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    /// Whether a single `$` also opens inline math.
52    /// Prettier (`singleDollarTextMath: false`) requires `$$`;
53    /// keep this off to leave prose like `$FOO` and `$10` alone.
54    pub math_text_single_dollar: bool,
55    /// `{% … %}` / `{{ … }}` template tags, multi-line capable.
56    /// Never interpreted, kept lossless.
57    pub liquid: bool,
58    /// `[[target]]` wiki links, alias-less minimal form, printed verbatim.
59    pub wiki_link: bool,
60
61    /// `::: … :::` container directives.
62    /// Fence lines are kept verbatim (dialects disagree on their grammar);
63    /// only children are markdown.
64    pub container_directive: bool,
65
66    /// MDX `import` / `export` statements at flow start.
67    pub mdx_esm: bool,
68    /// MDX `{expr}` as its own flow block.
69    pub mdx_expression_flow: bool,
70    /// MDX `{expr}` in text.
71    pub mdx_expression_text: bool,
72    /// MDX JSX elements at flow start.
73    pub mdx_jsx_flow: bool,
74    /// MDX JSX elements in text.
75    pub mdx_jsx_text: bool,
76}
77
78impl Constructs {
79    /// Markdown mode: CommonMark + GFM + Prettier's extensions + directives.
80    pub fn markdown() -> Self {
81        Self {
82            autolink: true,
83            code_indented: true,
84            html_flow: true,
85            html_text: true,
86            gfm_autolink_literal: true,
87            gfm_footnote: true,
88            gfm_strikethrough: true,
89            gfm_strikethrough_single_tilde: false,
90            gfm_table: true,
91            gfm_task_list_item: true,
92            math_flow: true,
93            math_text: true,
94            math_text_single_dollar: false,
95            liquid: true,
96            wiki_link: true,
97            container_directive: true,
98            mdx_esm: false,
99            mdx_expression_flow: false,
100            mdx_expression_text: false,
101            mdx_jsx_flow: false,
102            mdx_jsx_text: false,
103        }
104    }
105
106    /// MDX mode: `autolink` / `code_indented` / `html_flow` / `html_text` off, the `mdx_*` family on.
107    /// Unlike markdown mode, parsing can emit diagnostics (MDX has syntax errors; CommonMark does not).
108    ///
109    /// TODO: Not usable yet: the `mdx_*` constructs are unimplemented,
110    /// and disabling `code_indented` must also lift micromark's 4-space/3-column indent limits engine-wide (currently hardcoded),
111    /// so this preset misparses e.g. `    # hi`.
112    ///
113    /// NOTE: Design:
114    /// - Expression boundaries need JavaScript lexing (`{"}"}`, `{/* } */}`, template literals),
115    ///   so this preset lives behind an `mdx` cargo feature that adds `oxc_parser` as an optional dependency.
116    ///   No injection point: the scanner is this crate's own module, tested here.
117    ///   The oxc workspace patches `oxc_parser` to its path crate, as it does `oxc_allocator`.
118    /// - Semantics follow Prettier's MDX v3 parser (prettier/prettier#18533, micromark-extension-mdx-*):
119    ///   - `{expr}` (flow, text, JSX attribute values): at each candidate `}`,
120    ///     the text so far must be a complete expression (`Parser::parse_expression`, which rejects trailing content);
121    ///     otherwise continue to the next `}`, and fail at EOF.
122    ///   - `import` / `export`: ends at a blank line unless the parse error is at the end of input
123    ///     (label start == length, or an `Unterminated ...` lexer error), then continue past it.
124    ///     Prettier's `dummyAcorn` does not validate the body, so neither do we.
125    ///   - JSX tags are lexed here; only their `{...}` attribute values consult the scanner.
126    ///   - A hand-written brace scanner would agree on every input Prettier can format
127    ///     (they differ only inside invalid JS, where Prettier throws),
128    ///     which is why `oxc_lexer`, once published, can replace the probing without changing fixtures.
129    pub fn mdx() -> Self {
130        Self {
131            autolink: false,
132            code_indented: false,
133            html_flow: false,
134            html_text: false,
135            mdx_esm: true,
136            mdx_expression_flow: true,
137            mdx_expression_text: true,
138            mdx_jsx_flow: true,
139            mdx_jsx_text: true,
140            ..Self::markdown()
141        }
142    }
143}
144
145impl Default for Constructs {
146    fn default() -> Self {
147        Self::markdown()
148    }
149}
150
151/// Options for [`Parser`](crate::Parser).
152#[derive(Clone, Copy, Debug, Default, PartialEq, Eq)]
153pub struct ParserOptions {
154    pub constructs: Constructs,
155}
156
157impl ParserOptions {
158    /// Markdown mode (the default): [`Constructs::markdown`].
159    pub fn markdown() -> Self {
160        Self { constructs: Constructs::markdown() }
161    }
162
163    /// MDX mode: [`Constructs::mdx`].
164    pub fn mdx() -> Self {
165        Self { constructs: Constructs::mdx() }
166    }
167}
oxc_markdown_parser/options.rs

oxc_markdown_parser/
options.rs
