Merge 98e0cc4049 into 98d7d0b46b

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

.cspell.yml

@@ -9,4 +9,5 @@ words:
   - rustls
   - rustup
   - serde
+  - uring
   - zstd

Cargo.lock

@@ -44,7 +44,7 @@ dependencies = [
 
 [[package]]
 name = "actix-files"
-version = "0.6.6"
+version = "0.6.7"
 dependencies = [
  "actix-http",
  "actix-rt",
@@ -891,18 +891,18 @@ dependencies = [
 
 [[package]]
 name = "clap"
-version = "4.5.45"
+version = "4.5.46"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1fc0e74a703892159f5ae7d3aac52c8e6c392f5ae5f359c70b5881d60aaac318"
+checksum = "2c5e4fcf9c21d2e544ca1ee9d8552de13019a42aa7dbf32747fa7aaf1df76e57"
 dependencies = [
  "clap_builder",
 ]
 
 [[package]]
 name = "clap_builder"
-version = "4.5.44"
+version = "4.5.46"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b3e7f4214277f3c7aa526a59dd3fbe306a370daee1f8b7b8c987069cd8e888a8"
+checksum = "fecb53a0e6fcfb055f686001bc2e2592fa527efaf38dbe81a6a9563562e57d41"
 dependencies = [
  "anstyle",
  "clap_lex",
@@ -1482,7 +1482,7 @@ dependencies = [
  "cfg-if",
  "libc",
  "r-efi",
- "wasi 0.14.2+wasi-0.2.4",
+ "wasi 0.14.3+wasi-0.2.4",
 ]
 
 [[package]]
@@ -2305,9 +2305,9 @@ dependencies = [
 
 [[package]]
 name = "potential_utf"
-version = "0.1.2"
+version = "0.1.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e5a7c30837279ca13e7c867e9e40053bc68740f988cb07f7ca6df43cc734b585"
+checksum = "84df19adbe5b5a0782edcab45899906947ab039ccf4573713735ee7de1e6b08a"
 dependencies = [
  "zerovec",
 ]
@@ -3461,11 +3461,11 @@ checksum = "ccf3ec651a847eb01de73ccad15eb7d99f80485de043efb2f370cd654f4ea44b"
 
 [[package]]
 name = "wasi"
-version = "0.14.2+wasi-0.2.4"
+version = "0.14.3+wasi-0.2.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9683f9a5a998d873c0d21fcbe3c083009670149a8fab228644b8bd36b2c48cb3"
+checksum = "6a51ae83037bdd272a9e28ce236db8c07016dd0d50c27038b3f407533c030c95"
 dependencies = [
- "wit-bindgen-rt",
+ "wit-bindgen",
 ]
 
 [[package]]
@@ -3873,13 +3873,10 @@ dependencies = [
 ]
 
 [[package]]
-name = "wit-bindgen-rt"
-version = "0.39.0"
+name = "wit-bindgen"
+version = "0.45.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6f42320e61fe2cfd34354ecb597f86f413484a798ba44a8ca1165c58d42da6c1"
-dependencies = [
- "bitflags 2.9.3",
-]
+checksum = "052283831dbae3d879dc7f51f3d92703a316ca49f91540417d38591826127814"
 
 [[package]]
 name = "writeable"

actix-files/CHANGES.md

@@ -2,6 +2,9 @@
 
 ## Unreleased
 
+## 0.6.7
+
+- Add `{Files, NamedFile}::read_mode_threshold()` methods to allow faster synchronous reads of small files.
 - Minimum supported Rust version (MSRV) is now 1.75.
 
 ## 0.6.6

actix-files/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "actix-files"
-version = "0.6.6"
+version = "0.6.7"
 authors = ["Nikolay Kim <fafhrd91@gmail.com>", "Rob Ede <robjtede@icloud.com>"]
 description = "Static file serving for Actix Web"
 keywords = ["actix", "http", "async", "futures"]

actix-files/README.md

@@ -3,11 +3,11 @@
 <!-- prettier-ignore-start -->
 
 [![crates.io](https://img.shields.io/crates/v/actix-files?label=latest)](https://crates.io/crates/actix-files)
-[![Documentation](https://docs.rs/actix-files/badge.svg?version=0.6.6)](https://docs.rs/actix-files/0.6.6)
+[![Documentation](https://docs.rs/actix-files/badge.svg?version=0.6.7)](https://docs.rs/actix-files/0.6.7)
 ![Version](https://img.shields.io/badge/rustc-1.72+-ab6000.svg)
 ![License](https://img.shields.io/crates/l/actix-files.svg)
 <br />
-[![dependency status](https://deps.rs/crate/actix-files/0.6.6/status.svg)](https://deps.rs/crate/actix-files/0.6.6)
+[![dependency status](https://deps.rs/crate/actix-files/0.6.7/status.svg)](https://deps.rs/crate/actix-files/0.6.7)
 [![Download](https://img.shields.io/crates/d/actix-files.svg)](https://crates.io/crates/actix-files)
 [![Chat on Discord](https://img.shields.io/discord/771444961383153695?label=chat&logo=discord)](https://discord.gg/NWpN5mmg3x)
 

actix-files/src/chunked.rs

@@ -14,6 +14,12 @@ use pin_project_lite::pin_project;
 
 use super::named::File;
 
+#[derive(Debug, Clone, Copy)]
+pub(crate) enum ReadMode {
+    Sync,
+    Async,
+}
+
 pin_project! {
     /// Adapter to read a `std::file::File` in chunks.
     #[doc(hidden)]
@@ -24,6 +30,7 @@ pin_project! {
         state: ChunkedReadFileState<Fut>,
         counter: u64,
         callback: F,
+        read_mode: ReadMode,
     }
 }
 
@@ -57,6 +64,7 @@ pub(crate) fn new_chunked_read(
     size: u64,
     offset: u64,
     file: File,
+    read_mode_threshold: u64,
 ) -> impl Stream<Item = Result<Bytes, Error>> {
     ChunkedReadFile {
         size,
@@ -69,31 +77,50 @@ pub(crate) fn new_chunked_read(
         },
         counter: 0,
         callback: chunked_read_file_callback,
+        read_mode: if size < read_mode_threshold {
+            ReadMode::Sync
+        } else {
+            ReadMode::Async
+        },
     }
 }
 
 #[cfg(not(feature = "experimental-io-uring"))]
-async fn chunked_read_file_callback(
+fn chunked_read_file_callback_sync(
     mut file: File,
     offset: u64,
     max_bytes: usize,
-) -> Result<(File, Bytes), Error> {
+) -> Result<(File, Bytes), io::Error> {
     use io::{Read as _, Seek as _};
 
-    let res = actix_web::web::block(move || {
-        let mut buf = Vec::with_capacity(max_bytes);
+    let mut buf = Vec::with_capacity(max_bytes);
 
-        file.seek(io::SeekFrom::Start(offset))?;
+    file.seek(io::SeekFrom::Start(offset))?;
 
-        let n_bytes = file.by_ref().take(max_bytes as u64).read_to_end(&mut buf)?;
+    let n_bytes = file.by_ref().take(max_bytes as u64).read_to_end(&mut buf)?;
 
-        if n_bytes == 0 {
-            Err(io::Error::from(io::ErrorKind::UnexpectedEof))
-        } else {
-            Ok((file, Bytes::from(buf)))
+    if n_bytes == 0 {
+        Err(io::Error::from(io::ErrorKind::UnexpectedEof))
+    } else {
+        Ok((file, Bytes::from(buf)))
+    }
+}
+
+#[cfg(not(feature = "experimental-io-uring"))]
+#[inline]
+async fn chunked_read_file_callback(
+    file: File,
+    offset: u64,
+    max_bytes: usize,
+    read_mode: ReadMode,
+) -> Result<(File, Bytes), Error> {
+    let res = match read_mode {
+        ReadMode::Sync => chunked_read_file_callback_sync(file, offset, max_bytes)?,
+        ReadMode::Async => {
+            actix_web::web::block(move || chunked_read_file_callback_sync(file, offset, max_bytes))
+                .await??
         }
-    })
-    .await??;
+    };
 
     Ok(res)
 }
@@ -171,7 +198,7 @@ where
 #[cfg(not(feature = "experimental-io-uring"))]
 impl<F, Fut> Stream for ChunkedReadFile<F, Fut>
 where
-    F: Fn(File, u64, usize) -> Fut,
+    F: Fn(File, u64, usize, ReadMode) -> Fut,
     Fut: Future<Output = Result<(File, Bytes), Error>>,
 {
     type Item = Result<Bytes, Error>;
@@ -193,7 +220,7 @@ where
                         .take()
                         .expect("ChunkedReadFile polled after completion");
 
-                    let fut = (this.callback)(file, offset, max_bytes);
+                    let fut = (this.callback)(file, offset, max_bytes, *this.read_mode);
 
                     this.state
                         .project_replace(ChunkedReadFileState::Future { fut });

actix-files/src/files.rs

@@ -49,6 +49,7 @@ pub struct Files {
     use_guards: Option<Rc<dyn Guard>>,
     guards: Vec<Rc<dyn Guard>>,
     hidden_files: bool,
+    read_mode_threshold: u64,
 }
 
 impl fmt::Debug for Files {
@@ -73,6 +74,7 @@ impl Clone for Files {
             use_guards: self.use_guards.clone(),
             guards: self.guards.clone(),
             hidden_files: self.hidden_files,
+            read_mode_threshold: self.read_mode_threshold,
         }
     }
 }
@@ -119,6 +121,7 @@ impl Files {
             use_guards: None,
             guards: Vec::new(),
             hidden_files: false,
+            read_mode_threshold: 0,
         }
     }
 
@@ -204,6 +207,23 @@ impl Files {
         self
     }
 
+    /// Sets the size threshold that determines file read mode (sync/async).
+    ///
+    /// When a file is smaller than the threshold (bytes), the reader will switch from synchronous
+    /// (blocking) file-reads to async reads to avoid blocking the main-thread when processing large
+    /// files.
+    ///
+    /// Tweaking this value according to your expected usage may lead to signifiant performance
+    /// gains (or losses in other handlers, if `size` is too high).
+    ///
+    /// When the `experimental-io-uring` crate feature is enabled, file reads are always async.
+    ///
+    /// Default is 0, meaning all files are read asynchronously.
+    pub fn read_mode_threshold(mut self, size: u64) -> Self {
+        self.read_mode_threshold = size;
+        self
+    }
+
     /// Specifies whether to use ETag or not.
     ///
     /// Default is true.
@@ -367,6 +387,7 @@ impl ServiceFactory<ServiceRequest> for Files {
             file_flags: self.file_flags,
             guards: self.use_guards.clone(),
             hidden_files: self.hidden_files,
+            size_threshold: self.read_mode_threshold,
         };
 
         if let Some(ref default) = *self.default.borrow() {

actix-files/src/named.rs

@@ -80,6 +80,7 @@ pub struct NamedFile {
     pub(crate) content_type: Mime,
     pub(crate) content_disposition: ContentDisposition,
     pub(crate) encoding: Option<ContentEncoding>,
+    pub(crate) read_mode_threshold: u64,
 }
 
 #[cfg(not(feature = "experimental-io-uring"))]
@@ -200,6 +201,7 @@ impl NamedFile {
             encoding,
             status_code: StatusCode::OK,
             flags: Flags::default(),
+            read_mode_threshold: 0,
         })
     }
 
@@ -353,6 +355,23 @@ impl NamedFile {
         self
     }
 
+    /// Sets the size threshold that determines file read mode (sync/async).
+    ///
+    /// When a file is smaller than the threshold (bytes), the reader will switch from synchronous
+    /// (blocking) file-reads to async reads to avoid blocking the main-thread when processing large
+    /// files.
+    ///
+    /// Tweaking this value according to your expected usage may lead to signifiant performance
+    /// gains (or losses in other handlers, if `size` is too high).
+    ///
+    /// When the `experimental-io-uring` crate feature is enabled, file reads are always async.
+    ///
+    /// Default is 0, meaning all files are read asynchronously.
+    pub fn read_mode_threshold(mut self, size: u64) -> Self {
+        self.read_mode_threshold = size;
+        self
+    }
+
     /// Specifies whether to return `ETag` header in response.
     ///
     /// Default is true.
@@ -440,7 +459,8 @@ impl NamedFile {
                 res.insert_header((header::CONTENT_ENCODING, current_encoding.as_str()));
             }
 
-            let reader = chunked::new_chunked_read(self.md.len(), 0, self.file);
+            let reader =
+                chunked::new_chunked_read(self.md.len(), 0, self.file, self.read_mode_threshold);
 
             return res.streaming(reader);
         }
@@ -577,7 +597,7 @@ impl NamedFile {
                 .map_into_boxed_body();
         }
 
-        let reader = chunked::new_chunked_read(length, offset, self.file);
+        let reader = chunked::new_chunked_read(length, offset, self.file, self.read_mode_threshold);
 
         if offset != 0 || length != self.md.len() {
             res.status(StatusCode::PARTIAL_CONTENT);

actix-files/src/service.rs

@@ -39,6 +39,7 @@ pub struct FilesServiceInner {
     pub(crate) file_flags: named::Flags,
     pub(crate) guards: Option<Rc<dyn Guard>>,
     pub(crate) hidden_files: bool,
+    pub(crate) size_threshold: u64,
 }
 
 impl fmt::Debug for FilesServiceInner {
@@ -70,7 +71,9 @@ impl FilesService {
         named_file.flags = self.file_flags;
 
         let (req, _) = req.into_parts();
-        let res = named_file.into_response(&req);
+        let res = named_file
+            .read_mode_threshold(self.size_threshold)
+            .into_response(&req);
         ServiceResponse::new(req, res)
     }
 
@@ -169,17 +172,7 @@ impl Service<ServiceRequest> for FilesService {
                 }
             } else {
                 match NamedFile::open_async(&path).await {
-                    Ok(mut named_file) => {
-                        if let Some(ref mime_override) = this.mime_override {
-                            let new_disposition = mime_override(&named_file.content_type.type_());
-                            named_file.content_disposition.disposition = new_disposition;
-                        }
-                        named_file.flags = this.file_flags;
-
-                        let (req, _) = req.into_parts();
-                        let res = named_file.into_response(&req);
-                        Ok(ServiceResponse::new(req, res))
-                    }
+                    Ok(named_file) => Ok(this.serve_named_file(req, named_file)),
                     Err(err) => this.handle_err(err, req).await,
                 }
             }

actix-http/src/header/into_value.rs

@@ -1,7 +1,7 @@
 //! [`TryIntoHeaderValue`] trait and implementations.
 
 use bytes::Bytes;
-use http::{header::InvalidHeaderValue, Error as HttpError, HeaderValue};
+use http::{header::InvalidHeaderValue, Error as HttpError, HeaderValue, Uri};
 use mime::Mime;
 
 /// An interface for types that can be converted into a [`HeaderValue`].
@@ -129,3 +129,12 @@ impl TryIntoHeaderValue for Mime {
         HeaderValue::from_str(self.as_ref())
     }
 }
+
+impl TryIntoHeaderValue for Uri {
+    type Error = InvalidHeaderValue;
+
+    #[inline]
+    fn try_into_value(self) -> Result<HeaderValue, Self::Error> {
+        HeaderValue::from_str(&self.to_string())
+    }
+}

actix-web/CHANGES.md

@@ -36,6 +36,10 @@
 - On Windows, an error is now returned from `HttpServer::bind()` (or TLS variants) when binding to a socket that's already in use.
 - Update `brotli` dependency to `7`.
 - Minimum supported Rust version (MSRV) is now 1.75.
+- Add `TryIntoHeaderValue` for `Uri` type.
+- Add `http::header::ContentLocation` typed header.
+- Add `http::header::Location` typed header.
+- Add `http::header::Referer` typed header.
 
 ## 4.9.0
 

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

@@ -0,0 +1,36 @@
+use super::{Uri, CONTENT_LOCATION};
+
+crate::http::header::common_header! {
+    /// `Content-Location` header, defined
+    /// in [RFC 9110 §8.7](https://datatracker.ietf.org/doc/html/rfc9110#section-8.7)
+    ///
+    /// The "Content-Location" header field references a URI that can be used
+    /// as an identifier for a specific resource corresponding to the
+    /// representation in this message's content.
+    ///
+    /// # ABNF
+    /// ```plain
+    /// Content-Location = absolute-URI / partial-URI
+    /// ```
+    ///
+    /// # Example Values
+    /// * `http://www.example.org/hypertext/Overview.html`
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use actix_web::HttpResponse;
+    /// use actix_http::Uri;
+    /// use actix_web::http::header::ContentLocation;
+    ///
+    /// let mut builder = HttpResponse::Created();
+    /// builder.insert_header(
+    ///     ContentLocation("http://www.example.org".parse::<Uri>().unwrap())
+    /// );
+    /// ```
+    (ContentLocation, CONTENT_LOCATION) => [Uri]
+
+    test_parse_and_format {
+        crate::http::header::common_header_test!(test1, [b"http://www.example.org/hypertext/Overview.html"]);
+    }
+}

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

@@ -0,0 +1,37 @@
+use super::{Uri, LOCATION};
+
+crate::http::header::common_header! {
+    /// `Location` header, defined
+    /// in [RFC 9110 §10.2.2](https://datatracker.ietf.org/doc/html/rfc9110#section-10.2.2)
+    ///
+    /// The "Location" header field is used in some responses to refer to a
+    /// specific resource in relation to the response. The type of relationship
+    /// is defined by the combination of request method and status code
+    /// semantics.
+    ///
+    /// # ABNF
+    /// ```plain
+    /// Location = URI-reference
+    /// ```
+    ///
+    /// # Example Values
+    /// * `http://www.example.org/hypertext/Overview.html`
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use actix_web::HttpResponse;
+    /// use actix_http::Uri;
+    /// use actix_web::http::header::Location;
+    ///
+    /// let mut builder = HttpResponse::Ok();
+    /// builder.insert_header(
+    ///     Location("http://www.example.org".parse::<Uri>().unwrap())
+    /// );
+    /// ```
+    (Location, LOCATION) => [Uri]
+
+    test_parse_and_format {
+        crate::http::header::common_header_test!(test1, [b"http://www.example.org/hypertext/Overview.html"]);
+    }
+}

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

@@ -14,6 +14,7 @@ use std::fmt;
 // - the few typed headers from actix-http
 // - header parsing utils
 pub use actix_http::header::*;
+use actix_http::Uri;
 use bytes::{Bytes, BytesMut};
 
 mod accept;
@@ -25,6 +26,7 @@ mod cache_control;
 mod content_disposition;
 mod content_language;
 mod content_length;
+mod content_location;
 mod content_range;
 mod content_type;
 mod date;
@@ -38,9 +40,11 @@ mod if_none_match;
 mod if_range;
 mod if_unmodified_since;
 mod last_modified;
+mod location;
 mod macros;
 mod preference;
 mod range;
+mod referer;
 
 #[cfg(test)]
 pub(crate) use self::macros::common_header_test;
@@ -55,6 +59,7 @@ pub use self::{
     content_disposition::{ContentDisposition, DispositionParam, DispositionType},
     content_language::ContentLanguage,
     content_length::ContentLength,
+    content_location::ContentLocation,
     content_range::{ContentRange, ContentRangeSpec},
     content_type::ContentType,
     date::Date,
@@ -68,8 +73,10 @@ pub use self::{
     if_range::IfRange,
     if_unmodified_since::IfUnmodifiedSince,
     last_modified::LastModified,
+    location::Location,
     preference::Preference,
     range::{ByteRangeSpec, Range},
+    referer::Referer,
 };
 
 /// Format writer ([`fmt::Write`]) for a [`BytesMut`].

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

@@ -0,0 +1,36 @@
+use super::{Uri, REFERER};
+
+crate::http::header::common_header! {
+    /// `Referer` header, defined
+    /// in [RFC 9110 §10.1.3](https://datatracker.ietf.org/doc/html/rfc9110#section-10.1.3)
+    ///
+    /// The "Referer" (sic) header field allows the user agent to specify a URI
+    /// reference for the resource from which the target URI was obtained (i.e.,
+    /// the "referrer", though the field name is misspelled).
+    ///
+    /// # ABNF
+    /// ```plain
+    /// Referer = absolute-URI / partial-URI
+    /// ```
+    ///
+    /// # Example Values
+    /// * `http://www.example.org/hypertext/Overview.html`
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use actix_web::HttpResponse;
+    /// use actix_http::Uri;
+    /// use actix_web::http::header::Referer;
+    ///
+    /// let mut builder = HttpResponse::Ok();
+    /// builder.insert_header(
+    ///     Referer("http://www.example.org".parse::<Uri>().unwrap())
+    /// );
+    /// ```
+    (Referer, REFERER) => [Uri]
+
+    test_parse_and_format {
+        crate::http::header::common_header_test!(test1, [b"http://www.example.org/hypertext/Overview.html"]);
+    }
+}

actix-web/src/middleware/authors-guide.md

@@ -2,16 +2,80 @@
 
 ## What Is A Middleware?
 
+Middleware in Actix Web is a powerful mechanism that allows you to add additional behavior to request/response processing. It enables you to:
+
+- Pre-process incoming requests (e.g., path normalization, authentication)
+- Post-process outgoing responses (e.g., logging, compression)
+- Modify application state through ServiceRequest
+- Access external services (e.g., sessions, caching)
+
+Middleware is registered for each App, Scope, or Resource and executed in the reverse order of registration. This means the last registered middleware is the first to process the request.
+
 ## Middleware Traits
 
+Actix Web's middleware system is built on two main traits:
+
+1. `Transform<S, Req>`: The builder trait that creates the actual Service. It's responsible for:
+
+   - Creating new middleware instances
+   - Assembling the middleware chain
+   - Handling initialization errors
+
+2. `Service<Req>`: The trait that represents the actual middleware functionality. It:
+   - Processes requests and responses
+   - Can modify both request and response
+   - Can short-circuit request processing
+   - Must be implemented for the middleware to work
+
 ## Understanding Body Types
 
+When working with middleware, it's important to understand body types:
+
+- Middleware can work with different body types for requests and responses
+- The `MessageBody` trait is used to handle different body types
+- You can use `EitherBody` when you need to handle multiple body types
+- Be careful with body consumption - once a body is consumed, it cannot be read again
+
 ## Best Practices
 
+1. Keep middleware focused and single-purpose
+2. Handle errors appropriately and propagate them correctly
+3. Be mindful of performance impact
+4. Use appropriate body types and handle them correctly
+5. Consider middleware ordering carefully
+6. Document your middleware's behavior and requirements
+7. Test your middleware thoroughly
+
 ## Error Propagation
 
+Proper error handling is crucial in middleware:
+
+1. Always propagate errors from the inner service
+2. Use appropriate error types
+3. Handle initialization errors
+4. Consider using custom error types for specific middleware errors
+5. Document error conditions and handling
+
 ## When To (Not) Use Middleware
 
+Use middleware when you need to:
+
+- Add cross-cutting concerns
+- Modify requests/responses globally
+- Add authentication/authorization
+- Add logging or monitoring
+- Handle compression or caching
+
+Avoid middleware when:
+
+- The functionality is specific to a single route
+- The operation is better handled by a service
+- The overhead would be too high
+- The functionality can be implemented more simply
+
 ## Author's References
 
 - `EitherBody` + when is middleware appropriate: https://discord.com/channels/771444961383153695/952016890723729428
+- Actix Web Documentation: https://docs.rs/actix-web
+- Service Trait Documentation: https://docs.rs/actix-service
+- MessageBody Trait Documentation: https://docs.rs/actix-web/latest/actix_web/body/trait.MessageBody.html