From fe618ff6e38ec0ed4da72a3935fd9ea64ee1cef5 Mon Sep 17 00:00:00 2001 From: Titus Wormer Date: Mon, 19 Sep 2022 11:17:26 +0200 Subject: Add support for parsing MDX ESM, expressions This commit adds support for hooks that lets a user integrate another parser with `micromark-rs`, to parse ESM and expressions according to a particular grammar (such as a programming language, typically JavaScript). For an example integrating with SWC, see `tests/test_utils/mod.rs`. The integration occurs with two functions passed in `options`: `mdx_expression_parse` and `mdx_esm_parse`. The can signal back to micromark when they are successful, whether there is an error at the end (in which case micromark will try to parse more), or whether there is a syntax error (in which case micromark will crash). --- src/construct/mdx_esm.rs | 224 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 224 insertions(+) create mode 100644 src/construct/mdx_esm.rs (limited to 'src/construct/mdx_esm.rs') diff --git a/src/construct/mdx_esm.rs b/src/construct/mdx_esm.rs new file mode 100644 index 0000000..53f8beb --- /dev/null +++ b/src/construct/mdx_esm.rs @@ -0,0 +1,224 @@ +//! MDX ESM occurs in the [flow][] content type. +//! +//! ## Grammar +//! +//! MDX expression (flow) forms with the following BNF +//! (see [construct][crate::construct] for character groups): +//! +//! ```bnf +//! mdx_esm ::= word *line *(eol *line) +//! +//! word ::= 'e' 'x' 'p' 'o' 'r' 't' | 'i' 'm' 'p' 'o' 'r' 't' +//! ``` +//! +//! This construct must be followed by a blank line or eof (end of file). +//! It can include blank lines if [`MdxEsmParse`][crate::MdxEsmParse] passed in +//! `options.mdx_esm_parse` allows it. +//! +//! ## Tokens +//! +//! * [`LineEnding`][Name::LineEnding] +//! * [`MdxEsm`][Name::MdxEsm] +//! * [`MdxEsmData`][Name::MdxEsmData] +//! +//! ## References +//! +//! * [`syntax.js` in `micromark-extension-mdxjs-esm`](https://github.com/micromark/micromark-extension-mdxjs-esm/blob/main/dev/lib/syntax.js) +//! * [`mdxjs.com`](https://mdxjs.com) +//! +//! [flow]: crate::construct::flow + +use crate::event::Name; +use crate::state::{Name as StateName, State}; +use crate::tokenizer::Tokenizer; +use crate::util::{ + mdx_collect::{collect, place_to_point}, + slice::Slice, +}; +use crate::MdxSignal; +use alloc::format; + +/// Start of MDX ESM. +/// +/// ```markdown +/// > | import a from 'b' +/// ^ +/// ``` +pub fn start(tokenizer: &mut Tokenizer) -> State { + // If it’s turned on. + if tokenizer.parse_state.options.constructs.mdx_esm + // If there is a gnostic parser. + && tokenizer.parse_state.options.mdx_esm_parse.is_some() + // When not interrupting. + && !tokenizer.interrupt + // Only at the start of a line, not at whitespace or in a container. + && tokenizer.point.column == 1 + && matches!(tokenizer.current, Some(b'e' | b'i')) + { + // Place where keyword starts. + tokenizer.tokenize_state.start = tokenizer.point.index; + tokenizer.enter(Name::MdxEsm); + tokenizer.enter(Name::MdxEsmData); + tokenizer.consume(); + State::Next(StateName::MdxEsmWord) + } else { + State::Nok + } +} + +/// In keyword. +/// +/// ```markdown +/// > | import a from 'b' +/// ^^^^^^ +/// ``` +pub fn word(tokenizer: &mut Tokenizer) -> State { + if matches!(tokenizer.current, Some(b'a'..=b'z')) { + tokenizer.consume(); + State::Next(StateName::MdxEsmWord) + } else { + let slice = Slice::from_indices( + tokenizer.parse_state.bytes, + tokenizer.tokenize_state.start, + tokenizer.point.index, + ); + + if matches!(slice.as_str(), "export" | "import") && tokenizer.current == Some(b' ') { + tokenizer.concrete = true; + tokenizer.tokenize_state.start = tokenizer.events.len() - 1; + tokenizer.consume(); + State::Next(StateName::MdxEsmInside) + } else { + tokenizer.tokenize_state.start = 0; + State::Nok + } + } +} + +/// In data. +/// +/// ```markdown +/// > | import a from 'b' +/// ^ +/// ``` +pub fn inside(tokenizer: &mut Tokenizer) -> State { + match tokenizer.current { + None | Some(b'\n') => { + tokenizer.exit(Name::MdxEsmData); + State::Retry(StateName::MdxEsmLineStart) + } + _ => { + tokenizer.consume(); + State::Next(StateName::MdxEsmInside) + } + } +} + +/// At start of line. +/// +/// ```markdown +/// | import a from 'b' +/// > | export {a} +/// ^ +/// ``` +pub fn line_start(tokenizer: &mut Tokenizer) -> State { + match tokenizer.current { + None => State::Retry(StateName::MdxEsmAtEnd), + Some(b'\n') => { + tokenizer.check( + State::Next(StateName::MdxEsmAtEnd), + State::Next(StateName::MdxEsmContinuationStart), + ); + State::Retry(StateName::MdxEsmBlankLineBefore) + } + _ => { + tokenizer.enter(Name::MdxEsmData); + tokenizer.consume(); + State::Next(StateName::MdxEsmInside) + } + } +} + +/// At start of line that continues. +/// +/// ```markdown +/// | import a from 'b' +/// > | export {a} +/// ^ +/// ``` +pub fn continuation_start(tokenizer: &mut Tokenizer) -> State { + tokenizer.enter(Name::LineEnding); + tokenizer.consume(); + tokenizer.exit(Name::LineEnding); + State::Next(StateName::MdxEsmLineStart) +} + +/// At start of a potentially blank line. +/// +/// ```markdown +/// | import a from 'b' +/// > | export {a} +/// ^ +/// ``` +pub fn blank_line_before(tokenizer: &mut Tokenizer) -> State { + tokenizer.enter(Name::LineEnding); + tokenizer.consume(); + tokenizer.exit(Name::LineEnding); + State::Next(StateName::BlankLineStart) +} + +/// At end of line (blank or eof). +/// +/// ```markdown +/// > | import a from 'b' +/// ^ +/// ``` +pub fn at_end(tokenizer: &mut Tokenizer) -> State { + let result = parse_esm(tokenizer); + + // Done!. + if matches!(result, State::Ok) { + tokenizer.concrete = false; + tokenizer.exit(Name::MdxEsm); + } + + result +} + +/// Parse ESM with a given function. +fn parse_esm(tokenizer: &mut Tokenizer) -> State { + // We can `unwrap` because we don’t parse if this is `None`. + let parse = tokenizer + .parse_state + .options + .mdx_esm_parse + .as_ref() + .unwrap(); + + // Collect the body of the ESM and positional info for each run of it. + let result = collect( + tokenizer, + tokenizer.tokenize_state.start, + &[Name::MdxEsmData, Name::LineEnding], + ); + + // Parse and handle what was signaled back. + match parse(&result.value) { + MdxSignal::Ok => State::Ok, + MdxSignal::Error(message, place) => { + let point = place_to_point(&result, place); + State::Error(format!("{}:{}: {}", point.line, point.column, message)) + } + MdxSignal::Eof(message) => { + if tokenizer.current == None { + State::Error(format!( + "{}:{}: {}", + tokenizer.point.line, tokenizer.point.column, message + )) + } else { + tokenizer.tokenize_state.mdx_last_parse_error = Some(message); + State::Retry(StateName::MdxEsmContinuationStart) + } + } + } +} -- cgit