schala/schala-lang/language/src/schala.rs

377 lines
13 KiB
Rust
Raw Normal View History

use stopwatch::Stopwatch;
use std::cell::RefCell;
use std::rc::Rc;
use std::collections::HashSet;
use itertools::Itertools;
use schala_repl::{ProgrammingLanguageInterface,
ComputationRequest, ComputationResponse,
LangMetaRequest, LangMetaResponse, GlobalOutputStats,
DebugResponse, DebugAsk};
2019-10-23 02:51:34 -07:00
use crate::{ast, reduced_ast, tokenizing, parsing, eval, typechecking, symbol_table, source_map};
2019-10-21 04:09:43 -07:00
pub type SymbolTableHandle = Rc<RefCell<symbol_table::SymbolTable>>;
2019-10-23 02:51:34 -07:00
pub type SourceMapHandle = Rc<RefCell<source_map::SourceMap>>;
2019-10-21 04:09:43 -07:00
/// All the state necessary to parse and execute a Schala program are stored in this struct.
/// `state` represents the execution state for the AST-walking interpreter, the other fields
/// should be self-explanatory.
2021-10-13 00:53:32 -07:00
#[allow(dead_code)]
pub struct Schala {
source_reference: SourceReference,
2019-10-23 02:51:34 -07:00
source_map: SourceMapHandle,
state: eval::State<'static>,
2019-10-23 02:51:34 -07:00
symbol_table: SymbolTableHandle,
resolver: crate::scope_resolution::ScopeResolver<'static>,
type_context: typechecking::TypeContext<'static>,
active_parser: parsing::Parser,
}
impl Schala {
2021-10-13 23:45:54 -07:00
//TODO implement documentation for language items
/*
fn handle_docs(&self, source: String) -> LangMetaResponse {
LangMetaResponse::Docs {
doc_string: format!("Schala item `{}` : <<Schala-lang documentation not yet implemented>>", source)
}
}
2021-10-13 23:45:54 -07:00
*/
}
impl Schala {
/// Creates a new Schala environment *without* any prelude.
fn new_blank_env() -> Schala {
2019-10-23 02:51:34 -07:00
let source_map = Rc::new(RefCell::new(source_map::SourceMap::new()));
let symbols = Rc::new(RefCell::new(symbol_table::SymbolTable::new(source_map.clone())));
Schala {
2019-10-23 02:51:34 -07:00
//TODO maybe these can be the same structure
source_reference: SourceReference::new(),
symbol_table: symbols.clone(),
2019-10-23 02:51:34 -07:00
source_map: source_map.clone(),
resolver: crate::scope_resolution::ScopeResolver::new(symbols.clone()),
2019-11-09 19:52:05 -08:00
state: eval::State::new(),
type_context: typechecking::TypeContext::new(),
active_parser: parsing::Parser::new(source_map)
}
}
/// Creates a new Schala environment with the standard prelude, which is defined as ordinary
/// Schala code in the file `prelude.schala`
pub fn new() -> Schala {
let prelude = include_str!("prelude.schala");
let mut s = Schala::new_blank_env();
2021-10-13 23:45:54 -07:00
//TODO this shouldn't depend on schala-repl types
let request = ComputationRequest { source: prelude, debug_requests: HashSet::default() };
let response = s.run_computation(request);
if let Err(msg) = response.main_output {
panic!("Error in prelude, panicking: {}", msg);
}
s
}
2021-10-13 23:45:54 -07:00
/// This is where the actual action of interpreting/compilation happens.
/// Note: this should eventually use a query-based system for parallelization, cf.
/// https://rustc-dev-guide.rust-lang.org/overview.html
fn run_pipeline(&mut self, source: &str) -> Result<String, String> {
//TODO `run_pipeline` should return a formatted error struct
//TODO every stage should have a common error-format that gets turned into a repl-appropriate
//string in `run_computation`
// 1st stage - tokenization
// TODO tokenize should return its own error type
let tokens = tokenizing::tokenize(source);
let token_errors: Vec<String> = tokens.iter().filter_map(|t| t.get_error()).collect();
if token_errors.len() > 0 {
return Err(format!("{:?}", token_errors));
}
2021-10-13 23:45:54 -07:00
//2nd stage - parsing
self.active_parser.add_new_tokens(tokens);
let mut ast = self.active_parser.parse()
.map_err(|err| format_parse_error(err, &self.source_reference))?;
2021-10-13 23:45:54 -07:00
// Symbol table
self.symbol_table.borrow_mut().add_top_level_symbols(&ast)?;
// Scope resolution - requires mutating AST
self.resolver.resolve(&mut ast)?;
// Typechecking
let overall_type = self.type_context.typecheck(&ast)
.map_err(|err| format!("Type error: {}", err.msg))?;
// Reduce AST - this doesn't produce an error yet, but probably should
let symbol_table = self.symbol_table.borrow();
let reduced_ast = reduced_ast::reduce(&ast, &symbol_table);
// Tree-walking evaluator. TODO fix this
let evaluation_outputs = self.state.evaluate(reduced_ast, true);
let text_output: Result<Vec<String>, String> = evaluation_outputs
.into_iter()
.collect();
let eval_output: String = text_output
.map(|v| { Iterator::intersperse(v.into_iter(), "\n".to_owned()).collect() })?;
Ok(eval_output)
}
}
fn parsing(input: Vec<tokenizing::Token>, handle: &mut Schala, comp: Option<&mut PassDebugArtifact>) -> Result<ast::AST, String> {
2019-08-12 09:32:04 -07:00
use ParsingDebugType::*;
let ref mut parser = handle.active_parser;
parser.add_new_tokens(input);
let ast = parser.parse();
comp.map(|comp| {
2019-08-12 09:32:04 -07:00
let debug_format = comp.parsing.as_ref().unwrap_or(&CompactAST);
let debug_info = match debug_format {
2019-08-12 09:51:36 -07:00
CompactAST => match ast{
Ok(ref ast) => ast.compact_debug(),
2019-08-12 09:32:04 -07:00
Err(_) => "Error - see output".to_string(),
},
2019-08-12 09:51:36 -07:00
ExpandedAST => match ast{
Ok(ref ast) => ast.expanded_debug(),
2019-08-12 09:32:04 -07:00
Err(_) => "Error - see output".to_string(),
},
Trace => parser.format_parse_trace(),
};
comp.add_artifact(debug_info);
});
ast.map_err(|err| format_parse_error(err, &handle.source_reference))
}
fn format_parse_error(error: parsing::ParseError, source_reference: &SourceReference) -> String {
let line_num = error.token.location.line_num;
let ch = error.token.location.char_num;
let line_from_program = source_reference.get_line(line_num);
let location_pointer = format!("{}^", " ".repeat(ch));
let line_num_digits = format!("{}", line_num).chars().count();
let space_padding = " ".repeat(line_num_digits);
let production = match error.production_name {
Some(n) => format!("\n(from production \"{}\")", n),
None => "".to_string()
};
format!(r#"
{error_msg}{production}
{space_padding} |
{line_num} | {}
{space_padding} | {}
"#, line_from_program, location_pointer, error_msg=error.msg, space_padding=space_padding, line_num=line_num, production=production
)
}
fn symbol_table(input: ast::AST, handle: &mut Schala, comp: Option<&mut PassDebugArtifact>) -> Result<ast::AST, String> {
2019-08-05 03:37:37 -07:00
let () = handle.symbol_table.borrow_mut().add_top_level_symbols(&input)?;
comp.map(|comp| {
let debug = handle.symbol_table.borrow().debug_symbol_table();
comp.add_artifact(debug);
});
Ok(input)
}
fn scope_resolution(mut input: ast::AST, handle: &mut Schala, _com: Option<&mut PassDebugArtifact>) -> Result<ast::AST, String> {
let () = handle.resolver.resolve(&mut input)?;
2019-09-03 01:42:28 -07:00
Ok(input)
}
fn typechecking(input: ast::AST, handle: &mut Schala, comp: Option<&mut PassDebugArtifact>) -> Result<ast::AST, String> {
let result = handle.type_context.typecheck(&input);
comp.map(|comp| {
comp.add_artifact(match result {
Ok(ty) => ty.to_string(),
Err(err) => format!("Type error: {}", err.msg)
});
});
Ok(input)
}
fn ast_reducing(input: ast::AST, handle: &mut Schala, comp: Option<&mut PassDebugArtifact>) -> Result<reduced_ast::ReducedAST, String> {
let ref symbol_table = handle.symbol_table.borrow();
let output = reduced_ast::reduce(&input, symbol_table);
comp.map(|comp| comp.add_artifact(format!("{:?}", output)));
Ok(output)
}
fn eval(input: reduced_ast::ReducedAST, handle: &mut Schala, comp: Option<&mut PassDebugArtifact>) -> Result<String, String> {
comp.map(|comp| comp.add_artifact(handle.state.debug_print()));
let evaluation_outputs = handle.state.evaluate(input, true);
let text_output: Result<Vec<String>, String> = evaluation_outputs
.into_iter()
.collect();
let eval_output: Result<String, String> = text_output
2021-10-07 00:51:45 -07:00
.map(|v| { Iterator::intersperse(v.into_iter(), "\n".to_owned()).collect() });
eval_output
}
/// Represents lines of source code
struct SourceReference {
lines: Option<Vec<String>>
}
impl SourceReference {
fn new() -> SourceReference {
SourceReference { lines: None }
}
fn load_new_source(&mut self, source: &str) {
//TODO this is a lot of heap allocations - maybe there's a way to make it more efficient?
self.lines = Some(source.lines().map(|s| s.to_string()).collect()); }
fn get_line(&self, line: usize) -> String {
self.lines.as_ref().and_then(|x| x.get(line).map(|s| s.to_string())).unwrap_or(format!("NO LINE FOUND"))
}
}
enum ParsingDebugType {
CompactAST,
ExpandedAST,
Trace
}
#[derive(Default)]
struct PassDebugArtifact {
parsing: Option<ParsingDebugType>,
artifacts: Vec<String>
}
impl PassDebugArtifact {
fn add_artifact(&mut self, artifact: String) {
self.artifacts.push(artifact)
}
}
fn stage_names() -> Vec<&'static str> {
vec![
"tokenizing",
"parsing",
"symbol-table",
2019-09-03 01:42:28 -07:00
"scope-resolution",
"typechecking",
"ast-reduction",
"ast-walking-evaluation"
]
}
impl ProgrammingLanguageInterface for Schala {
2021-10-13 01:09:24 -07:00
fn language_name(&self) -> String {
"Schala".to_owned()
}
fn source_file_suffix(&self) -> String {
"schala".to_owned()
}
2021-10-13 23:45:54 -07:00
fn run_computation(&mut self, request: ComputationRequest) -> ComputationResponse {
let ComputationRequest { source, debug_requests } = request;
self.source_reference.load_new_source(source);
let sw = Stopwatch::start_new();
let main_output = self.run_pipeline(source);
let global_output_stats = GlobalOutputStats {
total_duration: sw.elapsed(),
stage_durations: vec![]
};
ComputationResponse {
main_output,
global_output_stats,
debug_responses: vec![]
}
}
// n.b. - old-style, overcomplicated `run_computation`. need to revamp the entire metadata system
/*
fn run_computation(&mut self, request: ComputationRequest) -> ComputationResponse {
struct PassToken<'a> {
schala: &'a mut Schala,
stage_durations: &'a mut Vec<(String, Duration)>,
sw: &'a Stopwatch,
debug_requests: &'a HashSet<DebugAsk>,
debug_responses: &'a mut Vec<DebugResponse>,
}
fn output_wrapper<Input, Output, F>(n: usize, func: F, input: Input, token: &mut PassToken) -> Result<Output, String>
where F: Fn(Input, &mut Schala, Option<&mut PassDebugArtifact>) -> Result<Output, String>
{
let stage_names = stage_names();
let cur_stage_name = stage_names[n];
let ask = token.debug_requests.iter().find(|ask| ask.is_for_stage(cur_stage_name));
2019-08-05 03:31:10 -07:00
let parsing = match ask {
Some(DebugAsk::ByStage { token, .. }) if cur_stage_name == "parsing" => Some(
token.as_ref().map(|token| match &token[..] {
"compact" => ParsingDebugType::CompactAST,
"expanded" => ParsingDebugType::ExpandedAST,
"trace" => ParsingDebugType::Trace,
_ => ParsingDebugType::CompactAST,
2019-08-05 03:31:10 -07:00
}).unwrap_or(ParsingDebugType::CompactAST)
),
_ => None,
};
let mut debug_artifact = ask.map(|_| PassDebugArtifact {
parsing, ..Default::default()
});
2019-08-05 03:31:10 -07:00
let output = func(input, token.schala, debug_artifact.as_mut());
2019-08-12 14:13:20 -07:00
//TODO I think this is not counting the time since the *previous* stage
token.stage_durations.push((cur_stage_name.to_string(), token.sw.elapsed()));
if let Some(artifact) = debug_artifact {
for value in artifact.artifacts.into_iter() {
let resp = DebugResponse { ask: ask.unwrap().clone(), value };
token.debug_responses.push(resp);
}
}
output
}
let ComputationRequest { source, debug_requests } = request;
self.source_reference.load_new_source(source);
let sw = Stopwatch::start_new();
let mut stage_durations = Vec::new();
let mut debug_responses = Vec::new();
let mut tok = PassToken { schala: self, stage_durations: &mut stage_durations, sw: &sw, debug_requests: &debug_requests, debug_responses: &mut debug_responses };
let main_output: Result<String, String> = Ok(source)
.and_then(|source| output_wrapper(0, tokenizing, source, &mut tok))
.and_then(|tokens| output_wrapper(1, parsing, tokens, &mut tok))
.and_then(|ast| output_wrapper(2, symbol_table, ast, &mut tok))
2019-09-03 01:42:28 -07:00
.and_then(|ast| output_wrapper(3, scope_resolution, ast, &mut tok))
.and_then(|ast| output_wrapper(4, typechecking, ast, &mut tok))
.and_then(|ast| output_wrapper(5, ast_reducing, ast, &mut tok))
.and_then(|reduced_ast| output_wrapper(6, eval, reduced_ast, &mut tok));
let total_duration = sw.elapsed();
let global_output_stats = GlobalOutputStats {
total_duration, stage_durations
};
ComputationResponse {
main_output,
global_output_stats,
debug_responses,
}
}
2021-10-13 23:45:54 -07:00
*/
fn request_meta(&mut self, request: LangMetaRequest) -> LangMetaResponse {
match request {
LangMetaRequest::StageNames => LangMetaResponse::StageNames(stage_names().iter().map(|s| s.to_string()).collect()),
2021-10-13 23:45:54 -07:00
_ => LangMetaResponse::Custom { kind: format!("not-implemented"), value: format!("") }
}
}
}