1 | use crate::core::binary::EncodeOptions; |
2 | use crate::core::*; |
3 | use crate::parser::{Parse, Parser, Result}; |
4 | use crate::token::{Id, Index, NameAnnotation, Span}; |
5 | use crate::{annotation, kw}; |
6 | |
7 | pub use crate::core::resolve::Names; |
8 | |
9 | /// A parsed WebAssembly core module. |
10 | #[derive (Debug)] |
11 | pub struct Module<'a> { |
12 | /// Where this `module` was defined |
13 | pub span: Span, |
14 | /// An optional identifier this module is known by |
15 | pub id: Option<Id<'a>>, |
16 | /// An optional `@name` annotation for this module |
17 | pub name: Option<NameAnnotation<'a>>, |
18 | /// What kind of module this was parsed as. |
19 | pub kind: ModuleKind<'a>, |
20 | } |
21 | |
22 | /// The different kinds of ways to define a module. |
23 | #[derive (Debug)] |
24 | pub enum ModuleKind<'a> { |
25 | /// A module defined in the textual s-expression format. |
26 | Text(Vec<ModuleField<'a>>), |
27 | /// A module that had its raw binary bytes defined via the `binary` |
28 | /// directive. |
29 | Binary(Vec<&'a [u8]>), |
30 | } |
31 | |
32 | impl<'a> Module<'a> { |
33 | /// Performs a name resolution pass on this [`Module`], resolving all |
34 | /// symbolic names to indices. |
35 | /// |
36 | /// The WAT format contains a number of shorthands to make it easier to |
37 | /// write, such as inline exports, inline imports, inline type definitions, |
38 | /// etc. Additionally it allows using symbolic names such as `$foo` instead |
39 | /// of using indices. This module will postprocess an AST to remove all of |
40 | /// this syntactic sugar, preparing the AST for binary emission. This is |
41 | /// where expansion and name resolution happens. |
42 | /// |
43 | /// This function will mutate the AST of this [`Module`] and replace all |
44 | /// [`Index`](crate::token::Index) arguments with `Index::Num`. This will |
45 | /// also expand inline exports/imports listed on fields and handle various |
46 | /// other shorthands of the text format. |
47 | /// |
48 | /// If successful the AST was modified to be ready for binary encoding. A |
49 | /// [`Names`] structure is also returned so if you'd like to do your own |
50 | /// name lookups on the result you can do so as well. |
51 | /// |
52 | /// # Errors |
53 | /// |
54 | /// If an error happens during resolution, such a name resolution error or |
55 | /// items are found in the wrong order, then an error is returned. |
56 | pub fn resolve(&mut self) -> std::result::Result<Names<'a>, crate::Error> { |
57 | let names = match &mut self.kind { |
58 | ModuleKind::Text(fields) => crate::core::resolve::resolve(fields)?, |
59 | ModuleKind::Binary(_blobs) => Default::default(), |
60 | }; |
61 | Ok(names) |
62 | } |
63 | |
64 | /// Encodes this [`Module`] to its binary form. |
65 | /// |
66 | /// This function will take the textual representation in [`Module`] and |
67 | /// perform all steps necessary to convert it to a binary WebAssembly |
68 | /// module, suitable for writing to a `*.wasm` file. This function may |
69 | /// internally modify the [`Module`], for example: |
70 | /// |
71 | /// * Name resolution is performed to ensure that `Index::Id` isn't present |
72 | /// anywhere in the AST. |
73 | /// |
74 | /// * Inline shorthands such as imports/exports/types are all expanded to be |
75 | /// dedicated fields of the module. |
76 | /// |
77 | /// * Module fields may be shuffled around to preserve index ordering from |
78 | /// expansions. |
79 | /// |
80 | /// After all of this expansion has happened the module will be converted to |
81 | /// its binary form and returned as a `Vec<u8>`. This is then suitable to |
82 | /// hand off to other wasm runtimes and such. |
83 | /// |
84 | /// # Errors |
85 | /// |
86 | /// This function can return an error for name resolution errors and other |
87 | /// expansion-related errors. |
88 | pub fn encode(&mut self) -> std::result::Result<Vec<u8>, crate::Error> { |
89 | EncodeOptions::default().encode_module(self) |
90 | } |
91 | |
92 | pub(crate) fn validate(&self, parser: Parser<'_>) -> Result<()> { |
93 | let mut starts = 0; |
94 | if let ModuleKind::Text(fields) = &self.kind { |
95 | for item in fields.iter() { |
96 | if let ModuleField::Start(_) = item { |
97 | starts += 1; |
98 | } |
99 | } |
100 | } |
101 | if starts > 1 { |
102 | return Err(parser.error("multiple start sections found" )); |
103 | } |
104 | Ok(()) |
105 | } |
106 | |
107 | pub(crate) fn parse_without_module_keyword( |
108 | module_keyword_span: Span, |
109 | parser: Parser<'a>, |
110 | ) -> Result<Self> { |
111 | let id = parser.parse()?; |
112 | let name = parser.parse()?; |
113 | |
114 | let kind = if parser.peek::<kw::binary>()? { |
115 | parser.parse::<kw::binary>()?; |
116 | let mut data = Vec::new(); |
117 | while !parser.is_empty() { |
118 | data.push(parser.parse()?); |
119 | } |
120 | ModuleKind::Binary(data) |
121 | } else { |
122 | ModuleKind::Text(ModuleField::parse_remaining(parser)?) |
123 | }; |
124 | Ok(Module { |
125 | span: module_keyword_span, |
126 | id, |
127 | name, |
128 | kind, |
129 | }) |
130 | } |
131 | } |
132 | |
133 | impl<'a> Parse<'a> for Module<'a> { |
134 | fn parse(parser: Parser<'a>) -> Result<Self> { |
135 | parser.with_standard_annotations_registered(|parser: Parser<'a>| { |
136 | let span: Span = parser.parse::<kw::module>()?.0; |
137 | Self::parse_without_module_keyword(span, parser) |
138 | }) |
139 | } |
140 | } |
141 | |
142 | /// A listing of all possible fields that can make up a WebAssembly module. |
143 | #[allow (missing_docs)] |
144 | #[derive (Debug)] |
145 | pub enum ModuleField<'a> { |
146 | Type(Type<'a>), |
147 | Rec(Rec<'a>), |
148 | Import(Import<'a>), |
149 | Func(Func<'a>), |
150 | Table(Table<'a>), |
151 | Memory(Memory<'a>), |
152 | Global(Global<'a>), |
153 | Export(Export<'a>), |
154 | Start(Index<'a>), |
155 | Elem(Elem<'a>), |
156 | Data(Data<'a>), |
157 | Tag(Tag<'a>), |
158 | Custom(Custom<'a>), |
159 | } |
160 | |
161 | impl<'a> ModuleField<'a> { |
162 | pub(crate) fn parse_remaining(parser: Parser<'a>) -> Result<Vec<ModuleField<'a>>> { |
163 | let mut fields: Vec> = Vec::new(); |
164 | while !parser.is_empty() { |
165 | fields.push(parser.parens(ModuleField::parse)?); |
166 | } |
167 | Ok(fields) |
168 | } |
169 | } |
170 | |
171 | impl<'a> Parse<'a> for ModuleField<'a> { |
172 | fn parse(parser: Parser<'a>) -> Result<Self> { |
173 | if parser.peek::<Type<'a>>()? { |
174 | return Ok(ModuleField::Type(parser.parse()?)); |
175 | } |
176 | if parser.peek::<kw::rec>()? { |
177 | return Ok(ModuleField::Rec(parser.parse()?)); |
178 | } |
179 | if parser.peek::<kw::import>()? { |
180 | return Ok(ModuleField::Import(parser.parse()?)); |
181 | } |
182 | if parser.peek::<kw::func>()? { |
183 | return Ok(ModuleField::Func(parser.parse()?)); |
184 | } |
185 | if parser.peek::<kw::table>()? { |
186 | return Ok(ModuleField::Table(parser.parse()?)); |
187 | } |
188 | if parser.peek::<kw::memory>()? { |
189 | return Ok(ModuleField::Memory(parser.parse()?)); |
190 | } |
191 | if parser.peek::<kw::global>()? { |
192 | return Ok(ModuleField::Global(parser.parse()?)); |
193 | } |
194 | if parser.peek::<kw::export>()? { |
195 | return Ok(ModuleField::Export(parser.parse()?)); |
196 | } |
197 | if parser.peek::<kw::start>()? { |
198 | parser.parse::<kw::start>()?; |
199 | return Ok(ModuleField::Start(parser.parse()?)); |
200 | } |
201 | if parser.peek::<kw::elem>()? { |
202 | return Ok(ModuleField::Elem(parser.parse()?)); |
203 | } |
204 | if parser.peek::<kw::data>()? { |
205 | return Ok(ModuleField::Data(parser.parse()?)); |
206 | } |
207 | if parser.peek::<kw::tag>()? { |
208 | return Ok(ModuleField::Tag(parser.parse()?)); |
209 | } |
210 | if parser.peek::<annotation::custom>()? |
211 | || parser.peek::<annotation::producers>()? |
212 | || parser.peek::<annotation::dylink_0>()? |
213 | { |
214 | return Ok(ModuleField::Custom(parser.parse()?)); |
215 | } |
216 | Err(parser.error("expected valid module field" )) |
217 | } |
218 | } |
219 | |