This commit is contained in:
Wojciech Danilo 2022-03-10 04:47:00 +01:00
parent 58fe42a20c
commit 2427605320
4 changed files with 620 additions and 0 deletions

8
Cargo.lock generated
View File

@ -944,6 +944,14 @@ dependencies = [
"wasm-bindgen-test",
]
[[package]]
name = "enso-formatter"
version = "0.1.0"
dependencies = [
"lazy_static",
"regex",
]
[[package]]
name = "enso-frp"
version = "0.1.0"

View File

@ -5,6 +5,7 @@
# where plausible.
members = [
"app/gui",
"build/enso-formatter",
"build/rust-scripts",
"lib/rust/*",
"lib/rust/profiler/data",

View File

@ -0,0 +1,10 @@
[package]
name = "enso-formatter"
version = "0.1.0"
authors = ["Enso Team <contact@enso.org>"]
edition = "2021"
[dependencies]
regex = "1"
lazy_static = "1.4.0"

View File

@ -0,0 +1,601 @@
//! This crate implements code formatter rules that are not implemented in rustfmt. These rules
//! are this codebase specific, and they may not be desired in other code bases, including:
//! - Sorting imports into groups (e.g. local imports, pub imports, etc.).
//! - Sorting module attributes into groups.
//! - Adding standard lint configuration to `lib.rs` and `main.rs` files.
//! - (Currently disabled) Emitting warnings about star imports that are not ending with `traits::*`
//! nor `prelude::*`.
//!
//! Possible extensions, not implemented yet:
//! - Sections are automatically keeping spacing.
// === Standard Linter Configuration ===
#![deny(non_ascii_idents)]
#![warn(unsafe_code)]
// === Non-Standard Linter Configuration ===
#![deny(keyword_idents)]
#![deny(macro_use_extern_crate)]
#![deny(missing_abi)]
#![deny(pointer_structural_match)]
#![deny(unsafe_op_in_unsafe_fn)]
#![deny(unconditional_recursion)]
#![warn(missing_docs)]
#![warn(absolute_paths_not_starting_with_crate)]
#![warn(elided_lifetimes_in_paths)]
#![warn(explicit_outlives_requirements)]
#![warn(missing_copy_implementations)]
#![warn(missing_debug_implementations)]
#![warn(noop_method_call)]
#![warn(single_use_lifetimes)]
#![warn(trivial_casts)]
#![warn(trivial_numeric_casts)]
#![warn(unused_crate_dependencies)]
#![warn(unused_extern_crates)]
#![warn(unused_import_braces)]
#![warn(unused_lifetimes)]
#![warn(unused_qualifications)]
#![warn(variant_size_differences)]
#![warn(unreachable_pub)]
use lazy_static::lazy_static;
use regex::Regex;
use std::collections::hash_map::DefaultHasher;
use std::collections::HashMap;
use std::ffi::OsStr;
use std::fmt::Debug;
use std::fs;
use std::hash::Hash;
use std::hash::Hasher;
use std::path::Path;
use std::path::PathBuf;
use std::process::Command;
use std::process::Stdio;
// =================
// === Constants ===
// =================
// TODO: The below lints should be uncommented, one-by-one, and the existing code should be
// adjusted.
/// Standard linter configuration. It will be used in every `main.rs` and `lib.rs` file in the
/// codebase.
const STD_LINTER_ATTRIBS: &[&str] = &[
// Rustc lints that are allowed by default:
// "warn(absolute_paths_not_starting_with_crate)",
// "warn(elided_lifetimes_in_paths)",
// "warn(explicit_outlives_requirements)",
// "deny(keyword_idents)",
// "deny(macro_use_extern_crate)",
// "deny(missing_abi)",
// "warn(missing_copy_implementations)",
// "warn(missing_debug_implementations)",
// "warn(missing_docs)",
"deny(non_ascii_idents)",
// "warn(noop_method_call)",
// "deny(pointer_structural_match)",
// "warn(single_use_lifetimes)",
// "warn(trivial_casts)",
// "warn(trivial_numeric_casts)",
"warn(unsafe_code)",
// "deny(unsafe_op_in_unsafe_fn)",
// "warn(unused_crate_dependencies)",
// "warn(unused_extern_crates)",
// "warn(unused_import_braces)",
// "warn(unused_lifetimes)",
// "warn(unused_qualifications)",
// "warn(variant_size_differences)",
// Rustc lints that emit a warning by default:
// "deny(unconditional_recursion)",
];
// =============
// === Utils ===
// =============
fn calculate_hash<T: Hash>(t: &T) -> u64 {
let mut s = DefaultHasher::new();
t.hash(&mut s);
s.finish()
}
fn read_file_with_hash(path: impl AsRef<Path>) -> std::io::Result<(u64, String)> {
fs::read_to_string(path).map(|t| (calculate_hash(&t), t))
}
// ===================
// === HeaderToken ===
// ===================
use HeaderToken::*;
/// A token that can be found in the header of a file.
#[derive(Debug, Copy, Clone, PartialEq, Eq, Hash)]
#[allow(missing_docs)]
pub enum HeaderToken {
Attrib,
ModuleAttrib,
ModuleAttribWarn,
ModuleAttribAllow,
ModuleAttribDeny,
ModuleAttribFeature,
ModuleAttribFeature2,
EmptyLine,
ModuleDoc,
Comment,
CrateUse,
CrateUseStar,
CratePubUse,
CratePubUseStar,
Use,
UseStar,
PubUse,
PubUseStar,
PubMod,
/// Special header token that is never parsed, but can be injected by the code.
StandardLinterConfig,
}
/// A header token with the matched string and possibly attached attributes.
#[derive(Clone)]
pub struct HeaderElement {
attrs: Vec<String>,
token: HeaderToken,
reg_match: String,
}
impl Debug for HeaderElement {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
write!(f, "{:?}({:?})", self.token, self.reg_match.as_str())
}
}
impl HeaderElement {
/// Constructor.
pub fn new(token: HeaderToken, reg_match: String) -> Self {
let attrs = Default::default();
Self { attrs, token, reg_match }
}
/// Check whether the element is empty.
pub fn is_empty(&self) -> bool {
self.reg_match.is_empty()
}
/// Length of the splice. Includes the length of the matched string and all attached attributes.
pub fn len(&self) -> usize {
let args_len: usize = self.attrs.iter().map(|t| t.len()).sum();
self.reg_match.len() + args_len
}
/// Convert the element to a string representation.
#[allow(clippy::inherent_to_string)]
pub fn to_string(&self) -> String {
format!("{}{}", self.attrs.join(""), self.reg_match)
}
}
/// Wrappers for [`Regex::find`] which returns [`Result::Err`] on element found. It allows combining
/// multiple calls to this function with the Rust `?` syntax.
fn find_with<T>(input: &str, regex: &Regex, f: impl FnOnce(String) -> T) -> Result<(), T> {
match regex.find(input) {
Some(t) => Err(f(t.as_str().into())),
None => Ok(()),
}
}
/// Regex constructor that starts on the beginning of a line, can be surrounded by whitespaces and
/// ends with a line break.
fn re(input: &str) -> Regex {
let str = format!(r"^ *{} *(; *)?((\r\n?)|\n)", input);
Regex::new(&str).unwrap()
}
macro_rules! define_rules {
($($name:ident = $re:tt;)*) => {
#[allow(non_upper_case_globals)]
mod static_re {
use super::*;
lazy_static! {
$(
pub static ref $name: Regex = re($re);
)*
}
}
fn match_header_internal(input: &str) -> Result<(), HeaderElement> {
$( find_with(input, &static_re::$name, |t| HeaderElement::new($name, t))?; )*
Ok(())
}
};
}
define_rules! {
EmptyLine = r"";
ModuleDoc = r"//![^\n\r]*";
Comment = r"//[^\n\r]*";
CrateUse = r"use +crate( *:: *[\w]+)*( +as +[\w]+)?";
CrateUseStar = r"use +crate( *:: *[\w*]+)*";
CratePubUse = r"pub +use +crate( *:: *[\w]+)*( +as +[\w]+)?";
CratePubUseStar = r"pub +use +crate( *:: *[\w*]+)*";
Use = r"use +[\w]+( *:: *[\w]+)*( +as +[\w]+)?";
UseStar = r"use +[\w]+( *:: *[\w*]+)*";
PubUse = r"pub +use +[\w]+( *:: *[\w]+)*( +as +[\w]+)?";
PubUseStar = r"pub +use +[\w]+( *:: *[\w*]+)*";
ModuleAttribFeature = r"#!\[feature[^\]]*\]";
ModuleAttribFeature2 = r"#!\[allow\(incomplete_features\)\]";
ModuleAttribWarn = r"#!\[warn[^\]]*\]";
ModuleAttribAllow = r"#!\[allow[^\]]*\]";
ModuleAttribDeny = r"#!\[deny[^\]]*\]";
ModuleAttrib = r"#!\[[^\]]*\]";
Attrib = r"#\[[^\]]*\]";
PubMod = r"pub +mod +[\w]+";
}
fn match_header(input: &str) -> Option<HeaderElement> {
match match_header_internal(input) {
Err(t) => Some(t),
Ok(_) => None,
}
}
// =======================
// === Pretty printing ===
// =======================
/// Prints H1 section if any of the provided tokens was used in the file being formatted.
fn print_h1(
out: &mut String,
map: &HashMap<HeaderToken, Vec<String>>,
tokens: &[HeaderToken],
str: &str,
) {
if tokens.iter().map(|tok| map.contains_key(tok)).any(|t| t) {
out.push('\n');
out.push_str(&format!("// ===={}====\n", "=".repeat(str.len())));
out.push_str(&format!("// === {} ===\n", str));
out.push_str(&format!("// ===={}====\n", "=".repeat(str.len())));
out.push('\n');
}
}
/// Prints H2 section if any of the provided tokens was used in the file being formatted.
fn print_h2(
out: &mut String,
map: &HashMap<HeaderToken, Vec<String>>,
tokens: &[HeaderToken],
str: &str,
) {
if tokens.iter().map(|tok| map.contains_key(tok)).any(|t| t) {
out.push_str(&format!("// === {} ===\n", str));
}
}
/// Prints all the entries associated with the provided tokens. If at least one entry was printed,
/// an empty line will be added in the end.
fn print(out: &mut String, map: &mut HashMap<HeaderToken, Vec<String>>, t: &[HeaderToken]) -> bool {
let sub_results: Vec<bool> = t.iter().map(|t| print_single(out, map, *t)).collect();
sub_results.iter().any(|t| *t)
}
/// Prints all the entries associated with the provided tokens. If at least one entry was printed,
/// an empty line will be added in the end.
fn print_section(out: &mut String, map: &mut HashMap<HeaderToken, Vec<String>>, t: &[HeaderToken]) {
if print(out, map, t) {
out.push('\n');
}
}
/// Print all the entries associated with the provided token.
fn print_single(
out: &mut String,
map: &mut HashMap<HeaderToken, Vec<String>>,
token: HeaderToken,
) -> bool {
match map.remove(&token) {
None => false,
Some(t) => {
out.push_str(&t.join(""));
true
}
}
}
// ==============
// === Action ===
// ==============
/// Possible commands this formatter can evaluate.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
#[allow(missing_docs)]
pub enum Action {
Format,
Preview,
FormatAndCheck,
}
// ==================
// === Processing ===
// ==================
/// Process all files of the given path recursively.
///
/// Please note that the [`hash_map`] variable contains hashes of all files before processing. After
/// the processing is done by this formatter, `rustfmt` is run on the whole codebase, and the hashes
/// are compared with new files. This allows checking if running the formatting changed the files.
/// An alternative design is possible we could run this formatter and pass its output to stdin of
/// `rustfmt`, run it in memory, and get the results without affecting files on the disk.
/// Unfortunately, such solution requires either running a separate `rustfmt` process per file, or
/// using its API. The former solution is very slow (16 seconds for the whole codebase), the second
/// uses non-documented API and is slow as well (8 seconds for the whole codebase). It should be
/// possible to improve the latter solution to get good performance, but it seems way harder than it
/// should be.
fn process_path(path: impl AsRef<Path>, action: Action) {
let paths = discover_paths(path);
let total = paths.len();
let mut hash_map = HashMap::<PathBuf, u64>::new();
for (i, sub_path) in paths.iter().enumerate() {
let file_name = sub_path.file_name().and_then(|s| s.to_str());
let is_main_file = file_name == Some("lib.rs") || file_name == Some("main.rs");
let dbg_msg = if is_main_file { " [main]" } else { "" };
println!("[{}/{}] Processing {:?}{}.", i + 1, total, sub_path, dbg_msg);
let hash = process_file(sub_path, action, is_main_file);
hash_map.insert(sub_path.into(), hash);
}
if action == Action::Format || action == Action::FormatAndCheck {
let mut child = Command::new("cargo")
.args(["fmt"])
.stdin(Stdio::null())
.stdout(Stdio::inherit())
.spawn()
.expect("'cargo fmt' failed to start.");
child.wait().unwrap();
}
if action == Action::FormatAndCheck {
let mut changed = Vec::new();
for sub_path in &paths {
let (hash, _) = read_file_with_hash(sub_path).unwrap();
if hash_map.get(sub_path) != Some(&hash) {
changed.push(sub_path.clone());
}
}
if !changed.is_empty() {
panic!("{} files changed:\n{:#?}", changed.len(), changed);
}
}
}
/// Discover all paths containing Rust sources, recursively.
fn discover_paths(path: impl AsRef<Path>) -> Vec<PathBuf> {
let mut vec = Vec::default();
discover_paths_internal(&mut vec, path);
vec
}
fn discover_paths_internal(vec: &mut Vec<PathBuf>, path: impl AsRef<Path>) {
let path = path.as_ref();
let md = fs::metadata(path).unwrap();
if md.is_dir() && path.file_name() != Some(OsStr::new("target")) {
let sub_paths = fs::read_dir(path).unwrap();
for sub_path in sub_paths {
discover_paths_internal(vec, &sub_path.unwrap().path())
}
} else if md.is_file() && path.extension() == Some(OsStr::new("rs")) {
vec.push(path.into());
}
}
fn process_file(path: impl AsRef<Path>, action: Action, is_main_file: bool) -> u64 {
let path = path.as_ref();
let (hash, input) = read_file_with_hash(path).unwrap();
match process_file_content(input, is_main_file) {
Err(e) => panic!("{:?}: {}", path, e),
Ok(out) => {
if action == Action::Preview {
println!("{}", out)
} else if action == Action::Format || action == Action::FormatAndCheck {
fs::write(path, out).expect("Unable to write back to the source file.")
}
hash
}
}
}
/// Process a single source file.
fn process_file_content(input: String, is_main_file: bool) -> Result<String, String> {
let mut str_ptr: &str = &input;
let mut attrs = vec![];
let mut header = vec![];
loop {
match match_header(str_ptr) {
None => break,
Some(mut m) => {
str_ptr = &str_ptr[m.len()..];
match m.token {
Attrib => attrs.push(m),
_ => {
if !attrs.is_empty() {
let old_attrs = std::mem::take(&mut attrs);
m.attrs = old_attrs.into_iter().map(|t| t.reg_match).collect();
}
header.push(m)
}
}
}
}
}
// Do not consume the leading comments.
let mut ending: Vec<&HeaderElement> = header
.iter()
.rev()
.take_while(|t| (t.token == Comment) || (t.token == EmptyLine))
.collect();
ending.reverse();
let incorrect_ending_len = ending.into_iter().skip_while(|t| t.token == EmptyLine).count();
header.truncate(header.len() - incorrect_ending_len);
let total_len: usize = header.iter().map(|t| t.len()).sum();
// Error if the import section contains comments.
let contains_comments =
header.iter().any(|t| t.token == Comment && !t.reg_match.starts_with("// ==="));
if contains_comments {
return Err(
"File contains comments in the import section. This is not allowed.".to_string()
);
}
// Error if the star import is used for non prelude- or traits-like imports.
// TODO: This is commented for now because it requires several non-trival changes in the code.
// let invalid_star_import = header.iter().any(|t| {
// t.token == UseStar
// && !t.reg_match.contains("prelude::*")
// && !t.reg_match.contains("traits::*")
// && !t.reg_match.contains("super::*")
// });
//
// if invalid_star_import {
// Err("Star imports only allowed for `prelude`, `traits`, and `super`
// modules.".to_string())?; }
// Build a mapping between tokens and registered entries.
let mut map = HashMap::<HeaderToken, Vec<String>>::new();
for elem in header {
map.entry(elem.token).or_default().push(elem.to_string());
}
// Remove standard linter configuration from the configuration found in the file.
if is_main_file {
let vec = map.entry(ModuleAttribAllow).or_default();
vec.retain(|t| !STD_LINTER_ATTRIBS.iter().map(|s| t.contains(s)).any(|b| b));
if vec.is_empty() {
map.remove(&ModuleAttribAllow);
}
let vec = map.entry(ModuleAttribDeny).or_default();
vec.retain(|t| !STD_LINTER_ATTRIBS.iter().map(|s| t.contains(s)).any(|b| b));
if vec.is_empty() {
map.remove(&ModuleAttribDeny);
}
let vec = map.entry(ModuleAttribWarn).or_default();
vec.retain(|t| !STD_LINTER_ATTRIBS.iter().map(|s| t.contains(s)).any(|b| b));
if vec.is_empty() {
map.remove(&ModuleAttribWarn);
}
let std_linter_attribs = STD_LINTER_ATTRIBS.iter().map(|t| format!("#![{}]\n", t));
map.entry(StandardLinterConfig).or_default().extend(std_linter_attribs);
}
// Print the results.
let mut out = String::new();
print_section(&mut out, &mut map, &[ModuleDoc]);
print_section(&mut out, &mut map, &[ModuleAttrib]);
print_h2(&mut out, &map, &[ModuleAttribFeature2, ModuleAttribFeature], "Features");
print_section(&mut out, &mut map, &[ModuleAttribFeature2, ModuleAttribFeature]);
if !STD_LINTER_ATTRIBS.is_empty() {
print_h2(&mut out, &map, &[StandardLinterConfig], "Standard Linter Configuration");
print_section(&mut out, &mut map, &[StandardLinterConfig]);
}
print_h2(
&mut out,
&map,
&[ModuleAttribAllow, ModuleAttribDeny, ModuleAttribWarn],
"Non-Standard Linter Configuration",
);
print_section(&mut out, &mut map, &[ModuleAttribAllow, ModuleAttribDeny, ModuleAttribWarn]);
print_section(&mut out, &mut map, &[CrateUseStar, UseStar]);
print_section(&mut out, &mut map, &[CrateUse]);
print_section(&mut out, &mut map, &[Use]);
print_h1(&mut out, &map, &[PubMod, CratePubUseStar, PubUseStar, CratePubUse, PubUse], "Export");
print_section(&mut out, &mut map, &[PubMod]);
print_section(&mut out, &mut map, &[CratePubUseStar, PubUseStar, CratePubUse, PubUse]);
out.push_str("\n\n");
out.push_str(&input[total_len..]);
Ok(out)
}
fn main() {
process_path(".", Action::Format);
}
// =============
// === Tests ===
// =============
#[test]
fn test_formatting() {
let input = r#"//! Module-level documentation
//! written in two lines.
#![warn(missing_copy_implementations)]
#![allow(incomplete_features)]
#![recursion_limit = "512"]
pub use lib_f::item_1;
pub mod mod1;
use crate::prelude::*;
use crate::lib_b;
use lib_c;
pub use crate::lib_e;
use crate::lib_a;
use lib_d::item_1;
use logger::traits::*;
pub mod mod2;
pub struct Struct1 {}
"#;
let output = r#"//! Module-level documentation
//! written in two lines.
#![recursion_limit = "512"]
// === Features ===
#![allow(incomplete_features)]
// === Non-Standard Linter Configuration ===
#![warn(missing_copy_implementations)]
use crate::prelude::*;
use logger::traits::*;
use crate::lib_b;
use crate::lib_a;
use lib_c;
use lib_d::item_1;
// ==============
// === Export ===
// ==============
pub mod mod1;
pub mod mod2;
pub use crate::lib_e;
pub use lib_f::item_1;
pub struct Struct1 {}
"#;
assert_eq!(process_file_content(input.into(), true), Ok(output.into()));
}