 @@ impl SliceMatcherDnC {
// Step through valid slicings based on reported size constraints in order
// from longest own section to shortest and from left to right
for (left, own, right) in self.valid_subdivisions(target) {
return Some(unwrap_or_continue!(
return Some(unwrap_or!(
self.apply_side_with_cache(Side::Left, left, cache)
.and_then(|lres| lres + self.apply_side_with_cache(Side::Right, right, cache))
.and_then(|side_res| side_res.insert_vec(name, own.as_ref()))
.and_then(|side_res| side_res.insert_vec(name, own.as_ref()));
continue
))
}
None

2
src/rule/executor/split_at_max_vec.rs
View File

@@ -1,7 +1,7 @@
use mappable_rc::Mrc;
use itertools::Itertools;
use crate::expression::{Expr, Clause};
use crate::ast::{Expr, Clause};
use crate::utils::{mrc_derive, mrc_try_derive};
pub type MaxVecSplit = (Mrc<[Expr]>, (Mrc<str>, usize, bool), Mrc<[Expr]>);

6
src/rule/executor/state.rs
View File

@@ -2,7 +2,7 @@ use std::{ops::{Add, Index}, rc::Rc, fmt::Debug};
use hashbrown::HashMap;
use crate::expression::Expr;
use crate::ast::Expr;
#[derive(Debug, PartialEq, Eq)]
pub enum Entry {
@@ -76,7 +76,9 @@ impl State {
{
if let Some(old) = self.0.get(k.as_ref()) {
if let Entry::NameOpt(val) = old {
if val.as_ref().map(|s| s.as_ref().as_str()) != v.map(|s| s.as_ref()) {return None}
if val.as_ref().map(|s| s.as_ref().as_str()) != v.map(|s| s.as_ref()) {
return None
}
} else {return None}
} else {
self.0.insert(k.to_string(), Entry::NameOpt(v.map(|s| Rc::new(s.to_string()))));

15
src/rule/repository.rs
View File

@@ -2,10 +2,11 @@ use std::fmt::Debug;
use mappable_rc::Mrc;
use crate::expression::Expr;
use crate::representations::ast::Expr;
use super::{super::expression::Rule, executor::execute, RuleError};
use super::{super::ast::Rule, executor::execute, RuleError};
/// Manages a priority queue of substitution rules and allows to apply them
pub struct Repository(Vec<Rule>);
impl Repository {
pub fn new(mut rules: Vec<Rule>) -> Self {
@@ -13,6 +14,7 @@ impl Repository {
Self(rules)
}
/// Attempt to run each rule in priority order once
pub fn step(&self, mut code: Mrc<[Expr]>) -> Result<Option<Mrc<[Expr]>>, RuleError> {
let mut ran_once = false;
for rule in self.0.iter() {
@@ -27,11 +29,16 @@ impl Repository {
Ok(if ran_once {Some(code)} else {None})
}
pub fn long_step(&self, mut code: Mrc<[Expr]>) -> Result<Mrc<[Expr]>, RuleError> {
/// Attempt to run each rule in priority order `limit` times. Returns the final
/// tree and the number of iterations left to the limit.
pub fn long_step(&self, mut code: Mrc<[Expr]>, mut limit: usize)
-> Result<(Mrc<[Expr]>, usize), RuleError> {
while let Some(tmp) = self.step(Mrc::clone(&code))? {
if 0 >= limit {break}
limit -= 1;
code = tmp
}
Ok(code)
Ok((code, limit))
}
}

52
src/types/hindley_milner.rs.proto Normal file
View File

@@ -0,0 +1,52 @@
use std::{borrow::Borrow};
use std::hash::Hash;
use hashbrown::HashMap;
use mappable_rc::Mrc;
use crate::{ast::{Expr, Clause}, utils::mrc_to_iter};
pub struct Substitution(HashMap<String, Mrc<Expr>>);
impl Substitution {
fn new() -> Self { Self(HashMap::new()) }
fn apply<Q: ?Sized + Hash + Eq>(&self, q: &Q) -> Option<Mrc<Expr>>
where String: Borrow<Q> {
self.0.get(q).map(Mrc::clone)
}
}
pub fn hindley_milner(a: Mrc<[Expr]>, b: Mrc<[Expr]>) -> Result<Substitution, ()> {
hindley_milner_rec(Substitution::new(), a, b)
}
pub fn hindley_milner_rec(mut s: Substitution, a: Mrc<[Expr]>, b: Mrc<[Expr]>)
-> Result<Substitution, ()> {
if a.len() != b.len() {return Err(())}
for (mut a, mut b) in mrc_to_iter(a).zip(mrc_to_iter(b)) {
if let Clause::Placeh{key, ..} = &a.0 {
if let Some(ex) = s.apply(key) { a = ex }
}
if let Clause::Placeh{key, ..} = &b.0 {
if let Some(ex) = s.apply(key) { b = ex }
}
if !matches!(&a.0, Clause::Placeh{..}) { (a, b) = (b, a) }
match (&a.0, &b.0) {
(Clause::Placeh{key:a_key,..}, Clause::Placeh{key:b_key,..}) =>
if a_key == b_key {return Ok(s)},
_ => return Err(())
}
if let (Clause::Placeh{key: a_key,..}, Clause::Placeh{key: b_key,..}) = (&a.0, &b.0) {
if a_key == b_key {return Ok(s)}
} else if let (Clause::S(_, a_body), Clause::S(_, b_body)) = (&a.0, &b.0) {
s = hindley_milner_rec(s, Mrc::clone(a_body), Mrc::clone(b_body))?
} else if let ()
}
Ok(s)
}
pub fn occurs(key: &str, val: &Expr) -> bool {
match val.0 {
Clause::Auto(_, _, body) => body.
}
}

13
src/types/mod.rs Normal file
View File

@@ -0,0 +1,13 @@
// mod hindley_milner;
#[derive(Clone, Hash, PartialEq, Eq)]
pub enum Expression<L, V, O, F> {
Literal(L),
Variable(V),
Operation(O, Vec<Expression<L, V, O, F>>),
Lazy(F)
}
pub struct Rule {
}

0
src/types/unifier.rs Normal file
View File

113
src/utils/bfs.rs Normal file
View File

@@ -0,0 +1,113 @@
use std::collections::{VecDeque, HashSet};
use std::iter;
use std::hash::Hash;
use crate::unwrap_or;
use crate::utils::BoxedIter;
/// Two-stage breadth-first search;
/// Instead of enumerating neighbors before returning a node, it puts visited but not yet
/// enumerated nodes in a separate queue and only enumerates them to refill the queue of children
/// one by one once it's empty. This method is preferable for generated graphs because it doesn't
/// allocate memory for the children until necessary, but it's also probably a bit slower since
/// it involves additional processing.
///
/// # Performance
/// `T` is cloned twice for each returned value.
pub fn bfs<T, F, I>(init: T, neighbors: F)
-> impl Iterator<Item = T>
where T: Eq + Hash + Clone + std::fmt::Debug,
F: Fn(T) -> I, I: Iterator<Item = T>
{
let mut visited: HashSet<T> = HashSet::new();
let mut visit_queue: VecDeque<T> = VecDeque::from([init]);
let mut unpack_queue: VecDeque<T> = VecDeque::new();
iter::from_fn(move || {
let next = {loop {
let next = unwrap_or!(visit_queue.pop_front(); break None);
if !visited.contains(&next) { break Some(next) }
}}.or_else(|| loop {
let unpacked = unwrap_or!(unpack_queue.pop_front(); break None);
let mut nbv = neighbors(unpacked).filter(|t| !visited.contains(t));
if let Some(next) = nbv.next() {
visit_queue.extend(nbv);
break Some(next)
}
})?;
visited.insert(next.clone());
unpack_queue.push_back(next.clone());
Some(next)
})
}
/// Same as [bfs] but with a recursion depth limit
///
/// The main intent is to effectively walk infinite graphs of unknown breadth without making the
/// recursion depth dependent on the number of nodes. If predictable runtime is more important
/// than predictable depth, [bfs] with [std::iter::Iterator::take] should be used instead
pub fn bfs_upto<'a, T: 'a, F: 'a, I: 'a>(init: T, neighbors: F, limit: usize)
-> impl Iterator<Item = T> + 'a
where T: Eq + Hash + Clone + std::fmt::Debug,
F: Fn(T) -> I, I: Iterator<Item = T>
{
/// Newtype to store the recursion depth but exclude it from equality comparisons
/// Because BFS visits nodes in increasing distance order, when a node is visited for the
/// second time it will never override the earlier version of itself. This is not the case
/// with Djikstra's algorithm, which can be conceptualised as a "weighted BFS".
#[derive(Eq, Clone, Debug)]
struct Wrap<U>(usize, U);
impl<U: PartialEq> PartialEq for Wrap<U> {
fn eq(&self, other: &Self) -> bool { self.1.eq(&other.1) }
}
impl<U: Hash> Hash for Wrap<U> {
fn hash<H: std::hash::Hasher>(&self, state: &mut H) { self.1.hash(state) }
}
bfs(Wrap(0, init), move |Wrap(dist, t)| -> BoxedIter<Wrap<T>> { // boxed because we branch
if dist == limit {Box::new(iter::empty())}
else {Box::new(neighbors(t).map(move |t| Wrap(dist + 1, t)))}
}).map(|Wrap(_, t)| t)
}
#[cfg(test)]
mod tests {
use itertools::Itertools;
use super::*;
type Graph = Vec<Vec<usize>>;
fn neighbors(graph: &Graph, pt: usize) -> impl Iterator<Item = usize> + '_ {
graph[pt].iter().copied()
}
fn from_neighborhood_matrix(matrix: Vec<Vec<usize>>) -> Graph {
matrix.into_iter().map(|v| {
v.into_iter().enumerate().filter_map(|(i, ent)| {
if ent > 1 {panic!("Neighborhood matrices must contain binary values")}
else if ent == 1 {Some(i)}
else {None}
}).collect()
}).collect()
}
#[test]
fn test_square() {
let simple_graph = from_neighborhood_matrix(vec![
vec![0,1,0,1,1,0,0,0],
vec![1,0,1,0,0,1,0,0],
vec![0,1,0,1,0,0,1,0],
vec![1,0,1,0,0,0,0,1],
vec![1,0,0,0,0,1,0,1],
vec![0,1,0,0,1,0,1,0],
vec![0,0,1,0,0,1,0,1],
vec![0,0,0,1,1,0,1,0],
]);
let scan = bfs(0, |n| neighbors(&simple_graph, n)).collect_vec();
assert_eq!(scan, vec![0, 1, 3, 4, 2, 5, 7, 6])
}
#[test]
fn test_stringbuilder() {
let scan = bfs("".to_string(), |s| {
vec![s.clone()+";", s.clone()+"a", s+"aaa"].into_iter()
}).take(30).collect_vec();
println!("{scan:?}")
}
}

91
src/utils/for_loop.rs Normal file
View File

@@ -0,0 +1,91 @@
/// Imitates a regular for loop with an exit clause using Rust's `loop` keyword.
/// This macro brings the break value to all existing Rust loops, by allowing you to specify
/// an exit expression in case the loop was broken by the condition and not an explicit `break`.
///
/// Since the exit expression can also be a block, this also allows you to execute other code when
/// the condition fails. This can also be used to re-enter the loop with an explicit `continue`
/// statement.
///
/// The macro also adds support for classic for loops familiar to everyone since C, except with
/// the addition of an exit statement these too can be turned into expressions.
///
/// ```
/// xloop!(for i in 0..10; {
/// connection.try_connect()
/// if connection.ready() {
/// break Some(connection)
/// }
/// }; None)
/// ```
///
/// While loop with reentry. This is a very convoluted example but displays the idea quite clearly.
///
/// ```
/// xloop!(while socket.is_open(); {
/// let (data, is_end) = socket.read();
/// all_data.append(data)
/// if is_end { break Ok(all_data) }
/// }; {
/// if let Ok(new_sock) = open_socket(socket.position()) {
/// new_sock.set_position(socket.position());
/// socket = new_sock;
/// continue
/// } else {
/// Err(DownloadError::ConnectionLost)
/// }
/// })
/// ```
///
/// CUDA algorythm for O(log n) summation using a C loop
///
/// ```
/// xloop!(let mut leap = 1; own_id*2 + leap < batch_size; leap *= 2; {
/// batch[own_id*2] += batch[own_id*2 + leap]
/// })
/// ```
///
/// The above loop isn't used as an expression, but an exit expression - or block - can be added
/// to these as well just like the others. In all cases the exit expression is optional, its
/// default value is `()`.
///
/// **todo** find a valid use case for While let for a demo
#[macro_export]
macro_rules! xloop {
(for $p:pat in $it:expr; $body:stmt) => {
xloop!(for $p in $it; $body; ())
};
(for $p:pat in $it:expr; $body:stmt; $exit:stmt) => {
{
let mut __xloop__ = $it.into_iter();
xloop!(let Some($p) = __xloop__.next(); $body; $exit)
}
};
(let $p:pat = $e:expr; $body:stmt) => {
xloop!(let $p = $e; $body; ())
};
(let $p:pat = $e:expr; $body:stmt; $exit:stmt) => {
{
loop {
if let $p = $e { $body }
else { break { $exit } }
}
}
};
(while $cond:expr; $body:stmt) => {
xloop!($cond; $body; ())
};
(while $cond:expr; $body:stmt; $exit:stmt) => {
{
loop {
if $cond { break { $exit } }
else { $body }
}
}
};
($init:stmt; $cond:expr; $step:stmt; $body:stmt) => {
xloop!(for ( $init; $cond; $step ) $body; ())
};
($init:stmt; $cond:expr; $step:stmt; $body:stmt; $exit:stmt) => {
{ $init; xloop!(while !($cond); { $body; $step }; $exit) }
};
}

36
src/utils/mod.rs
View File

@@ -2,8 +2,13 @@ mod cache;
mod substack;
mod side;
mod merge_sorted;
mod unwrap_or_continue;
mod unwrap_or;
pub mod iter;
mod bfs;
mod unless_let;
mod string_from_charset;
mod for_loop;
mod protomap;
pub use cache::Cache;
use mappable_rc::Mrc;
@@ -11,6 +16,7 @@ pub use substack::Stackframe;
pub use side::Side;
pub use merge_sorted::merge_sorted;
pub use iter::BoxedIter;
pub use string_from_charset::string_from_charset;
pub fn mrc_derive<T: ?Sized, P, U: ?Sized>(m: &Mrc<T>, p: P) -> Mrc<U>
where P: for<'a> FnOnce(&'a T) -> &'a U {
@@ -37,3 +43,31 @@ pub fn mrc_derive_slice<T>(mv: &Mrc<Vec<T>>) -> Mrc<[T]> {
pub fn one_mrc_slice<T>(t: T) -> Mrc<[T]> {
Mrc::map(Mrc::new([t; 1]), |v| v.as_slice())
}
pub fn mrc_to_iter<T>(ms: Mrc<[T]>) -> impl Iterator<Item = Mrc<T>> {
let mut i = 0;
std::iter::from_fn(move || if i < ms.len() {
let out = Some(mrc_derive(&ms, |s| &s[i]));
i += 1;
out
} else {None})
}
pub fn mrc_unnest<T>(m: &Mrc<Mrc<T>>) -> Mrc<T> {
Mrc::clone(m.as_ref())
}
pub fn mrc_slice_to_only<T>(m: Mrc<[T]>) -> Result<Mrc<T>, ()> {
Mrc::try_map(m, |slice| {
if slice.len() != 1 {None}
else {Some(&slice[0])}
}).map_err(|_| ())
}
pub fn mrc_slice_to_only_option<T>(m: Mrc<[T]>) -> Result<Option<Mrc<T>>, ()> {
if m.len() > 1 {return Err(())}
Ok(Mrc::try_map(m, |slice| {
if slice.len() == 0 {None}
else {Some(&slice[0])}
}).ok())
}

152
src/utils/protomap.rs Normal file
View File

@@ -0,0 +1,152 @@
use std::{iter, ops::{Index, Add}, borrow::Borrow};
use smallvec::SmallVec;
const INLINE_ENTRIES: usize = 2;
/// Linked-array-list of key-value pairs.
/// Lookup and modification is O(n + cachemiss * n / m)
/// Can be extended by reference in O(m) < O(n)
pub struct ProtoMap<'a, K, V> {
entries: SmallVec<[(K, Option<V>); INLINE_ENTRIES]>,
prototype: Option<&'a ProtoMap<'a, K, V>>
}
impl<'a, K, V> ProtoMap<'a, K, V> {
pub fn new() -> Self {
Self {
entries: SmallVec::new(),
prototype: None
}
}
/// Mutable reference to entry without checking proto in O(m)
fn local_entry_mut<'b, Q: ?Sized>(&'b mut self, query: &Q)
-> Option<(usize, &'b mut K, &'b mut Option<V>)>
where K: Borrow<Q>, Q: Eq
{
self.entries.iter_mut().enumerate().find_map(|(i, (k, v))| {
if query.eq((*k).borrow()) { Some((i, k, v)) } else { None }
})
}
/// Entry without checking proto in O(m)
fn local_entry<'b, Q: ?Sized>(&'b self, query: &Q)
-> Option<(usize, &'b K, &'b Option<V>)>
where K: Borrow<Q>, Q: Eq
{
self.entries.iter().enumerate().find_map(|(i, (k, v))| {
if query.eq((*k).borrow()) { Some((i, k, v)) } else { None }
})
}
/// Find entry in prototype chain in O(n)
pub fn get<'b, Q: ?Sized>(&'b self, query: &Q) -> Option<&'b V>
where K: Borrow<Q>, Q: Eq
{
if let Some((_, _, v)) = self.local_entry(query) {
v.as_ref()
} else {
self.prototype?.get(query)
}
}
/// Record a value for the given key in O(m)
pub fn set(&mut self, key: &K, value: V) where K: Eq + Clone {
if let Some((_, _, v)) = self.local_entry_mut(key) {
*v = Some(value);
} else {
self.entries.push((key.clone(), Some(value)))
}
}
/// Delete in a memory-efficient way in O(n)
pub fn delete_small(&mut self, key: &K) where K: Eq + Clone {
let exists_up = self.prototype.and_then(|p| p.get(key)).is_some();
let local_entry = self.local_entry_mut(key);
match (exists_up, local_entry) {
(false, None) => (), // nothing to do
(false, Some((i, _, _))) => { self.entries.remove(i); }, // forget locally
(true, Some((_, _, v))) => *v = None, // update local override to cover
(true, None) => self.entries.push((key.clone(), None)), // create new
}
}
/// Delete in O(m) without checking the prototype chain
/// May produce unnecessary cover over previously unknown key
pub fn delete_fast(&mut self, key: &K) where K: Eq + Clone {
if let Some((_, _, v)) = self.local_entry_mut(key) {
*v = None
} else {
self.entries.push((key.clone(), None))
}
}
/// Iterate over the values defined herein and on the prototype chain
/// Note that this will visit keys multiple times
pub fn iter(&self) -> impl Iterator<Item = &(K, Option<V>)> {
let mut map = self;
iter::from_fn(move || {
let pairs = map.entries.iter();
map = map.prototype?;
Some(pairs)
}).flatten()
}
/// Visit the keys in an unsafe random order, repeated arbitrarily many times
pub fn keys(&self) -> impl Iterator<Item = &K> {
self.iter().map(|(k, _)| k)
}
/// Visit the values in random order
pub fn values(&self) -> impl Iterator<Item = &V> {
self.iter().filter_map(|(_, v)| v.as_ref())
}
/// Update the prototype, and correspondingly the lifetime of the map
pub fn set_proto<'b>(self, proto: &'b ProtoMap<'b, K, V>) -> ProtoMap<'b, K, V> {
ProtoMap {
entries: self.entries,
prototype: Some(proto)
}
}
}
impl<T, K, V> From<T> for ProtoMap<'_, K, V> where T: IntoIterator<Item = (K, V)> {
fn from(value: T) -> Self {
Self {
entries: value.into_iter().map(|(k, v)| (k, Some(v))).collect(),
prototype: None
}
}
}
impl<Q: ?Sized, K, V> Index<&Q> for ProtoMap<'_, K, V> where K: Borrow<Q>, Q: Eq {
type Output = V;
fn index(&self, index: &Q) -> &Self::Output {
self.get(index).expect("Index not found in map")
}
}
impl<K: Clone, V: Clone> Clone for ProtoMap<'_, K, V> {
fn clone(&self) -> Self {
Self {
entries: self.entries.clone(),
prototype: self.prototype
}
}
}
impl<'a, K: 'a, V: 'a> Add<(K, V)> for &'a ProtoMap<'a, K, V> {
type Output = ProtoMap<'a, K, V>;
fn add(self, rhs: (K, V)) -> Self::Output {
ProtoMap::from([rhs]).set_proto(self)
}
}
#[macro_export]
macro_rules! protomap {
($($ent:expr),*) => {
ProtoMap::from([$($ent:expr),*])
};
}

14
src/utils/string_from_charset.rs Normal file
View File

@@ -0,0 +1,14 @@
fn string_from_charset_rec(val: usize, digits: &str) -> String {
let radix = digits.len();
let mut prefix = if val > radix {
string_from_charset_rec(val / radix, digits)
} else {String::new()};
prefix.push(digits.chars().nth(val - 1).unwrap_or_else(|| {
panic!("Overindexed digit set \"{}\" with {}", digits, val - 1)
}));
prefix
}
pub fn string_from_charset(val: usize, digits: &str) -> String {
string_from_charset_rec(val + 1, digits)
}

6
src/utils/unless_let.rs Normal file
View File

@@ -0,0 +1,6 @@
#[macro_export]
macro_rules! unless_let {
($m:pat_param = $expr:tt) => {
if let $m = $expr {} else
}
}

6
src/utils/unwrap_or.rs Normal file
View File

@@ -0,0 +1,6 @@
#[macro_export]
macro_rules! unwrap_or {
($m:expr; $fail:expr) => {
{ if let Some(res) = ($m) {res} else {$fail} }
}
}

6
src/utils/unwrap_or_continue.rs
View File

@@ -1,6 +0,0 @@
#[macro_export]
macro_rules! unwrap_or_continue {
($m:expr) => {
{ if let Some(res) = ($m) {res} else {continue} }
}
}