From bf417fadb206d6d2351a13cd7c6988977a46dd33 Mon Sep 17 00:00:00 2001 From: Nadrieril Date: Wed, 18 Sep 2019 22:37:30 +0200 Subject: Extract pest_consume into its own crate --- pest_consume_macros/Cargo.toml | 19 -- pest_consume_macros/src/lib.rs | 27 --- pest_consume_macros/src/make_parser.rs | 424 --------------------------------- pest_consume_macros/src/match_nodes.rs | 248 ------------------- 4 files changed, 718 deletions(-) delete mode 100644 pest_consume_macros/Cargo.toml delete mode 100644 pest_consume_macros/src/lib.rs delete mode 100644 pest_consume_macros/src/make_parser.rs delete mode 100644 pest_consume_macros/src/match_nodes.rs (limited to 'pest_consume_macros') diff --git a/pest_consume_macros/Cargo.toml b/pest_consume_macros/Cargo.toml deleted file mode 100644 index aaa0b10..0000000 --- a/pest_consume_macros/Cargo.toml +++ /dev/null @@ -1,19 +0,0 @@ -[package] -name = "pest_consume_macros" -version = "0.1.0" # remember to update html_root_url -authors = ["Nadrieril "] -license = "MIT OR Apache-2.0" -edition = "2018" -description = "Macros for pest_consume" -readme = "README.md" -repository = "https://github.com/Nadrieril/dhall-rust" - -[lib] -proc-macro = true -doctest = false - -[dependencies] -quote = "1.0.2" -proc-macro2 = "1.0.2" -proc-macro-hack = "0.5.9" -syn = { version = "1.0.5", features = ["full"] } diff --git a/pest_consume_macros/src/lib.rs b/pest_consume_macros/src/lib.rs deleted file mode 100644 index d726b5d..0000000 --- a/pest_consume_macros/src/lib.rs +++ /dev/null @@ -1,27 +0,0 @@ -//! This crate contains the code-generation primitives for the [dhall-rust][dhall-rust] crate. -//! This is highly unstable and breaks regularly; use at your own risk. -//! -//! [dhall-rust]: https://github.com/Nadrieril/dhall-rust - -extern crate proc_macro; - -mod make_parser; -mod match_nodes; - -use proc_macro::TokenStream; - -#[proc_macro_attribute] -pub fn parser(attrs: TokenStream, input: TokenStream) -> TokenStream { - TokenStream::from(match make_parser::make_parser(attrs, input) { - Ok(tokens) => tokens, - Err(err) => err.to_compile_error(), - }) -} - -#[proc_macro_hack::proc_macro_hack] -pub fn match_nodes(input: TokenStream) -> TokenStream { - TokenStream::from(match match_nodes::match_nodes(input) { - Ok(tokens) => tokens, - Err(err) => err.to_compile_error(), - }) -} diff --git a/pest_consume_macros/src/make_parser.rs b/pest_consume_macros/src/make_parser.rs deleted file mode 100644 index 5bf8fe3..0000000 --- a/pest_consume_macros/src/make_parser.rs +++ /dev/null @@ -1,424 +0,0 @@ -use std::collections::HashMap; -use std::iter; - -use quote::quote; -use syn::parse::{Parse, ParseStream, Result}; -use syn::spanned::Spanned; -use syn::{ - parse_quote, Error, Expr, FnArg, Ident, ImplItem, ImplItemMethod, ItemImpl, - LitBool, Pat, Path, Token, -}; - -/// Ext. trait adding `partition_filter` to `Vec`. Would like to use `Vec::drain_filter` -/// but it's unstable for now. -pub trait VecPartitionFilterExt { - fn partition_filter(&mut self, predicate: F) -> Vec - where - F: FnMut(&mut Item) -> bool; -} - -impl VecPartitionFilterExt for Vec { - fn partition_filter(&mut self, mut predicate: F) -> Vec - where - F: FnMut(&mut Item) -> bool, - { - let mut ret = Vec::new(); - let mut i = 0; - while i != self.len() { - if predicate(&mut self[i]) { - ret.push(self.remove(i)) - } else { - i += 1; - } - } - ret - } -} - -mod kw { - syn::custom_keyword!(shortcut); - syn::custom_keyword!(rule); - syn::custom_keyword!(parser); -} - -struct MakeParserAttrs { - parser: Path, - rule_enum: Path, -} - -struct AliasArgs { - target: Ident, - is_shortcut: bool, -} - -struct PrecClimbArgs { - child_rule: Ident, - climber: Expr, -} - -struct AliasSrc { - ident: Ident, - is_shortcut: bool, -} - -struct ParsedFn<'a> { - // Body of the function - function: &'a mut ImplItemMethod, - // Name of the function. - fn_name: Ident, - // Name of the first argument of the function, which should be of type `Node`. - input_arg: Ident, - // List of aliases pointing to this function - alias_srcs: Vec, -} - -impl Parse for MakeParserAttrs { - fn parse(input: ParseStream) -> Result { - // By default, the pest parser is the same type as the pest_consume one - let mut parser = parse_quote!(Self); - // By default, use the `Rule` type in scope - let mut rule_enum = parse_quote!(Rule); - - while !input.is_empty() { - let lookahead = input.lookahead1(); - if lookahead.peek(kw::parser) { - let _: kw::parser = input.parse()?; - let _: Token![=] = input.parse()?; - parser = input.parse()?; - } else if lookahead.peek(kw::rule) { - let _: kw::rule = input.parse()?; - let _: Token![=] = input.parse()?; - rule_enum = input.parse()?; - } else { - return Err(lookahead.error()); - } - - if input.peek(Token![,]) { - let _: Token![,] = input.parse()?; - } else { - break; - } - } - - Ok(MakeParserAttrs { parser, rule_enum }) - } -} - -impl Parse for AliasArgs { - fn parse(input: ParseStream) -> Result { - let target = input.parse()?; - let is_shortcut = if input.peek(Token![,]) { - // #[alias(rule, shortcut = true)] - let _: Token![,] = input.parse()?; - let _: kw::shortcut = input.parse()?; - let _: Token![=] = input.parse()?; - let b: LitBool = input.parse()?; - b.value - } else { - // #[alias(rule)] - false - }; - Ok(AliasArgs { - target, - is_shortcut, - }) - } -} - -impl Parse for PrecClimbArgs { - fn parse(input: ParseStream) -> Result { - let child_rule = input.parse()?; - let _: Token![,] = input.parse()?; - let climber = input.parse()?; - Ok(PrecClimbArgs { - child_rule, - climber, - }) - } -} - -fn collect_aliases( - imp: &mut ItemImpl, -) -> Result>> { - let functions = imp.items.iter_mut().flat_map(|item| match item { - ImplItem::Method(m) => Some(m), - _ => None, - }); - - let mut alias_map = HashMap::new(); - for function in functions { - let fn_name = function.sig.ident.clone(); - let mut alias_attrs = function - .attrs - .partition_filter(|attr| attr.path.is_ident("alias")) - .into_iter(); - - if let Some(attr) = alias_attrs.next() { - let args: AliasArgs = attr.parse_args()?; - alias_map.entry(args.target).or_insert_with(Vec::new).push( - AliasSrc { - ident: fn_name, - is_shortcut: args.is_shortcut, - }, - ); - } else { - // Self entry - alias_map - .entry(fn_name.clone()) - .or_insert_with(Vec::new) - .push(AliasSrc { - ident: fn_name, - is_shortcut: false, - }); - } - if let Some(attr) = alias_attrs.next() { - return Err(Error::new( - attr.span(), - "expected at most one alias attribute", - )); - } - } - - Ok(alias_map) -} - -fn parse_fn<'a>( - function: &'a mut ImplItemMethod, - alias_map: &mut HashMap>, -) -> Result> { - let fn_name = function.sig.ident.clone(); - // Get the name of the first (`input`) function argument - let input_arg = function.sig.inputs.first().ok_or_else(|| { - Error::new( - function.sig.inputs.span(), - "a rule function needs an `input` argument", - ) - })?; - let input_arg = match &input_arg { - FnArg::Receiver(_) => return Err(Error::new( - input_arg.span(), - "a rule function should not have a `self` argument", - )), - FnArg::Typed(input_arg) => match &*input_arg.pat{ - Pat::Ident(ident) => ident.ident.clone(), - _ => return Err(Error::new( - input_arg.span(), - "this argument should be a plain identifier instead of a pattern", - )), - } - }; - - let alias_srcs = alias_map.remove(&fn_name).unwrap_or_else(Vec::new); - - Ok(ParsedFn { - function, - fn_name, - input_arg, - alias_srcs, - }) -} - -fn apply_special_attrs(f: &mut ParsedFn, rule_enum: &Path) -> Result<()> { - let function = &mut *f.function; - let fn_name = &f.fn_name; - let input_arg = &f.input_arg; - - *function = parse_quote!( - #[allow(non_snake_case)] - #function - ); - - // `prec_climb` attr - let prec_climb_attrs: Vec<_> = function - .attrs - .partition_filter(|attr| attr.path.is_ident("prec_climb")); - - if prec_climb_attrs.len() > 1 { - return Err(Error::new( - prec_climb_attrs[1].span(), - "expected at most one prec_climb attribute", - )); - } else if prec_climb_attrs.is_empty() { - // do nothing - } else { - let attr = prec_climb_attrs.into_iter().next().unwrap(); - let PrecClimbArgs { - child_rule, - climber, - } = attr.parse_args()?; - - function.block = parse_quote!({ - #function - - #climber.climb( - #input_arg.as_pair().clone().into_inner(), - |p| Self::#child_rule(#input_arg.with_pair(p)), - |l, op, r| { - #fn_name(#input_arg.clone(), l?, op, r?) - }, - ) - }); - // Remove the 3 last arguments to keep only the `input` one - function.sig.inputs.pop(); - function.sig.inputs.pop(); - function.sig.inputs.pop(); - // Check that an argument remains - function.sig.inputs.first().ok_or_else(|| { - Error::new( - function.sig.inputs.span(), - "a prec_climb function needs 4 arguments", - ) - })?; - } - - // `alias` attr - // f.alias_srcs has always at least 1 element because it has an entry pointing from itself. - if f.alias_srcs.len() > 1 { - let aliases = f - .alias_srcs - .iter() - .map(|src| &src.ident) - .filter(|i| i != &fn_name); - let block = &function.block; - let self_ty = quote!(); - function.block = parse_quote!({ - let mut #input_arg = #input_arg; - // While the current rule allows shortcutting, and there is a single child, and the - // child can still be parsed by the current function, then skip to that child. - while #self_ty::allows_shortcut(#input_arg.as_rule()) { - if let ::std::option::Option::Some(child) = #input_arg.single_child() { - if child.as_aliased_rule::() == #self_ty::AliasedRule::#fn_name { - #input_arg = child; - continue; - } - } - break - } - - match #input_arg.as_rule() { - #(#rule_enum::#aliases => Self::#aliases(#input_arg),)* - #rule_enum::#fn_name => #block, - r => ::std::unreachable!( - "make_parser: called {} on {:?}", - ::std::stringify!(#fn_name), - r - ) - } - }); - } - - Ok(()) -} - -pub fn make_parser( - attrs: proc_macro::TokenStream, - input: proc_macro::TokenStream, -) -> Result { - let attrs: MakeParserAttrs = syn::parse(attrs)?; - let parser = &attrs.parser; - let rule_enum = &attrs.rule_enum; - let mut imp: ItemImpl = syn::parse(input)?; - - let mut alias_map = collect_aliases(&mut imp)?; - let rule_alias_branches: Vec<_> = alias_map - .iter() - .flat_map(|(tgt, srcs)| iter::repeat(tgt).zip(srcs)) - .map(|(tgt, src)| { - let ident = &src.ident; - quote!( - #rule_enum::#ident => Self::AliasedRule::#tgt, - ) - }) - .collect(); - let aliased_rule_variants: Vec<_> = - alias_map.iter().map(|(tgt, _)| tgt.clone()).collect(); - let shortcut_branches: Vec<_> = alias_map - .iter() - .flat_map(|(_tgt, srcs)| srcs) - .map(|AliasSrc { ident, is_shortcut }| { - quote!( - #rule_enum::#ident => #is_shortcut, - ) - }) - .collect(); - - let fn_map: HashMap = imp - .items - .iter_mut() - .flat_map(|item| match item { - ImplItem::Method(m) => Some(m), - _ => None, - }) - .map(|method| { - let mut f = parse_fn(method, &mut alias_map)?; - apply_special_attrs(&mut f, &rule_enum)?; - Ok((f.fn_name.clone(), f)) - }) - .collect::>()?; - - // Entries that remain in the alias map don't have a matching method, so we create one. - let extra_fns: Vec<_> = alias_map - .iter() - .map(|(tgt, srcs)| { - // Get the signature of one of the functions that has this alias. They should all have - // essentially the same signature anyways. - let f = fn_map.get(&srcs.first().unwrap().ident).unwrap(); - let input_arg = f.input_arg.clone(); - let mut sig = f.function.sig.clone(); - sig.ident = tgt.clone(); - let srcs = srcs.iter().map(|src| &src.ident); - - Ok(parse_quote!( - #sig { - match #input_arg.as_rule() { - #(#rule_enum::#srcs => Self::#srcs(#input_arg),)* - // We can't match on #rule_enum::#tgt since `tgt` might be an arbitrary - // identifier. - r if &::std::format!("{:?}", r) == ::std::stringify!(#tgt) => - return ::std::result::Result::Err(#input_arg.error(::std::format!( - "make_parser: missing method for rule {}", - ::std::stringify!(#tgt), - ))), - r => ::std::unreachable!( - "make_parser: called {} on {:?}", - ::std::stringify!(#tgt), - r - ) - } - } - )) - }) - .collect::>()?; - imp.items.extend(extra_fns); - - let ty = &imp.self_ty; - let (impl_generics, _, where_clause) = imp.generics.split_for_impl(); - Ok(quote!( - #[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)] - #[allow(non_camel_case_types)] - enum AliasedRule { - #(#aliased_rule_variants,)* - } - - impl #impl_generics ::pest_consume::Parser for #ty #where_clause { - type Rule = #rule_enum; - type AliasedRule = AliasedRule; - type Parser = #parser; - fn rule_alias(rule: Self::Rule) -> Self::AliasedRule { - match rule { - #(#rule_alias_branches)* - // TODO: return a proper error ? - r => ::std::unreachable!("Rule {:?} does not have a corresponding parsing method", r), - } - } - fn allows_shortcut(rule: Self::Rule) -> bool { - match rule { - #(#shortcut_branches)* - _ => false, - } - } - } - - #imp - )) -} diff --git a/pest_consume_macros/src/match_nodes.rs b/pest_consume_macros/src/match_nodes.rs deleted file mode 100644 index a91d1f4..0000000 --- a/pest_consume_macros/src/match_nodes.rs +++ /dev/null @@ -1,248 +0,0 @@ -use proc_macro2::{Span, TokenStream}; -use quote::quote; -use syn::parse::{Parse, ParseStream, Result}; -use syn::punctuated::Punctuated; -use syn::spanned::Spanned; -use syn::{ - bracketed, parenthesized, parse_quote, token, Error, Expr, Ident, Pat, - Token, Type, -}; - -#[derive(Clone)] -struct MatchBranch { - // Patterns all have the form [a, b, c.., d], with a bunch of simple patterns, - // optionally a multiple pattern, and then some more simple patterns. - singles_before_multiple: Vec<(Ident, Pat)>, - multiple: Option<(Ident, Ident)>, - singles_after_multiple: Vec<(Ident, Pat)>, - - body: Expr, -} - -#[derive(Clone)] -enum MatchBranchPatternItem { - Single { - rule_name: Ident, - binder: Pat, - }, - Multiple { - rule_name: Ident, - binder: Ident, - slice_token: Token![..], - }, -} - -#[derive(Clone)] -struct MacroInput { - parser: Type, - input_expr: Expr, - branches: Punctuated, -} - -impl Parse for MatchBranch { - fn parse(input: ParseStream) -> Result { - let contents; - let _: token::Bracket = bracketed!(contents in input); - - let pattern: Punctuated = - Punctuated::parse_terminated(&contents)?; - use MatchBranchPatternItem::{Multiple, Single}; - let mut singles_before_multiple = Vec::new(); - let mut multiple = None; - let mut singles_after_multiple = Vec::new(); - for item in pattern.clone() { - match item { - Single { rule_name, binder } => { - if multiple.is_none() { - singles_before_multiple.push((rule_name, binder)) - } else { - singles_after_multiple.push((rule_name, binder)) - } - } - Multiple { - rule_name, - binder, - slice_token, - } => { - if multiple.is_none() { - multiple = Some((rule_name, binder)) - } else { - return Err(Error::new( - slice_token.span(), - "multiple variable-length patterns are not allowed", - )); - } - } - } - } - - let _: Token![=>] = input.parse()?; - let body = input.parse()?; - - Ok(MatchBranch { - singles_before_multiple, - multiple, - singles_after_multiple, - body, - }) - } -} - -impl Parse for MatchBranchPatternItem { - fn parse(input: ParseStream) -> Result { - let contents; - let rule_name = input.parse()?; - parenthesized!(contents in input); - if input.peek(Token![..]) { - let binder = contents.parse()?; - let slice_token = input.parse()?; - Ok(MatchBranchPatternItem::Multiple { - rule_name, - binder, - slice_token, - }) - } else if input.is_empty() || input.peek(Token![,]) { - let binder = contents.parse()?; - Ok(MatchBranchPatternItem::Single { rule_name, binder }) - } else { - Err(input.error("expected `..` or nothing")) - } - } -} - -impl Parse for MacroInput { - fn parse(input: ParseStream) -> Result { - let parser = if input.peek(token::Lt) { - let _: token::Lt = input.parse()?; - let parser = input.parse()?; - let _: token::Gt = input.parse()?; - let _: Token![;] = input.parse()?; - parser - } else { - parse_quote!(Self) - }; - let input_expr = input.parse()?; - let _: Token![;] = input.parse()?; - let branches = Punctuated::parse_terminated(input)?; - - Ok(MacroInput { - parser, - input_expr, - branches, - }) - } -} - -fn make_branch( - branch: &MatchBranch, - i_nodes: &Ident, - i_node_rules: &Ident, - parser: &Type, -) -> Result { - let aliased_rule = quote!(<#parser as ::pest_consume::Parser>::AliasedRule); - - // Find which branch to take - let mut conditions = Vec::new(); - let start = branch.singles_before_multiple.len(); - let end = branch.singles_after_multiple.len(); - conditions.push(quote!( - #start + #end <= #i_node_rules.len() - )); - for (i, (rule_name, _)) in branch.singles_before_multiple.iter().enumerate() - { - conditions.push(quote!( - #i_node_rules[#i] == #aliased_rule::#rule_name - )) - } - for (i, (rule_name, _)) in branch.singles_after_multiple.iter().enumerate() - { - conditions.push(quote!( - #i_node_rules[#i_node_rules.len()-1 - #i] == #aliased_rule::#rule_name - )) - } - if let Some((rule_name, _)) = &branch.multiple { - conditions.push(quote!( - { - // We can't use .all() directly in the pattern guard; see - // https://github.com/rust-lang/rust/issues/59803. - let all_match = |slice: &[_]| { - slice.iter().all(|r| - *r == #aliased_rule::#rule_name - ) - }; - all_match(&#i_node_rules[#start..#i_node_rules.len() - #end]) - } - )) - } else { - // No variable-length pattern, so the size must be exactly the number of patterns - conditions.push(quote!( - #start + #end == #i_node_rules.len() - )) - } - - // Once we have found a branch that matches, we need to parse the nodes. - let mut parses = Vec::new(); - for (rule_name, binder) in branch.singles_before_multiple.iter() { - parses.push(quote!( - let #binder = #parser::#rule_name( - #i_nodes.next().unwrap() - )?; - )) - } - // Note the `rev()`: we are taking nodes from the end of the iterator in reverse order, so that - // only the unmatched nodes are left in the iterator for the variable-length pattern, if any. - for (rule_name, binder) in branch.singles_after_multiple.iter().rev() { - parses.push(quote!( - let #binder = #parser::#rule_name( - #i_nodes.next_back().unwrap() - )?; - )) - } - if let Some((rule_name, binder)) = &branch.multiple { - parses.push(quote!( - let #binder = #i_nodes - .map(|i| #parser::#rule_name(i)) - .collect::<::std::result::Result<::std::vec::Vec<_>, _>>()? - .into_iter(); - )) - } - - let body = &branch.body; - Ok(quote!( - _ if #(#conditions &&)* true => { - #(#parses)* - #body - } - )) -} - -pub fn match_nodes( - input: proc_macro::TokenStream, -) -> Result { - let input: MacroInput = syn::parse(input)?; - - let i_nodes = Ident::new("___nodes", input.input_expr.span()); - let i_node_rules = Ident::new("___node_rules", Span::call_site()); - - let input_expr = &input.input_expr; - let parser = &input.parser; - let branches = input - .branches - .iter() - .map(|br| make_branch(br, &i_nodes, &i_node_rules, parser)) - .collect::>>()?; - - Ok(quote!({ - #[allow(unused_mut)] - let mut #i_nodes = #input_expr; - let #i_node_rules = #i_nodes.aliased_rules::<#parser>(); - - #[allow(unreachable_code)] - match () { - #(#branches,)* - _ => return ::std::result::Result::Err(#i_nodes.error( - std::format!("Nodes didn't match any pattern: {:?}", #i_node_rules) - )), - } - })) -} -- cgit v1.2.3