perf: Use FxHash (#1224)

This PR introduces the `FxHashMap`, `FxHashSet`, `FxIndexMap` and `FxIndexSet` types which use the non-cryptographic [fxhash](https://github.com/rust-lang/rustc-hash) algorithm also used in the Rust compiler. Doing so improves the `plc check` times by ~300 to 500 ms in a bigger internal project of ours.

Cargo.toml

@@ -42,13 +42,14 @@ lazy_static.workspace = true
 serde_json.workspace = true
 serde.workspace = true
 toml.workspace = true
+rustc-hash.workspace = true
 
 [dev-dependencies]
 num = "0.4"
 insta = "1.31.0"
 pretty_assertions = "1.3.0"
 driver = { path = "./compiler/plc_driver/", package = "plc_driver" }
-project = { path = "./compiler/plc_project/", package = "plc_project", features = ["integration"]}
+project = { path = "./compiler/plc_project/", package = "plc_project", features = ["integration"] }
 plc_xml = { path = "./compiler/plc_xml" }
 serial_test = "*"
 tempfile = "3"
@@ -90,3 +91,4 @@ lazy_static = "1.4.0"
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1"
 toml = "0.5"
+rustc-hash = "1.1.0"

compiler/plc_ast/Cargo.toml

@@ -7,6 +7,7 @@ edition = "2021"
 
 [dependencies]
 plc_util = { path = "../plc_util" }
-plc_source = {path = "../plc_source"}
+plc_source = { path = "../plc_source" }
 chrono = { version = "0.4", default-features = false }
 serde = { version = "1.0", features = ["derive"] }
+rustc-hash.workspace = true

compiler/plc_ast/src/pre_processor.rs

@@ -1,6 +1,6 @@
 // Copyright (c) 2020 Ghaith Hachem and Mathias Rieder
 
-use std::collections::HashMap;
+use rustc_hash::FxHashMap;
 
 use plc_util::convention::internal_type_name;
 
@@ -176,7 +176,7 @@ fn build_enum_initializer(
 }
 
 fn preprocess_generic_structs(pou: &mut Pou) -> Vec<UserTypeDeclaration> {
-    let mut generic_types = HashMap::new();
+    let mut generic_types = FxHashMap::default();
     let mut types = vec![];
     for binding in &pou.generics {
         let new_name = format!("__{}__{}", pou.name, binding.name); // TODO: Naming convention (see plc_util/src/convention.rs)
@@ -269,7 +269,7 @@ fn add_nested_datatypes(
     }
 }
 
-fn replace_generic_type_name(dt: &mut DataTypeDeclaration, generics: &HashMap<String, String>) {
+fn replace_generic_type_name(dt: &mut DataTypeDeclaration, generics: &FxHashMap<String, String>) {
     match dt {
         DataTypeDeclaration::DataTypeDefinition { data_type, .. } => match data_type {
             DataType::ArrayType { referenced_type, .. }

compiler/plc_diagnostics/Cargo.toml

@@ -15,3 +15,4 @@ toml.workspace = true
 anyhow.workspace = true
 lazy_static.workspace = true
 log.workspace = true
+rustc-hash.workspace = true

compiler/plc_diagnostics/src/diagnostician.rs

@@ -1,4 +1,4 @@
-use std::collections::HashMap;
+use rustc_hash::FxHashMap;
 
 use crate::{
     diagnostics::{
@@ -16,7 +16,7 @@ use crate::{
 pub struct Diagnostician {
     reporter: Box<dyn DiagnosticReporter>,
     assessor: Box<dyn DiagnosticAssessor>,
-    filename_fileid_mapping: HashMap<String, usize>,
+    filename_fileid_mapping: FxHashMap<String, usize>,
 }
 
 impl Diagnostician {
@@ -75,7 +75,7 @@ impl Diagnostician {
         Diagnostician {
             assessor: Box::<DiagnosticsRegistry>::default(),
             reporter: Box::<NullDiagnosticReporter>::default(),
-            filename_fileid_mapping: HashMap::new(),
+            filename_fileid_mapping: FxHashMap::default(),
         }
     }
 
@@ -84,7 +84,7 @@ impl Diagnostician {
         Diagnostician {
             assessor: Box::<DiagnosticsRegistry>::default(),
             reporter: Box::new(CodeSpanDiagnosticReporter::buffered()),
-            filename_fileid_mapping: HashMap::new(),
+            filename_fileid_mapping: FxHashMap::default(),
         }
     }
 
@@ -93,7 +93,7 @@ impl Diagnostician {
         Diagnostician {
             reporter: Box::<ClangFormatDiagnosticReporter>::default(),
             assessor: Box::<DiagnosticsRegistry>::default(),
-            filename_fileid_mapping: HashMap::new(),
+            filename_fileid_mapping: FxHashMap::default(),
         }
     }
 
@@ -151,7 +151,7 @@ impl Default for Diagnostician {
         Self {
             reporter: Box::<CodeSpanDiagnosticReporter>::default(),
             assessor: Box::<DiagnosticsRegistry>::default(),
-            filename_fileid_mapping: HashMap::new(),
+            filename_fileid_mapping: FxHashMap::default(),
         }
     }
 }

compiler/plc_diagnostics/src/diagnostics/diagnostics_registry.rs

@@ -1,7 +1,6 @@
 /// Returns a diagnostics map with the error code, default severity and a description
-use std::collections::HashMap;
-
 use lazy_static::lazy_static;
+use rustc_hash::FxHashMap;
 use serde::{Deserialize, Serialize};
 
 use crate::diagnostician::DiagnosticAssessor;
@@ -13,7 +12,7 @@ use super::{
 
 macro_rules! add_diagnostic {
     ($($number:ident, $severity:expr, $desc:expr,)*) => {
-        { let mut m : HashMap<&str, DiagnosticEntry> = HashMap::new();
+        { let mut m : FxHashMap<&str, DiagnosticEntry> = FxHashMap::default();
             $(
                 {
                     let code = stringify!($number);
@@ -24,7 +23,7 @@ macro_rules! add_diagnostic {
         }}
 }
 
-pub struct DiagnosticsRegistry(HashMap<&'static str, DiagnosticEntry>);
+pub struct DiagnosticsRegistry(FxHashMap<&'static str, DiagnosticEntry>);
 
 #[derive(Default, Clone, Copy, PartialEq, Eq, PartialOrd, Ord)]
 pub struct DiagnosticEntry {
@@ -41,7 +40,7 @@ impl Default for DiagnosticsRegistry {
 
 impl DiagnosticsRegistry {
     /// Creates an empty registry.
-    fn new(map: HashMap<&'static str, DiagnosticEntry>) -> Self {
+    fn new(map: FxHashMap<&'static str, DiagnosticEntry>) -> Self {
         DiagnosticsRegistry(map)
     }
 
@@ -87,7 +86,7 @@ impl DiagnosticAssessor for DiagnosticsRegistry {
 
 #[derive(Serialize, Deserialize, PartialEq, Eq, Default)]
 #[serde(transparent)]
-pub struct DiagnosticsConfiguration(HashMap<Severity, Vec<String>>);
+pub struct DiagnosticsConfiguration(FxHashMap<Severity, Vec<String>>);
 
 impl From<&DiagnosticsRegistry> for DiagnosticsConfiguration {
     fn from(registry: &DiagnosticsRegistry) -> Self {
@@ -101,7 +100,7 @@ impl From<&DiagnosticsRegistry> for DiagnosticsConfiguration {
 }
 
 lazy_static! {
-    static ref DIAGNOSTICS: HashMap<&'static str, DiagnosticEntry> = add_diagnostic!(
+    static ref DIAGNOSTICS: FxHashMap<&'static str, DiagnosticEntry> = add_diagnostic!(
     E001,
     Error,
     include_str!("./error_codes/E001.md"), //General Error

compiler/plc_driver/src/pipelines.rs

@@ -10,7 +10,8 @@ use ast::{
     ast::{pre_process, CompilationUnit, LinkageType},
     provider::IdProvider,
 };
-use indexmap::IndexSet;
+
+use plc::index::FxIndexSet;
 use plc::{
     codegen::{CodegenContext, GeneratedModule},
     index::Index,
@@ -192,7 +193,7 @@ impl<T: SourceContainer + Sync> IndexedProject<T> {
 /// A project that has been annotated with information about different types and used units
 pub struct AnnotatedProject<T: SourceContainer + Sync> {
     pub project: Project<T>,
-    pub units: Vec<(CompilationUnit, IndexSet<Dependency>, StringLiterals)>,
+    pub units: Vec<(CompilationUnit, FxIndexSet<Dependency>, StringLiterals)>,
     pub index: Index,
     pub annotations: AstAnnotations,
 }
@@ -266,7 +267,7 @@ impl<T: SourceContainer + Sync> AnnotatedProject<T> {
         context: &'ctx CodegenContext,
         compile_options: &CompileOptions,
         unit: &CompilationUnit,
-        dependencies: &IndexSet<Dependency>,
+        dependencies: &FxIndexSet<Dependency>,
         literals: &StringLiterals,
     ) -> Result<GeneratedModule<'ctx>, Diagnostic> {
         let mut code_generator = plc::codegen::CodeGen::new(

compiler/plc_index/Cargo.toml

@@ -10,5 +10,6 @@ plc_ast = { path = "../plc_ast" }
 plc_source = { path = "../plc_source" }
 plc_diagnostics = { path = "../plc_diagnostics" }
 #plc_project = { path = "../plc_project" } # TODO: This will create a circular dependency
+rustc-hash.workspace = true
 
 encoding_rs = "0.8"

compiler/plc_index/src/lib.rs

@@ -3,13 +3,13 @@ use plc_ast::provider::IdProvider;
 use plc_diagnostics::diagnostics::Diagnostic;
 use plc_source::source_location::SourceLocation;
 use plc_source::{SourceCode, SourceContainer};
-use std::collections::HashMap;
+use rustc_hash::FxHashMap;
 
 #[derive(Debug, Default)]
 pub struct GlobalContext {
     /// HashMap containing all read, i.e. parsed, sources where the key represents
     /// the relative file path and the value some [`SourceCode`]
-    sources: HashMap<&'static str, SourceCode>,
+    sources: FxHashMap<&'static str, SourceCode>,
 
     /// [`IdProvider`] used during the parsing session
     provider: IdProvider,
@@ -24,7 +24,7 @@ pub struct GlobalContext {
 
 impl GlobalContext {
     pub fn new() -> Self {
-        Self { sources: HashMap::new(), provider: IdProvider::default() }
+        Self { sources: FxHashMap::default(), provider: IdProvider::default() }
     }
 
     /// Inserts a single [`SourceCode`] to the internal source map

compiler/plc_xml/Cargo.toml

@@ -16,6 +16,7 @@ ast = { path = "../plc_ast/", package = "plc_ast" }
 plc_diagnostics = { path = "../plc_diagnostics" }
 plc_source = { path = "../plc_source" }
 itertools.workspace = true
+rustc-hash.workspace = true
 
 [features]
 default = []

compiler/plc_xml/src/extensions.rs

@@ -1,8 +1,9 @@
 //! In this file extension traits are defined to be used within the `plc_xml` crate.
 
-use std::{borrow::Cow, collections::HashMap};
+use std::borrow::Cow;
 
 use quick_xml::name::QName;
+use rustc_hash::FxHashMap;
 
 use crate::error::Error;
 
@@ -34,7 +35,7 @@ impl TryToString for Cow<'_, [u8]> {
     }
 }
 
-impl GetOrErr for HashMap<String, String> {
+impl GetOrErr for FxHashMap<String, String> {
     fn get_or_err(&self, key: &str) -> Result<String, Error> {
         self.get(key).map(|it| it.to_owned()).ok_or(Error::MissingAttribute(key.to_string()))
     }

compiler/plc_xml/src/model/block.rs

@@ -1,6 +1,7 @@
-use std::{borrow::Cow, collections::HashMap};
+use std::borrow::Cow;
 
 use quick_xml::events::{BytesStart, Event};
+use rustc_hash::FxHashMap;
 
 use crate::{
     error::Error,
@@ -21,7 +22,7 @@ pub(crate) struct Block<'xml> {
 }
 
 impl<'xml> Block<'xml> {
-    pub fn new(mut hm: HashMap<String, String>, variables: Vec<BlockVariable>) -> Result<Self, Error> {
+    pub fn new(mut hm: FxHashMap<String, String>, variables: Vec<BlockVariable>) -> Result<Self, Error> {
         Ok(Self {
             local_id: hm.get_or_err("localId").map(|it| it.parse())??,
             type_name: Cow::from(hm.get_or_err("typeName")?),

compiler/plc_xml/src/model/connector.rs

@@ -1,6 +1,7 @@
-use std::{borrow::Cow, collections::HashMap};
+use std::borrow::Cow;
 
 use quick_xml::events::Event;
+use rustc_hash::FxHashMap;
 
 use crate::{
     error::Error,
@@ -19,7 +20,7 @@ pub(crate) struct Connector<'xml> {
 }
 
 impl<'xml> Connector<'xml> {
-    pub fn new(mut hm: HashMap<String, String>, kind: ConnectorKind) -> Result<Self, Error> {
+    pub fn new(mut hm: FxHashMap<String, String>, kind: ConnectorKind) -> Result<Self, Error> {
         Ok(Self {
             kind,
             name: Cow::from(hm.get_or_err("name")?),

compiler/plc_xml/src/model/control.rs

@@ -1,6 +1,7 @@
-use std::{borrow::Cow, collections::HashMap, str::FromStr};
+use std::{borrow::Cow, str::FromStr};
 
 use quick_xml::events::Event;
+use rustc_hash::FxHashMap;
 
 use crate::{
     error::Error,
@@ -20,7 +21,7 @@ pub(crate) struct Control<'xml> {
 }
 
 impl<'xml> Control<'xml> {
-    pub fn new(mut hm: HashMap<String, String>, kind: ControlKind) -> Result<Self, Error> {
+    pub fn new(mut hm: FxHashMap<String, String>, kind: ControlKind) -> Result<Self, Error> {
         Ok(Self {
             kind,
             name: hm.remove("label").map(Cow::from),