Add unicode feature to switch between regex and regex-lite crates as a trade-off between full unicode support and binary size (#3291)

* - Add `unicode` feature to switch between `regex` and `regex-lite`

as a trade-off between full unicode support and binary size.

* Update CHANGES.md

* Update CHANGES.md

* refactor: move regexset code selection to own module

* docs: add docs within RegexSet module

* chore: restore manifests

* test: ensure all actix-router codepaths are tested

---------

Co-authored-by: Rob Ede <robjtede@icloud.com>

.github/workflows/ci.yml

@@ -70,6 +70,7 @@ jobs:
         shell: bash
         run: |
           set -e
+          cargo test --lib --tests -p=actix-router --no-default-features
           cargo test --lib --tests -p=actix-router --all-features
           cargo test --lib --tests -p=actix-http --all-features
           cargo test --lib --tests -p=actix-web --features=rustls-0_20,rustls-0_21,rustls-0_22,openssl -- --skip=test_reading_deflate_encoding_large_random_rustls

actix-router/CHANGES.md

@@ -2,6 +2,7 @@
 
 ## Unreleased
 
+- Add `unicode` crate feature (on-by-default) to switch between `regex` and `regex-lite` as a trade-off between full unicode support and binary size.
 - Minimum supported Rust version (MSRV) is now 1.72.
 
 ## 0.5.2

actix-router/Cargo.toml

@@ -17,12 +17,16 @@ name = "actix_router"
 path = "src/lib.rs"
 
 [features]
-default = ["http"]
+default = ["http", "unicode"]
+http = ["dep:http"]
+unicode = ["dep:regex"]
 
 [dependencies]
 bytestring = ">=0.1.5, <2"
+cfg-if = "1"
 http = { version = "0.2.7", optional = true }
-regex = "1.5"
+regex = { version = "1.5", optional = true }
+regex-lite = "0.1"
 serde = "1"
 tracing = { version = "0.1.30", default-features = false, features = ["log"] }
 
@@ -35,6 +39,7 @@ percent-encoding = "2.1"
 [[bench]]
 name = "router"
 harness = false
+required-features = ["unicode"]
 
 [[bench]]
 name = "quoter"

actix-router/src/lib.rs

@@ -10,6 +10,7 @@ mod de;
 mod path;
 mod pattern;
 mod quoter;
+mod regex_set;
 mod resource;
 mod resource_path;
 mod router;

actix-router/src/regex_set.rs

@@ -0,0 +1,66 @@
+//! Abstraction over `regex` and `regex-lite` depending on whether we have `unicode` crate feature
+//! enabled.
+
+use cfg_if::cfg_if;
+#[cfg(feature = "unicode")]
+pub(crate) use regex::{escape, Regex};
+#[cfg(not(feature = "unicode"))]
+pub(crate) use regex_lite::{escape, Regex};
+
+#[cfg(feature = "unicode")]
+#[derive(Debug, Clone)]
+pub(crate) struct RegexSet(regex::RegexSet);
+
+#[cfg(not(feature = "unicode"))]
+#[derive(Debug, Clone)]
+pub(crate) struct RegexSet(Vec<regex_lite::Regex>);
+
+impl RegexSet {
+    /// Create a new regex set.
+    ///
+    /// # Panics
+    ///
+    /// Panics if any path patterns are malformed.
+    pub(crate) fn new(re_set: Vec<String>) -> Self {
+        cfg_if! {
+            if #[cfg(feature = "unicode")] {
+                Self(regex::RegexSet::new(re_set).unwrap())
+            } else {
+                Self(re_set.iter().map(|re| Regex::new(re).unwrap()).collect())
+            }
+        }
+    }
+
+    /// Create a new empty regex set.
+    pub(crate) fn empty() -> Self {
+        cfg_if! {
+            if #[cfg(feature = "unicode")] {
+                Self(regex::RegexSet::empty())
+            } else {
+                Self(Vec::new())
+            }
+        }
+    }
+
+    /// Returns true if regex set matches `path`.
+    pub(crate) fn is_match(&self, path: &str) -> bool {
+        cfg_if! {
+            if #[cfg(feature = "unicode")] {
+                self.0.is_match(path)
+            } else {
+                self.0.iter().any(|re| re.is_match(path))
+            }
+        }
+    }
+
+    /// Returns index within `path` of first match.
+    pub(crate) fn first_match_idx(&self, path: &str) -> Option<usize> {
+        cfg_if! {
+            if #[cfg(feature = "unicode")] {
+                self.0.matches(path).into_iter().next()
+            } else {
+                Some(self.0.iter().enumerate().find(|(_, re)| re.is_match(path))?.0)
+            }
+        }
+    }
+}

actix-router/src/resource.rs

@@ -5,10 +5,13 @@ use std::{
     mem,
 };
 
-use regex::{escape, Regex, RegexSet};
 use tracing::error;
 
-use crate::{path::PathItem, IntoPatterns, Patterns, Resource, ResourcePath};
+use crate::{
+    path::PathItem,
+    regex_set::{escape, Regex, RegexSet},
+    IntoPatterns, Patterns, Resource, ResourcePath,
+};
 
 const MAX_DYNAMIC_SEGMENTS: usize = 16;
 
@@ -233,7 +236,7 @@ enum PatternSegment {
     Var(String),
 }
 
-#[derive(Clone, Debug)]
+#[derive(Debug, Clone)]
 #[allow(clippy::large_enum_variant)]
 enum PatternType {
     /// Single constant/literal segment.
@@ -603,7 +606,7 @@ impl ResourceDef {
             PatternType::Dynamic(re, _) => Some(re.captures(path)?[1].len()),
 
             PatternType::DynamicSet(re, params) => {
-                let idx = re.matches(path).into_iter().next()?;
+                let idx = re.first_match_idx(path)?;
                 let (ref pattern, _) = params[idx];
                 Some(pattern.captures(path)?[1].len())
             }
@@ -706,7 +709,7 @@ impl ResourceDef {
 
             PatternType::DynamicSet(re, params) => {
                 let path = path.unprocessed();
-                let (pattern, names) = match re.matches(path).into_iter().next() {
+                let (pattern, names) = match re.first_match_idx(path) {
                     Some(idx) => &params[idx],
                     _ => return false,
                 };
@@ -870,7 +873,7 @@ impl ResourceDef {
                     }
                 }
 
-                let pattern_re_set = RegexSet::new(re_set).unwrap();
+                let pattern_re_set = RegexSet::new(re_set);
                 let segments = segments.unwrap_or_default();
 
                 (

actix-web/CHANGES.md

@@ -4,6 +4,7 @@
 
 ### Changed
 
+- Add `unicode` crate feature (on-by-default) to switch between `regex` and `regex-lite` as a trade-off between full unicode support and binary size.
 - Minimum supported Rust version (MSRV) is now 1.72.
 
 ## 4.5.1

actix-web/Cargo.toml

@@ -40,7 +40,7 @@ name = "actix_web"
 path = "src/lib.rs"
 
 [features]
-default = ["macros", "compress-brotli", "compress-gzip", "compress-zstd", "cookies", "http2"]
+default = ["macros", "compress-brotli", "compress-gzip", "compress-zstd", "cookies", "http2", "unicode"]
 
 # Brotli algorithm content-encoding support
 compress-brotli = ["actix-http/compress-brotli", "__compress"]
@@ -72,6 +72,9 @@ rustls-0_21 = ["http2", "actix-http/rustls-0_21", "actix-tls/accept", "actix-tls
 # TLS via Rustls v0.22
 rustls-0_22 = ["http2", "actix-http/rustls-0_22", "actix-tls/accept", "actix-tls/rustls-0_22"]
 
+# Full unicode support
+unicode = ["dep:regex", "actix-router/unicode"]
+
 # Internal (PRIVATE!) features used to aid testing and checking feature status.
 # Don't rely on these whatsoever. They may disappear at anytime.
 __compress = []
@@ -89,7 +92,7 @@ actix-utils = "3"
 actix-tls = { version = "3.3", default-features = false, optional = true }
 
 actix-http = { version = "3.6", features = ["ws"] }
-actix-router = "0.5"
+actix-router = { version = "0.5", default-features = false, features = ["http"] }
 actix-web-codegen = { version = "4.2", optional = true }
 
 ahash = "0.8"
@@ -107,7 +110,8 @@ log = "0.4"
 mime = "0.3"
 once_cell = "1.5"
 pin-project-lite = "0.2.7"
-regex = "1.5.5"
+regex = { version = "1.5.5", optional = true }
+regex-lite = "0.1"
 serde = "1.0"
 serde_json = "1.0"
 serde_urlencoded = "0.7"

actix-web/src/http/header/content_disposition.rs

@@ -13,7 +13,10 @@
 use std::fmt::{self, Write};
 
 use once_cell::sync::Lazy;
+#[cfg(feature = "unicode")]
 use regex::Regex;
+#[cfg(not(feature = "unicode"))]
+use regex_lite::Regex;
 
 use super::{ExtendedValue, Header, TryIntoHeaderValue, Writer};
 use crate::http::header;

actix-web/src/middleware/logger.rs

@@ -18,7 +18,10 @@ use bytes::Bytes;
 use futures_core::ready;
 use log::{debug, warn};
 use pin_project_lite::pin_project;
-use regex::{Regex, RegexSet};
+#[cfg(feature = "unicode")]
+use regex::Regex;
+#[cfg(not(feature = "unicode"))]
+use regex_lite::Regex;
 use time::{format_description::well_known::Rfc3339, OffsetDateTime};
 
 use crate::{
@@ -87,7 +90,7 @@ pub struct Logger(Rc<Inner>);
 struct Inner {
     format: Format,
     exclude: HashSet<String>,
-    exclude_regex: RegexSet,
+    exclude_regex: Vec<Regex>,
     log_target: Cow<'static, str>,
 }
 
@@ -97,7 +100,7 @@ impl Logger {
         Logger(Rc::new(Inner {
             format: Format::new(format),
             exclude: HashSet::new(),
-            exclude_regex: RegexSet::empty(),
+            exclude_regex: Vec::new(),
             log_target: Cow::Borrowed(module_path!()),
         }))
     }
@@ -114,10 +117,7 @@ impl Logger {
     /// Ignore and do not log access info for paths that match regex.
     pub fn exclude_regex<T: Into<String>>(mut self, path: T) -> Self {
         let inner = Rc::get_mut(&mut self.0).unwrap();
-        let mut patterns = inner.exclude_regex.patterns().to_vec();
+        inner.exclude_regex.push(Regex::new(&path.into()).unwrap());
-        patterns.push(path.into());
-        let regex_set = RegexSet::new(patterns).unwrap();
-        inner.exclude_regex = regex_set;
         self
     }
 
@@ -240,7 +240,7 @@ impl Default for Logger {
         Logger(Rc::new(Inner {
             format: Format::default(),
             exclude: HashSet::new(),
-            exclude_regex: RegexSet::empty(),
+            exclude_regex: Vec::new(),
             log_target: Cow::Borrowed(module_path!()),
         }))
     }
@@ -300,7 +300,11 @@ where
 
     fn call(&self, req: ServiceRequest) -> Self::Future {
         let excluded = self.inner.exclude.contains(req.path())
-            || self.inner.exclude_regex.is_match(req.path());
+            || self
+                .inner
+                .exclude_regex
+                .iter()
+                .any(|r| r.is_match(req.path()));
 
         if excluded {
             LoggerResponse {

actix-web/src/middleware/normalize.rs

@@ -4,7 +4,10 @@ use actix_http::uri::{PathAndQuery, Uri};
 use actix_service::{Service, Transform};
 use actix_utils::future::{ready, Ready};
 use bytes::Bytes;
+#[cfg(feature = "unicode")]
 use regex::Regex;
+#[cfg(not(feature = "unicode"))]
+use regex_lite::Regex;
 
 use crate::{
     service::{ServiceRequest, ServiceResponse},