oxc_markdown_parser/
options.rs

1//! Parser configuration.
2//!
3//! MD and MDX are "the same syntax with different meanings":
4//! the syntactic difference reduces to which constructs are enabled,
5//! so both dialects share one parser, one AST, and one API.
6//! Downstream code never branches on a dialect enum, only on which nodes actually appear.
7
8/// Which syntax constructs the parser recognizes.
9///
10/// CommonMark's core (paragraphs, headings, thematic breaks, blockquotes,
11/// lists, fenced code, emphasis, links, hard breaks) is always on.
12/// Everything listed here is a construct that some dialect turns off or bolts on.
13///
14/// The default set mirrors Prettier's parser composition:
15/// GFM + math + wiki-link + liquid,
16/// plus container directives (`:::`), which Prettier lacks and consequently corrupts.
17#[derive(Clone, Copy, Debug, PartialEq, Eq)]
18pub struct Constructs {
19    /// `<https://example.com>` / `<user@example.com>`. Syntax error in MDX.
20    pub autolink: bool,
21    /// 4-space indented code blocks.
22    /// In MDX the indent is an ordinary paragraph (continuation) instead.
23    pub code_indented: bool,
24    /// HTML blocks (CommonMark types 1–7).
25    /// In MDX `<` at flow start is JSX.
26    pub html_flow: bool,
27    /// Inline HTML tags/comments.
28    /// In MDX `<` in text is JSX.
29    pub html_text: bool,
30
31    /// GFM bare-URL links (`www.example.com`).
32    /// Printed verbatim.
33    pub gfm_autolink_literal: bool,
34    /// GFM footnotes (`[^label]`).
35    pub gfm_footnote: bool,
36    /// GFM strikethrough (`~~x~~`).
37    pub gfm_strikethrough: bool,
38    /// Whether a single `~` also opens strikethrough.
39    /// GitHub and Prettier (`singleTilde: false`) require `~~`;
40    /// keep this off to leave subscript conventions like `H~2~O` alone.
41    pub gfm_strikethrough_single_tilde: bool,
42    /// GFM tables.
43    pub gfm_table: bool,
44    /// GFM task list items (`- [x]`).
45    pub gfm_task_list_item: bool,
46
47    /// `$$ … $$` display math.
48    pub math_flow: bool,
49    /// `$ … $` inline math, printed verbatim.
50    pub math_text: bool,
51    /// `{% … %}` / `{{ … }}` template tags, multi-line capable.
52    /// Never interpreted, kept lossless.
53    pub liquid: bool,
54    /// `[[target]]` wiki links, alias-less minimal form, printed verbatim.
55    pub wiki_link: bool,
56
57    /// `::: … :::` container directives.
58    /// Fence lines are kept verbatim (dialects disagree on their grammar);
59    /// only children are markdown.
60    pub container_directive: bool,
61
62    /// MDX `import` / `export` statements at flow start.
63    pub mdx_esm: bool,
64    /// MDX `{expr}` as its own flow block.
65    pub mdx_expression_flow: bool,
66    /// MDX `{expr}` in text.
67    pub mdx_expression_text: bool,
68    /// MDX JSX elements at flow start.
69    pub mdx_jsx_flow: bool,
70    /// MDX JSX elements in text.
71    pub mdx_jsx_text: bool,
72}
73
74impl Constructs {
75    /// Markdown mode: CommonMark + GFM + Prettier's extensions + directives.
76    pub fn markdown() -> Self {
77        Self {
78            autolink: true,
79            code_indented: true,
80            html_flow: true,
81            html_text: true,
82            gfm_autolink_literal: true,
83            gfm_footnote: true,
84            gfm_strikethrough: true,
85            gfm_strikethrough_single_tilde: false,
86            gfm_table: true,
87            gfm_task_list_item: true,
88            math_flow: true,
89            math_text: true,
90            liquid: true,
91            wiki_link: true,
92            container_directive: true,
93            mdx_esm: false,
94            mdx_expression_flow: false,
95            mdx_expression_text: false,
96            mdx_jsx_flow: false,
97            mdx_jsx_text: false,
98        }
99    }
100
101    /// MDX mode: `autolink` / `code_indented` / `html_flow` / `html_text` off, the `mdx_*` family on.
102    /// Unlike markdown mode, parsing can emit diagnostics (MDX has syntax errors; CommonMark does not).
103    ///
104    /// TODO: Not usable yet: the `mdx_*` constructs are unimplemented,
105    /// and disabling `code_indented` must also lift micromark's 4-space/3-column indent limits engine-wide (currently hardcoded),
106    /// so this preset misparses e.g. `    # hi`.
107    ///
108    /// NOTE: Design:
109    /// - Expression boundaries need JavaScript lexing (`{"}"}`, `{/* } */}`, template literals),
110    ///   so this preset lives behind an `mdx` cargo feature that adds `oxc_parser` as an optional dependency.
111    ///   No injection point: the scanner is this crate's own module, tested here.
112    ///   The oxc workspace patches `oxc_parser` to its path crate, as it does `oxc_allocator`.
113    /// - Semantics follow Prettier's MDX v3 parser (prettier/prettier#18533, micromark-extension-mdx-*):
114    ///   - `{expr}` (flow, text, JSX attribute values): at each candidate `}`,
115    ///     the text so far must be a complete expression (`Parser::parse_expression`, which rejects trailing content);
116    ///     otherwise continue to the next `}`, and fail at EOF.
117    ///   - `import` / `export`: ends at a blank line unless the parse error is at the end of input
118    ///     (label start == length, or an `Unterminated ...` lexer error), then continue past it.
119    ///     Prettier's `dummyAcorn` does not validate the body, so neither do we.
120    ///   - JSX tags are lexed here; only their `{...}` attribute values consult the scanner.
121    ///   - A hand-written brace scanner would agree on every input Prettier can format
122    ///     (they differ only inside invalid JS, where Prettier throws),
123    ///     which is why `oxc_lexer`, once published, can replace the probing without changing fixtures.
124    pub fn mdx() -> Self {
125        Self {
126            autolink: false,
127            code_indented: false,
128            html_flow: false,
129            html_text: false,
130            mdx_esm: true,
131            mdx_expression_flow: true,
132            mdx_expression_text: true,
133            mdx_jsx_flow: true,
134            mdx_jsx_text: true,
135            ..Self::markdown()
136        }
137    }
138}
139
140impl Default for Constructs {
141    fn default() -> Self {
142        Self::markdown()
143    }
144}
145
146/// Options for [`Parser`](crate::Parser).
147#[derive(Clone, Copy, Debug, Default, PartialEq, Eq)]
148pub struct ParserOptions {
149    pub constructs: Constructs,
150}
151
152impl ParserOptions {
153    /// Markdown mode (the default): [`Constructs::markdown`].
154    pub fn markdown() -> Self {
155        Self { constructs: Constructs::markdown() }
156    }
157
158    /// MDX mode: [`Constructs::mdx`].
159    pub fn mdx() -> Self {
160        Self { constructs: Constructs::mdx() }
161    }
162}
oxc_markdown_parser/options.rs

oxc_markdown_parser/
options.rs
