0
0
Fork 0
mirror of https://github.com/denoland/deno.git synced 2025-02-24 06:02:18 -05:00
denoland-deno/cli/tools/lint/ast_buffer/buffer.rs
Bartek Iwańczuk f08ca6414b
feat(lint): add JavaScript plugin support (#27203)
This commit adds an unstable lint plugin API.

Plugins are specified in the `deno.json` file under
`lint.plugins` option like so:

```
{
  "lint": {
    "plugins": [
      "./plugins/my-plugin.ts",
      "jsr:@deno/lint-plugin1",
      "npm:@deno/lint-plugin2"
    ]
  }
}
```

The API is considered unstable and might be subject
to changes in the future.

Plugin API was modelled after ESLint API for the 
most part, but there are no guarantees for compatibility.
The AST format exposed to plugins is closely modelled
after the AST that `typescript-eslint` uses.

Lint plugins use the visitor pattern and can add
diagnostics like so:

```
export default {
  name: "lint-plugin",
  rules: {
    "plugin-rule": {
      create(context) {
        return {
          Identifier(node) {
            if (node.name === "a") {
              context.report({
                node,
                message: "should be b",
                fix(fixer) {
                  return fixer.replaceText(node, "_b");
                },
              });
            }
          },
        };
      },
    },
  },
} satisfies Deno.lint.Plugin;
```

Besides reporting errors (diagnostics) plugins can provide
automatic fixes that use text replacement to apply changes.

---------

Co-authored-by: Marvin Hagemeister <marvin@deno.com>
Co-authored-by: David Sherret <dsherret@gmail.com>
2025-02-05 16:59:24 +01:00

604 lines
16 KiB
Rust

// Copyright 2018-2025 the Deno authors. MIT license.
use std::fmt::Display;
use deno_ast::swc::common::Span;
use deno_ast::swc::common::DUMMY_SP;
use indexmap::IndexMap;
use crate::util::text_encoding::Utf16Map;
/// Each property has this flag to mark what kind of value it holds-
/// Plain objects and arrays are not supported yet, but could be easily
/// added if needed.
#[derive(Debug, PartialEq)]
pub enum PropFlags {
Ref,
RefArr,
String,
Number,
Bool,
Null,
Undefined,
Object,
Regex,
BigInt,
Array,
}
impl From<PropFlags> for u8 {
fn from(m: PropFlags) -> u8 {
m as u8
}
}
impl TryFrom<u8> for PropFlags {
type Error = &'static str;
fn try_from(value: u8) -> Result<Self, Self::Error> {
match value {
0 => Ok(PropFlags::Ref),
1 => Ok(PropFlags::RefArr),
2 => Ok(PropFlags::String),
3 => Ok(PropFlags::Number),
4 => Ok(PropFlags::Bool),
5 => Ok(PropFlags::Null),
6 => Ok(PropFlags::Undefined),
7 => Ok(PropFlags::Object),
8 => Ok(PropFlags::Regex),
9 => Ok(PropFlags::BigInt),
10 => Ok(PropFlags::Array),
_ => Err("Unknown Prop flag"),
}
}
}
pub type Index = u32;
const GROUP_KIND: u8 = 1;
const MASK_U32_1: u32 = 0b11111111_00000000_00000000_00000000;
const MASK_U32_2: u32 = 0b00000000_11111111_00000000_00000000;
const MASK_U32_3: u32 = 0b00000000_00000000_11111111_00000000;
const MASK_U32_4: u32 = 0b00000000_00000000_00000000_11111111;
#[inline]
fn append_u32(result: &mut Vec<u8>, value: u32) {
let v1: u8 = ((value & MASK_U32_1) >> 24) as u8;
let v2: u8 = ((value & MASK_U32_2) >> 16) as u8;
let v3: u8 = ((value & MASK_U32_3) >> 8) as u8;
let v4: u8 = (value & MASK_U32_4) as u8;
result.push(v1);
result.push(v2);
result.push(v3);
result.push(v4);
}
fn append_usize(result: &mut Vec<u8>, value: usize) {
let raw = u32::try_from(value).unwrap();
append_u32(result, raw);
}
#[derive(Debug)]
pub struct StringTable {
id: usize,
table: IndexMap<String, usize>,
}
impl StringTable {
pub fn new() -> Self {
Self {
id: 0,
table: IndexMap::new(),
}
}
pub fn insert(&mut self, s: &str) -> usize {
if let Some(id) = self.table.get(s) {
return *id;
}
let id = self.id;
self.id += 1;
self.table.insert(s.to_string(), id);
id
}
pub fn serialize(&mut self) -> Vec<u8> {
let mut result: Vec<u8> = vec![];
append_u32(&mut result, self.table.len() as u32);
// Assume that it's sorted by id
for (s, _id) in &self.table {
let bytes = s.as_bytes();
append_u32(&mut result, bytes.len() as u32);
result.append(&mut bytes.to_vec());
}
result
}
}
#[derive(Debug, Clone, Copy, PartialEq)]
pub struct NodeRef(pub Index);
#[derive(Debug, Clone, Copy, PartialEq)]
pub struct PendingRef(pub Index);
pub trait AstBufSerializer {
fn serialize(&mut self) -> Vec<u8>;
}
/// <type u8>
/// <prop offset u32>
/// <child idx u32>
/// <next idx u32>
/// <parent idx u32>
#[derive(Debug)]
struct Node {
kind: u8,
prop_offset: u32,
child: u32,
next: u32,
parent: u32,
}
#[derive(Debug)]
pub struct SerializeCtx {
root_idx: Index,
nodes: Vec<Node>,
prop_stack: Vec<Vec<u8>>,
field_count: Vec<usize>,
field_buf: Vec<u8>,
prev_sibling_stack: Vec<Index>,
/// Vec of spans
spans: Vec<u32>,
/// Maps string id to the actual string
str_table: StringTable,
/// Maps kind id to string id
kind_name_map: Vec<usize>,
/// Maps prop id to string id
prop_name_map: Vec<usize>,
}
/// This is the internal context used to allocate and fill the buffer. The point
/// is to be able to write absolute offsets directly in place.
///
/// The typical workflow is to reserve all necessary space for the currrent
/// node with placeholders for the offsets of the child nodes. Once child
/// nodes have been traversed, we know their offsets and can replace the
/// placeholder values with the actual ones.
impl SerializeCtx {
pub fn new(kind_len: u8, prop_len: u8) -> Self {
let kind_size = kind_len as usize;
let prop_size = prop_len as usize;
let mut ctx = Self {
spans: Vec::with_capacity(512),
root_idx: 0,
nodes: Vec::with_capacity(512),
prop_stack: vec![vec![]],
prev_sibling_stack: vec![0],
field_count: vec![0],
field_buf: Vec::with_capacity(1024),
str_table: StringTable::new(),
kind_name_map: vec![0; kind_size],
prop_name_map: vec![0; prop_size],
};
let empty_str = ctx.str_table.insert("");
// Placeholder node is always 0
ctx.append_node(0, &DUMMY_SP);
ctx.kind_name_map[0] = empty_str;
ctx.kind_name_map[1] = empty_str;
// Insert default props that are always present
let type_str = ctx.str_table.insert("type");
let parent_str = ctx.str_table.insert("parent");
let range_str = ctx.str_table.insert("range");
let length_str = ctx.str_table.insert("length");
// These values are expected to be in this order on the JS side
ctx.prop_name_map[0] = empty_str;
ctx.prop_name_map[1] = type_str;
ctx.prop_name_map[2] = parent_str;
ctx.prop_name_map[3] = range_str;
ctx.prop_name_map[4] = length_str;
ctx
}
pub fn set_root_idx(&mut self, idx: Index) {
self.root_idx = idx;
}
pub fn map_utf8_spans_to_utf16(&mut self, map: &Utf16Map) {
for value in &mut self.spans {
*value = map
.utf8_to_utf16_offset((*value).into())
.unwrap_or_else(|| panic!("Failed converting '{value}' to utf16."))
.into();
}
}
/// Allocate a node's header
fn field_header<P>(&mut self, prop: P, prop_flags: PropFlags)
where
P: Into<u8> + Display + Clone,
{
let flags: u8 = prop_flags.into();
let n: u8 = prop.clone().into();
if let Some(v) = self.prop_name_map.get::<usize>(n.into()) {
if *v == 0 {
let id = self.str_table.insert(&format!("{prop}"));
self.prop_name_map[n as usize] = id;
}
}
// Increment field counter
let idx = self.field_count.len() - 1;
let count = self.field_count[idx];
self.field_count[idx] = count + 1;
let buf = self.prop_stack.last_mut().unwrap();
buf.push(n);
buf.push(flags);
}
fn get_node(&mut self, id: Index) -> &mut Node {
self.nodes.get_mut(id as usize).unwrap()
}
fn set_parent(&mut self, child_id: Index, parent_id: Index) {
let child = self.get_node(child_id);
child.parent = parent_id;
}
fn set_child(&mut self, parent_id: Index, child_id: Index) {
let parent = self.get_node(parent_id);
parent.child = child_id;
}
fn set_next(&mut self, node_id: Index, next_id: Index) {
let node = self.get_node(node_id);
node.next = next_id;
}
fn update_ref_links(&mut self, parent_id: Index, child_id: Index) {
let last_idx = self.prev_sibling_stack.len() - 1;
let parent = self.get_node(parent_id);
if parent.child == 0 {
parent.child = child_id;
} else {
let prev_id = self.prev_sibling_stack[last_idx];
self.set_next(prev_id, child_id);
}
self.prev_sibling_stack[last_idx] = child_id;
self.set_parent(child_id, parent_id);
}
pub fn append_node<K>(&mut self, kind: K, span: &Span) -> PendingRef
where
K: Into<u8> + Display + Clone,
{
let (start, end) = if *span == DUMMY_SP {
(0, 0)
} else {
// -1 is because swc stores spans 1-indexed
(span.lo.0 - 1, span.hi.0 - 1)
};
self.append_inner(kind, start, end)
}
pub fn append_inner<K>(
&mut self,
kind: K,
span_lo: u32,
span_hi: u32,
) -> PendingRef
where
K: Into<u8> + Display + Clone,
{
let kind_u8: u8 = kind.clone().into();
let id: Index = self.nodes.len() as u32;
self.nodes.push(Node {
kind: kind_u8,
prop_offset: 0,
child: 0,
next: 0,
parent: 0,
});
if let Some(v) = self.kind_name_map.get::<usize>(kind_u8.into()) {
if *v == 0 {
let s_id = self.str_table.insert(&format!("{kind}"));
self.kind_name_map[kind_u8 as usize] = s_id;
}
}
self.field_count.push(0);
self.prop_stack.push(vec![]);
self.prev_sibling_stack.push(0);
// write spans
self.spans.push(span_lo);
self.spans.push(span_hi);
PendingRef(id)
}
pub fn commit_node(&mut self, id: PendingRef) -> NodeRef {
let mut buf = self.prop_stack.pop().unwrap();
let count = self.field_count.pop().unwrap();
let offset = self.field_buf.len();
// All nodes have <10 fields
self.field_buf.push(count as u8);
self.field_buf.append(&mut buf);
let node = self.nodes.get_mut(id.0 as usize).unwrap();
node.prop_offset = offset as u32;
self.prev_sibling_stack.pop();
NodeRef(id.0)
}
// Allocate an object field
pub fn open_obj(&mut self) {
self.field_count.push(0);
self.prop_stack.push(vec![]);
}
pub fn commit_obj<P>(&mut self, prop: P)
where
P: Into<u8> + Display + Clone,
{
let mut buf = self.prop_stack.pop().unwrap();
let count = self.field_count.pop().unwrap();
let offset = self.field_buf.len();
append_usize(&mut self.field_buf, count);
self.field_buf.append(&mut buf);
self.field_header(prop, PropFlags::Object);
let buf = self.prop_stack.last_mut().unwrap();
append_usize(buf, offset);
}
/// Allocate an null field
pub fn write_null<P>(&mut self, prop: P)
where
P: Into<u8> + Display + Clone,
{
self.field_header(prop, PropFlags::Null);
let buf = self.prop_stack.last_mut().unwrap();
append_u32(buf, 0);
}
/// Allocate an null field
pub fn write_undefined<P>(&mut self, prop: P)
where
P: Into<u8> + Display + Clone,
{
self.field_header(prop, PropFlags::Undefined);
let buf = self.prop_stack.last_mut().unwrap();
append_u32(buf, 0);
}
/// Allocate a number field
pub fn write_num<P>(&mut self, prop: P, value: &str)
where
P: Into<u8> + Display + Clone,
{
self.field_header(prop, PropFlags::Number);
let id = self.str_table.insert(value);
let buf = self.prop_stack.last_mut().unwrap();
append_usize(buf, id);
}
/// Allocate a bigint field
pub fn write_bigint<P>(&mut self, prop: P, value: &str)
where
P: Into<u8> + Display + Clone,
{
self.field_header(prop, PropFlags::BigInt);
let id = self.str_table.insert(value);
let buf = self.prop_stack.last_mut().unwrap();
append_usize(buf, id);
}
/// Allocate a RegExp field
pub fn write_regex<P>(&mut self, prop: P, value: &str)
where
P: Into<u8> + Display + Clone,
{
self.field_header(prop, PropFlags::Regex);
let id = self.str_table.insert(value);
let buf = self.prop_stack.last_mut().unwrap();
append_usize(buf, id);
}
/// Store the string in our string table and save the id of the string
/// in the current field.
pub fn write_str<P>(&mut self, prop: P, value: &str)
where
P: Into<u8> + Display + Clone,
{
self.field_header(prop, PropFlags::String);
let id = self.str_table.insert(value);
let buf = self.prop_stack.last_mut().unwrap();
append_usize(buf, id);
}
/// Write a bool to a field.
pub fn write_bool<P>(&mut self, prop: P, value: bool)
where
P: Into<u8> + Display + Clone,
{
self.field_header(prop, PropFlags::Bool);
let n = if value { 1 } else { 0 };
let buf = self.prop_stack.last_mut().unwrap();
append_u32(buf, n);
}
/// Replace the placeholder of a reference field with the actual offset
/// to the node we want to point to.
pub fn write_ref<P>(&mut self, prop: P, parent: &PendingRef, value: NodeRef)
where
P: Into<u8> + Display + Clone,
{
self.field_header(prop, PropFlags::Ref);
let buf = self.prop_stack.last_mut().unwrap();
append_u32(buf, value.0);
if parent.0 > 0 {
self.update_ref_links(parent.0, value.0);
}
}
/// Helper for writing optional node offsets
pub fn write_maybe_ref<P>(
&mut self,
prop: P,
parent: &PendingRef,
value: Option<NodeRef>,
) where
P: Into<u8> + Display + Clone,
{
if let Some(v) = value {
self.write_ref(prop, parent, v);
} else {
self.write_null(prop);
};
}
/// Write a vec of node offsets into the property. The necessary space
/// has been reserved earlier.
pub fn write_ref_vec<P>(
&mut self,
prop: P,
parent_ref: &PendingRef,
value: Vec<NodeRef>,
) where
P: Into<u8> + Display + Clone,
{
self.field_header(prop, PropFlags::RefArr);
let group_id = self.append_node(GROUP_KIND, &DUMMY_SP);
let group_id = self.commit_node(group_id).0;
let buf = self.prop_stack.last_mut().unwrap();
append_u32(buf, group_id);
self.update_ref_links(parent_ref.0, group_id);
let mut prev_id = 0;
for (i, item) in value.iter().enumerate() {
self.set_parent(item.0, group_id);
if i == 0 {
self.set_child(group_id, item.0);
} else {
self.set_next(prev_id, item.0);
}
prev_id = item.0;
}
}
/// Serialize all information we have into a buffer that can be sent to JS.
/// It has the following structure:
///
/// <...ast>
/// <string table>
/// <node kind map> <- node kind id maps to string id
/// <node prop map> <- node property id maps to string id
/// <spans> <- List of spans, rarely needed
/// <offset spans>
/// <offset kind map>
/// <offset prop map>
/// <offset str table>
pub fn serialize(&mut self) -> Vec<u8> {
let mut buf: Vec<u8> = vec![];
// The buffer starts with the serialized AST first, because that
// contains absolute offsets. By butting this at the start of the
// message we don't have to waste time updating any offsets.
for node in &self.nodes {
buf.push(node.kind);
append_u32(&mut buf, node.prop_offset);
append_u32(&mut buf, node.child);
append_u32(&mut buf, node.next);
append_u32(&mut buf, node.parent);
}
// Next follows the string table. We'll keep track of the offset
// in the message of where the string table begins
let offset_str_table = buf.len();
// Serialize string table
buf.append(&mut self.str_table.serialize());
// Next, serialize the mappings of kind -> string of encountered
// nodes in the AST. We use this additional lookup table to compress
// the message so that we can save space by using a u8 . All nodes of
// JS, TS and JSX together are <200
let offset_kind_map = buf.len();
// Write the total number of entries in the kind -> str mapping table
// TODO: make this a u8
append_usize(&mut buf, self.kind_name_map.len());
for v in &self.kind_name_map {
append_usize(&mut buf, *v);
}
// Store offset to prop -> string map. It's the same as with node kind
// as the total number of properties is <120 which allows us to store it
// as u8.
let offset_prop_map = buf.len();
// Write the total number of entries in the kind -> str mapping table
append_usize(&mut buf, self.prop_name_map.len());
for v in &self.prop_name_map {
append_usize(&mut buf, *v);
}
// Spans are rarely needed, so they're stored in a separate array.
// They're indexed by the node id.
let offset_spans = buf.len();
for v in &self.spans {
append_u32(&mut buf, *v);
}
// The field value table. They're detached from nodes as they're not
// as frequently needed as the nodes themselves. The most common
// operation is traversal and we can traverse nodes without knowing
// about the fields.
let offset_props = buf.len();
buf.append(&mut self.field_buf);
// Putting offsets of relevant parts of the buffer at the end. This
// allows us to hop to the relevant part by merely looking at the last
// for values in the message. Each value represents an offset into the
// buffer.
append_usize(&mut buf, offset_props);
append_usize(&mut buf, offset_spans);
append_usize(&mut buf, offset_kind_map);
append_usize(&mut buf, offset_prop_map);
append_usize(&mut buf, offset_str_table);
append_u32(&mut buf, self.root_idx);
buf
}
}