Support evaluating most expressions

Blocks, paths, and others

Plus a bunch of other infra improvements
This commit is contained in:
Yehuda Katz
2019-06-29 01:55:42 -07:00
parent 7074d3ffcc
commit 3379c23a49
13 changed files with 497 additions and 150 deletions

View File

@ -9,11 +9,20 @@ use derive_new::new;
use getset::Getters;
crate use baseline_parse::baseline_parse_single_token;
crate use baseline_parse_tokens::{baseline_parse_next_expr, ExpressionKindHint};
crate use baseline_parse_tokens::{baseline_parse_next_expr, ExpressionKindHint, TokensIterator};
crate use binary::Binary;
crate use named::NamedArguments;
crate use path::Path;
pub fn path(head: impl Into<Expression>, tail: Vec<Spanned<impl Into<String>>>) -> Path {
Path::new(
head.into(),
tail.into_iter()
.map(|item| item.map(|string| string.into()))
.collect(),
)
}
#[derive(Debug, Clone, Eq, PartialEq, Getters, new)]
pub struct Call {
#[get = "crate"]
@ -29,7 +38,7 @@ pub enum RawExpression {
Literal(Literal),
Variable(Variable),
Binary(Box<Binary>),
Block(Box<Expression>),
Block(Vec<Expression>),
Path(Box<Path>),
#[allow(unused)]
@ -76,6 +85,12 @@ impl Expression {
}
}
impl From<Spanned<Path>> for Expression {
fn from(path: Spanned<Path>) -> Expression {
path.map(|p| RawExpression::Path(Box::new(p)))
}
}
#[derive(Debug, Clone, Copy, Eq, PartialEq, Ord, PartialOrd, Hash)]
pub enum Literal {
Integer(i64),

View File

@ -1,24 +1,27 @@
use crate::errors::ShellError;
use crate::parser::registry::CommandRegistry;
use crate::parser::{hir, hir::baseline_parse_single_token, Span, Spanned, TokenNode};
use crate::Text;
use crate::parser::{
hir, hir::baseline_parse_single_token, DelimitedNode, Delimiter, PathNode, RawToken, Span,
Spanned, TokenNode,
};
use crate::{SpannedItem, Text};
use derive_new::new;
use log::trace;
pub fn baseline_parse_tokens(
token_nodes: &[TokenNode],
token_nodes: &mut TokensIterator<'_>,
registry: &dyn CommandRegistry,
source: &Text,
) -> Result<Vec<hir::Expression>, ShellError> {
let mut exprs: Vec<hir::Expression> = vec![];
let mut rest = token_nodes;
loop {
if rest.len() == 0 {
if token_nodes.at_end() {
break;
}
let (expr, remainder) = baseline_parse_next_expr(rest, registry, source, None)?;
let expr = baseline_parse_next_expr(token_nodes, registry, source, None)?;
exprs.push(expr);
rest = remainder;
}
Ok(exprs)
@ -35,25 +38,21 @@ pub enum ExpressionKindHint {
}
pub fn baseline_parse_next_expr(
token_nodes: &'nodes [TokenNode],
_registry: &dyn CommandRegistry,
tokens: &mut TokensIterator,
registry: &dyn CommandRegistry,
source: &Text,
coerce_hint: Option<ExpressionKindHint>,
) -> Result<(hir::Expression, &'nodes [TokenNode]), ShellError> {
let mut tokens = token_nodes.iter().peekable();
let first = next_token(&mut tokens);
let first = match first {
) -> Result<hir::Expression, ShellError> {
let first = match tokens.next() {
None => return Err(ShellError::string("Expected token, found none")),
Some(token) => baseline_parse_semantic_token(token, source)?,
Some(token) => baseline_parse_semantic_token(token, registry, source)?,
};
let possible_op = tokens.peek();
let op = match possible_op {
Some(TokenNode::Operator(op)) => op,
_ => return Ok((first, &token_nodes[1..])),
Some(TokenNode::Operator(op)) => op.clone(),
_ => return Ok(first),
};
tokens.next();
@ -64,7 +63,7 @@ pub fn baseline_parse_next_expr(
"Expected op followed by another expr, found nothing",
))
}
Some(token) => baseline_parse_semantic_token(token, source)?,
Some(token) => baseline_parse_semantic_token(token, registry, source)?,
};
// We definitely have a binary expression here -- let's see if we should coerce it into a block
@ -72,11 +71,11 @@ pub fn baseline_parse_next_expr(
match coerce_hint {
None => {
let span = (first.span.start, second.span.end);
let binary = hir::Binary::new(first, *op, second);
let binary = hir::Binary::new(first, op, second);
let binary = hir::RawExpression::Binary(Box::new(binary));
let binary = Spanned::from_item(binary, span);
Ok((binary, &token_nodes[3..]))
Ok(binary)
}
Some(hint) => match hint {
@ -133,14 +132,14 @@ pub fn baseline_parse_next_expr(
}
};
let binary = hir::Binary::new(path, *op, second);
let binary = hir::Binary::new(path, op, second);
let binary = hir::RawExpression::Binary(Box::new(binary));
let binary = Spanned::from_item(binary, span);
let block = hir::RawExpression::Block(Box::new(binary));
let block = hir::RawExpression::Block(vec![binary]);
let block = Spanned::from_item(block, span);
Ok((block, &token_nodes[3..]))
Ok(block)
}
other => unimplemented!("coerce hint {:?}", other),
@ -150,27 +149,199 @@ pub fn baseline_parse_next_expr(
pub fn baseline_parse_semantic_token(
token: &TokenNode,
registry: &dyn CommandRegistry,
source: &Text,
) -> Result<hir::Expression, ShellError> {
match token {
TokenNode::Token(token) => Ok(baseline_parse_single_token(token, source)),
TokenNode::Call(_call) => unimplemented!(),
TokenNode::Delimited(_delimited) => unimplemented!(),
TokenNode::Delimited(delimited) => baseline_parse_delimited(delimited, registry, source),
TokenNode::Pipeline(_pipeline) => unimplemented!(),
TokenNode::Operator(_op) => unreachable!(),
TokenNode::Flag(_flag) => unimplemented!(),
TokenNode::Identifier(_span) => unreachable!(),
TokenNode::Whitespace(_span) => unreachable!(),
TokenNode::Error(error) => Err(*error.item.clone()),
TokenNode::Path(_path) => unimplemented!(),
TokenNode::Path(path) => baseline_parse_path(path, registry, source),
}
}
fn next_token(nodes: &mut impl Iterator<Item = &'a TokenNode>) -> Option<&'a TokenNode> {
loop {
match nodes.next() {
Some(TokenNode::Whitespace(_)) => continue,
other => return other,
pub fn baseline_parse_delimited(
token: &Spanned<DelimitedNode>,
registry: &dyn CommandRegistry,
source: &Text,
) -> Result<hir::Expression, ShellError> {
match token.delimiter() {
Delimiter::Brace => {
let children = token.children();
let exprs =
baseline_parse_tokens(&mut TokensIterator::new(children), registry, source)?;
let expr = hir::RawExpression::Block(exprs);
Ok(Spanned::from_item(expr, token.span()))
}
Delimiter::Paren => unimplemented!(),
Delimiter::Square => unimplemented!(),
}
}
pub fn baseline_parse_path(
token: &Spanned<PathNode>,
registry: &dyn CommandRegistry,
source: &Text,
) -> Result<hir::Expression, ShellError> {
let head = baseline_parse_semantic_token(token.head(), registry, source)?;
let mut tail = vec![];
for part in token.tail() {
let string = match part {
TokenNode::Token(token) => match token.item() {
RawToken::Bare => token.span().slice(source),
RawToken::String(span) => span.slice(source),
RawToken::Integer(_) | RawToken::Size(..) | RawToken::Variable(_) => {
return Err(ShellError::type_error(
"String",
token.type_name().spanned(part),
))
}
},
TokenNode::Identifier(span) => span.slice(source),
// TODO: Make this impossible
other => unreachable!("{:?}", other),
}
.to_string();
tail.push(string.spanned(part));
}
Ok(hir::path(head, tail).spanned(token).into())
}
#[derive(Debug, new)]
pub struct TokensIterator<'a> {
tokens: &'a [TokenNode],
#[new(default)]
index: usize,
#[new(default)]
seen: indexmap::IndexSet<usize>,
}
impl TokensIterator<'a> {
pub fn remove(&mut self, position: usize) {
self.seen.insert(position);
}
pub fn len(&self) -> usize {
self.tokens.len()
}
pub fn at_end(&self) -> bool {
for index in self.index..self.tokens.len() {
if !self.seen.contains(&index) {
return false;
}
}
true
}
pub fn advance(&mut self) {
self.seen.insert(self.index);
self.index += 1;
}
pub fn extract<T>(&mut self, f: impl Fn(&TokenNode) -> Option<T>) -> Option<(usize, T)> {
for (i, item) in self.tokens.iter().enumerate() {
if self.seen.contains(&i) {
continue;
}
match f(item) {
None => {
continue;
}
Some(value) => {
self.seen.insert(i);
return Some((i, value));
}
}
}
None
}
pub fn move_to(&mut self, pos: usize) {
self.index = pos;
}
pub fn restart(&mut self) {
self.index = 0;
}
pub fn clone(&self) -> TokensIterator {
TokensIterator {
tokens: self.tokens,
index: self.index,
seen: self.seen.clone(),
}
}
pub fn peek(&self) -> Option<&TokenNode> {
let mut tokens = self.clone();
tokens.next()
}
pub fn debug_remaining(&self) -> Vec<TokenNode> {
let mut tokens = self.clone();
tokens.restart();
tokens.cloned().collect()
}
}
impl Iterator for TokensIterator<'a> {
type Item = &'a TokenNode;
fn next(&mut self) -> Option<&'a TokenNode> {
loop {
if self.index >= self.tokens.len() {
return None;
}
if self.seen.contains(&self.index) {
self.advance();
continue;
}
if self.index >= self.tokens.len() {
return None;
}
match &self.tokens[self.index] {
TokenNode::Whitespace(_) => {
self.advance();
}
other => {
self.advance();
return Some(other);
}
}
}
}
}
pub fn trace_remaining(desc: &'static str, tail: hir::TokensIterator<'a>, source: &Text) {
trace!(
"{} = {:?}",
desc,
itertools::join(
tail.debug_remaining()
.iter()
.map(|i| format!("%{:?}%", i.debug(source))),
" "
)
);
}

View File

@ -12,6 +12,14 @@ pub struct Spanned<T> {
pub item: T,
}
pub trait SpannedItem: Sized {
fn spanned(self, span: impl Into<Span>) -> Spanned<Self> {
Spanned::from_item(self, span.into())
}
}
impl<T> SpannedItem for T {}
impl<T> std::ops::Deref for Spanned<T> {
type Target = T;
@ -56,6 +64,12 @@ pub struct Span {
// source: &'source str,
}
impl<T> From<&Spanned<T>> for Span {
fn from(input: &Spanned<T>) -> Span {
input.span
}
}
impl From<&Span> for Span {
fn from(input: &Span) -> Span {
*input

View File

@ -3,6 +3,8 @@ use crate::parser::parse::{call_node::*, flag::*, operator::*, pipeline::*, span
use crate::Text;
use derive_new::new;
use enum_utils::FromStr;
use getset::Getters;
use std::fmt;
#[derive(Debug, Clone, Eq, PartialEq, Ord, PartialOrd)]
pub enum TokenNode {
@ -20,6 +22,67 @@ pub enum TokenNode {
Path(Spanned<PathNode>),
}
pub struct DebugTokenNode<'a> {
node: &'a TokenNode,
source: &'a Text,
}
impl fmt::Debug for DebugTokenNode<'a> {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
match self.node {
TokenNode::Token(t) => write!(f, "{:?}", t.debug(self.source)),
TokenNode::Call(s) => {
write!(f, "(")?;
write!(f, "{:?}", s.head().debug(self.source))?;
if let Some(children) = s.children() {
for child in children {
write!(f, "{:?}", child.debug(self.source))?;
}
}
write!(f, ")")
}
TokenNode::Delimited(d) => {
write!(
f,
"{}",
match d.delimiter {
Delimiter::Brace => "{",
Delimiter::Paren => "(",
Delimiter::Square => "[",
}
)?;
for child in d.children() {
write!(f, "{:?}", child.debug(self.source))?;
}
write!(
f,
"{}",
match d.delimiter {
Delimiter::Brace => "}",
Delimiter::Paren => ")",
Delimiter::Square => "]",
}
)
}
TokenNode::Pipeline(s) => write!(f, "<todo:pipeline>"),
TokenNode::Error(s) => write!(f, "<error> for {:?}", s.span().slice(self.source)),
rest => write!(f, "{}", rest.span().slice(self.source)),
}
}
}
impl From<&TokenNode> for Span {
fn from(token: &TokenNode) -> Span {
token.span()
}
}
impl TokenNode {
pub fn span(&self) -> Span {
match self {
@ -36,6 +99,10 @@ impl TokenNode {
}
}
pub fn debug(&'a self, source: &'a Text) -> DebugTokenNode<'a> {
DebugTokenNode { node: self, source }
}
pub fn as_external_arg(&self, source: &Text) -> String {
self.span().slice(source).to_string()
}
@ -73,7 +140,8 @@ impl TokenNode {
}
}
#[derive(Debug, Clone, Eq, PartialEq, Ord, PartialOrd, new)]
#[derive(Debug, Clone, Eq, PartialEq, Ord, PartialOrd, Getters, new)]
#[get = "crate"]
pub struct DelimitedNode {
delimiter: Delimiter,
children: Vec<TokenNode>,
@ -86,7 +154,8 @@ pub enum Delimiter {
Square,
}
#[derive(Debug, Clone, Eq, PartialEq, Ord, PartialOrd, new)]
#[derive(Debug, Clone, Eq, PartialEq, Ord, PartialOrd, Getters, new)]
#[get = "crate"]
pub struct PathNode {
head: Box<TokenNode>,
tail: Vec<TokenNode>,

View File

@ -1,5 +1,7 @@
use crate::parser::parse::span::*;
use crate::parser::parse::unit::*;
use crate::Text;
use std::fmt;
#[derive(Debug, Clone, Copy, Eq, PartialEq, Ord, PartialOrd, Hash)]
pub enum RawToken {
@ -10,4 +12,36 @@ pub enum RawToken {
Bare,
}
impl RawToken {
pub fn type_name(&self) -> &'static str {
match self {
RawToken::Integer(_) => "Integer",
RawToken::Size(..) => "Size",
RawToken::String(_) => "String",
RawToken::Variable(_) => "Variable",
RawToken::Bare => "String",
}
}
}
pub type Token = Spanned<RawToken>;
impl Token {
pub fn debug(&self, source: &'a Text) -> DebugToken<'a> {
DebugToken {
node: *self,
source,
}
}
}
pub struct DebugToken<'a> {
node: Token,
source: &'a Text,
}
impl fmt::Debug for DebugToken<'a> {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(f, "{}", self.node.span().slice(self.source))
}
}

View File

@ -67,77 +67,81 @@ fn parse_command_tail(
tail: Option<Vec<TokenNode>>,
source: &Text,
) -> Result<Option<(Option<Vec<hir::Expression>>, Option<NamedArguments>)>, ShellError> {
let mut tail = match tail {
let tail = &mut match &tail {
None => return Ok(None),
Some(tail) => tail,
Some(tail) => hir::TokensIterator::new(tail),
};
let mut named = NamedArguments::new();
trace_remaining("nodes", tail.clone(), source);
for (name, kind) in config.named() {
trace!("looking for {} : {:?}", name, kind);
match kind {
NamedType::Switch => {
let (rest, flag) = extract_switch(name, tail, source);
tail = rest;
let flag = extract_switch(name, tail, source);
named.insert_switch(name, flag);
}
NamedType::Mandatory(kind) => match extract_mandatory(name, tail, source) {
Err(err) => return Err(err), // produce a correct diagnostic
Ok((rest, pos, _flag)) => {
let (expr, rest) = hir::baseline_parse_next_expr(
&rest[pos..],
Ok((pos, _flag)) => {
tail.move_to(pos);
let expr = hir::baseline_parse_next_expr(
tail,
registry,
source,
kind.to_coerce_hint(),
)?;
tail = rest.to_vec();
tail.restart();
named.insert_mandatory(name, expr);
}
},
NamedType::Optional(kind) => match extract_optional(name, tail, source) {
Err(err) => return Err(err), // produce a correct diagnostic
Ok((rest, Some((pos, _flag)))) => {
let (expr, rest) = hir::baseline_parse_next_expr(
&rest[pos..],
Ok(Some((pos, _flag))) => {
tail.move_to(pos);
let expr = hir::baseline_parse_next_expr(
tail,
registry,
source,
kind.to_coerce_hint(),
)?;
tail = rest.to_vec();
tail.restart();
named.insert_optional(name, Some(expr));
}
Ok((rest, None)) => {
tail = rest;
Ok(None) => {
tail.restart();
named.insert_optional(name, None);
}
},
};
}
trace_remaining("after named", tail.clone(), source);
let mut positional = vec![];
let mandatory = config.mandatory_positional();
for arg in mandatory {
trace!("Processing mandatory {:?}", arg);
if tail.len() == 0 {
return Err(ShellError::unimplemented("Missing mandatory argument"));
}
let (result, rest) =
hir::baseline_parse_next_expr(&tail, registry, source, arg.to_coerce_hint())?;
let result = hir::baseline_parse_next_expr(tail, registry, source, arg.to_coerce_hint())?;
positional.push(result);
tail = rest.to_vec();
}
trace_remaining("after mandatory", tail.clone(), source);
let optional = config.optional_positional();
for arg in optional {
@ -145,18 +149,19 @@ fn parse_command_tail(
break;
}
let (result, rest) =
hir::baseline_parse_next_expr(&tail, registry, source, arg.to_coerce_hint())?;
let result = hir::baseline_parse_next_expr(tail, registry, source, arg.to_coerce_hint())?;
positional.push(result);
tail = rest.to_vec();
}
trace_remaining("after optional", tail.clone(), source);
// TODO: Only do this if rest params are specified
let remainder = baseline_parse_tokens(&tail, registry, source)?;
let remainder = baseline_parse_tokens(tail, registry, source)?;
positional.extend(remainder);
trace_remaining("after rest", tail.clone(), source);
trace!("Constructed positional={:?} named={:?}", positional, named);
let positional = match positional {
@ -174,38 +179,20 @@ fn parse_command_tail(
Ok(Some((positional, named)))
}
fn extract_switch(
name: &str,
mut tokens: Vec<TokenNode>,
source: &Text,
) -> (Vec<TokenNode>, Option<Flag>) {
let pos = tokens
.iter()
.enumerate()
.filter_map(|(i, t)| t.as_flag(name, source).map(|f| (i, f)))
.nth(0);
match pos {
None => (tokens, None),
Some((pos, flag)) => {
tokens.remove(pos);
(tokens, Some(*flag))
}
}
fn extract_switch(name: &str, tokens: &mut hir::TokensIterator<'_>, source: &Text) -> Option<Flag> {
tokens
.extract(|t| t.as_flag(name, source))
.map(|(_pos, flag)| flag.item)
}
fn extract_mandatory(
name: &str,
mut tokens: Vec<TokenNode>,
tokens: &mut hir::TokensIterator<'a>,
source: &Text,
) -> Result<(Vec<TokenNode>, usize, Flag), ShellError> {
let pos = tokens
.iter()
.enumerate()
.filter_map(|(i, t)| t.as_flag(name, source).map(|f| (i, f)))
.nth(0);
) -> Result<(usize, Flag), ShellError> {
let flag = tokens.extract(|t| t.as_flag(name, source));
match pos {
match flag {
None => Err(ShellError::unimplemented(
"Better error: mandatory flags must be present",
)),
@ -218,24 +205,20 @@ fn extract_mandatory(
tokens.remove(pos);
Ok((tokens, pos, *flag))
Ok((pos, *flag))
}
}
}
fn extract_optional(
name: &str,
mut tokens: Vec<TokenNode>,
tokens: &mut hir::TokensIterator<'a>,
source: &Text,
) -> Result<(Vec<TokenNode>, Option<(usize, Flag)>), ShellError> {
let pos = tokens
.iter()
.enumerate()
.filter_map(|(i, t)| t.as_flag(name, source).map(|f| (i, f)))
.nth(0);
) -> Result<(Option<(usize, Flag)>), ShellError> {
let flag = tokens.extract(|t| t.as_flag(name, source));
match pos {
None => Ok((tokens, None)),
match flag {
None => Ok(None),
Some((pos, flag)) => {
if tokens.len() <= pos {
return Err(ShellError::unimplemented(
@ -245,7 +228,20 @@ fn extract_optional(
tokens.remove(pos);
Ok((tokens, Some((pos, *flag))))
Ok(Some((pos, *flag)))
}
}
}
pub fn trace_remaining(desc: &'static str, tail: hir::TokensIterator<'a>, source: &Text) {
trace!(
"{} = {:?}",
desc,
itertools::join(
tail.debug_remaining()
.iter()
.map(|i| format!("%{:?}%", i.debug(source))),
" "
)
);
}