use super::table::{Index, Table}; use super::{huffman, Header}; use bytes::{BufMut, BytesMut}; use http::header::{HeaderName, HeaderValue}; #[derive(Debug)] pub struct Encoder { table: Table, size_update: Option, } #[derive(Debug, Copy, Clone, Eq, PartialEq)] enum SizeUpdate { One(usize), Two(usize, usize), // min, max } impl Encoder { pub fn new(max_size: usize, capacity: usize) -> Encoder { Encoder { table: Table::new(max_size, capacity), size_update: None, } } /// Queues a max size update. /// /// The next call to `encode` will include a dynamic size update frame. pub fn update_max_size(&mut self, val: usize) { match self.size_update { Some(SizeUpdate::One(old)) => { if val > old { if old > self.table.max_size() { self.size_update = Some(SizeUpdate::One(val)); } else { self.size_update = Some(SizeUpdate::Two(old, val)); } } else { self.size_update = Some(SizeUpdate::One(val)); } } Some(SizeUpdate::Two(min, _)) => { if val < min { self.size_update = Some(SizeUpdate::One(val)); } else { self.size_update = Some(SizeUpdate::Two(min, val)); } } None => { if val != self.table.max_size() { // Don't bother writing a frame if the value already matches // the table's max size. self.size_update = Some(SizeUpdate::One(val)); } } } } /// Encode a set of headers into the provide buffer pub fn encode(&mut self, headers: I, dst: &mut BytesMut) where I: IntoIterator>>, { let span = tracing::trace_span!("hpack::encode"); let _e = span.enter(); self.encode_size_updates(dst); let mut last_index = None; for header in headers { match header.reify() { // The header has an associated name. In which case, try to // index it in the table. Ok(header) => { let index = self.table.index(header); self.encode_header(&index, dst); last_index = Some(index); } // The header does not have an associated name. This means that // the name is the same as the previously yielded header. In // which case, we skip table lookup and just use the same index // as the previous entry. Err(value) => { self.encode_header_without_name( last_index.as_ref().unwrap_or_else(|| { panic!("encoding header without name, but no previous index to use for name"); }), &value, dst, ); } } } } fn encode_size_updates(&mut self, dst: &mut BytesMut) { match self.size_update.take() { Some(SizeUpdate::One(val)) => { self.table.resize(val); encode_size_update(val, dst); } Some(SizeUpdate::Two(min, max)) => { self.table.resize(min); self.table.resize(max); encode_size_update(min, dst); encode_size_update(max, dst); } None => {} } } fn encode_header(&mut self, index: &Index, dst: &mut BytesMut) { match *index { Index::Indexed(idx, _) => { encode_int(idx, 7, 0x80, dst); } Index::Name(idx, _) => { let header = self.table.resolve(index); encode_not_indexed(idx, header.value_slice(), header.is_sensitive(), dst); } Index::Inserted(_) => { let header = self.table.resolve(index); assert!(!header.is_sensitive()); dst.put_u8(0b0100_0000); encode_str(header.name().as_slice(), dst); encode_str(header.value_slice(), dst); } Index::InsertedValue(idx, _) => { let header = self.table.resolve(index); assert!(!header.is_sensitive()); encode_int(idx, 6, 0b0100_0000, dst); encode_str(header.value_slice(), dst); } Index::NotIndexed(_) => { let header = self.table.resolve(index); encode_not_indexed2( header.name().as_slice(), header.value_slice(), header.is_sensitive(), dst, ); } } } fn encode_header_without_name( &mut self, last: &Index, value: &HeaderValue, dst: &mut BytesMut, ) { match *last { Index::Indexed(..) | Index::Name(..) | Index::Inserted(..) | Index::InsertedValue(..) => { let idx = self.table.resolve_idx(last); encode_not_indexed(idx, value.as_ref(), value.is_sensitive(), dst); } Index::NotIndexed(_) => { let last = self.table.resolve(last); encode_not_indexed2( last.name().as_slice(), value.as_ref(), value.is_sensitive(), dst, ); } } } } impl Default for Encoder { fn default() -> Encoder { Encoder::new(4096, 0) } } fn encode_size_update(val: usize, dst: &mut BytesMut) { encode_int(val, 5, 0b0010_0000, dst) } fn encode_not_indexed(name: usize, value: &[u8], sensitive: bool, dst: &mut BytesMut) { if sensitive { encode_int(name, 4, 0b10000, dst); } else { encode_int(name, 4, 0, dst); } encode_str(value, dst); } fn encode_not_indexed2(name: &[u8], value: &[u8], sensitive: bool, dst: &mut BytesMut) { if sensitive { dst.put_u8(0b10000); } else { dst.put_u8(0); } encode_str(name, dst); encode_str(value, dst); } fn encode_str(val: &[u8], dst: &mut BytesMut) { if !val.is_empty() { let idx = position(dst); // Push a placeholder byte for the length header dst.put_u8(0); // Encode with huffman huffman::encode(val, dst); let huff_len = position(dst) - (idx + 1); if encode_int_one_byte(huff_len, 7) { // Write the string head dst[idx] = 0x80 | huff_len as u8; } else { // Write the head to a placeholder const PLACEHOLDER_LEN: usize = 8; let mut buf = [0u8; PLACEHOLDER_LEN]; let head_len = { let mut head_dst = &mut buf[..]; encode_int(huff_len, 7, 0x80, &mut head_dst); PLACEHOLDER_LEN - head_dst.remaining_mut() }; // This is just done to reserve space in the destination dst.put_slice(&buf[1..head_len]); // Shift the header forward for i in 0..huff_len { let src_i = idx + 1 + (huff_len - (i + 1)); let dst_i = idx + head_len + (huff_len - (i + 1)); dst[dst_i] = dst[src_i]; } // Copy in the head for i in 0..head_len { dst[idx + i] = buf[i]; } } } else { // Write an empty string dst.put_u8(0); } } /// Encode an integer into the given destination buffer fn encode_int( mut value: usize, // The integer to encode prefix_bits: usize, // The number of bits in the prefix first_byte: u8, // The base upon which to start encoding the int dst: &mut B, ) { if encode_int_one_byte(value, prefix_bits) { dst.put_u8(first_byte | value as u8); return; } let low = (1 << prefix_bits) - 1; value -= low; dst.put_u8(first_byte | low as u8); while value >= 128 { dst.put_u8(0b1000_0000 | value as u8); value >>= 7; } dst.put_u8(value as u8); } /// Returns true if the in the int can be fully encoded in the first byte. fn encode_int_one_byte(value: usize, prefix_bits: usize) -> bool { value < (1 << prefix_bits) - 1 } fn position(buf: &BytesMut) -> usize { buf.len() } #[cfg(test)] mod test { use super::*; use crate::hpack::Header; use http::*; #[test] fn test_encode_method_get() { let mut encoder = Encoder::default(); let res = encode(&mut encoder, vec![method("GET")]); assert_eq!(*res, [0x80 | 2]); assert_eq!(encoder.table.len(), 0); } #[test] fn test_encode_method_post() { let mut encoder = Encoder::default(); let res = encode(&mut encoder, vec![method("POST")]); assert_eq!(*res, [0x80 | 3]); assert_eq!(encoder.table.len(), 0); } #[test] fn test_encode_method_patch() { let mut encoder = Encoder::default(); let res = encode(&mut encoder, vec![method("PATCH")]); assert_eq!(res[0], 0b01000000 | 2); // Incremental indexing w/ name pulled from table assert_eq!(res[1], 0x80 | 5); // header value w/ huffman coding assert_eq!("PATCH", huff_decode(&res[2..7])); assert_eq!(encoder.table.len(), 1); let res = encode(&mut encoder, vec![method("PATCH")]); assert_eq!(1 << 7 | 62, res[0]); assert_eq!(1, res.len()); } #[test] fn test_encode_indexed_name_literal_value() { let mut encoder = Encoder::default(); let res = encode(&mut encoder, vec![header("content-language", "foo")]); assert_eq!(res[0], 0b01000000 | 27); // Indexed name assert_eq!(res[1], 0x80 | 2); // header value w/ huffman coding assert_eq!("foo", huff_decode(&res[2..4])); // Same name, new value should still use incremental let res = encode(&mut encoder, vec![header("content-language", "bar")]); assert_eq!(res[0], 0b01000000 | 27); // Indexed name assert_eq!(res[1], 0x80 | 3); // header value w/ huffman coding assert_eq!("bar", huff_decode(&res[2..5])); } #[test] fn test_repeated_headers_are_indexed() { let mut encoder = Encoder::default(); let res = encode(&mut encoder, vec![header("foo", "hello")]); assert_eq!(&[0b01000000, 0x80 | 2], &res[0..2]); assert_eq!("foo", huff_decode(&res[2..4])); assert_eq!(0x80 | 4, res[4]); assert_eq!("hello", huff_decode(&res[5..])); assert_eq!(9, res.len()); assert_eq!(1, encoder.table.len()); let res = encode(&mut encoder, vec![header("foo", "hello")]); assert_eq!([0x80 | 62], *res); assert_eq!(encoder.table.len(), 1); } #[test] fn test_evicting_headers() { let mut encoder = Encoder::default(); // Fill the table for i in 0..64 { let key = format!("x-hello-world-{:02}", i); let res = encode(&mut encoder, vec![header(&key, &key)]); assert_eq!(&[0b01000000, 0x80 | 12], &res[0..2]); assert_eq!(key, huff_decode(&res[2..14])); assert_eq!(0x80 | 12, res[14]); assert_eq!(key, huff_decode(&res[15..])); assert_eq!(27, res.len()); // Make sure the header can be found... let res = encode(&mut encoder, vec![header(&key, &key)]); // Only check that it is found assert_eq!(0x80, res[0] & 0x80); } assert_eq!(4096, encoder.table.size()); assert_eq!(64, encoder.table.len()); // Find existing headers for i in 0..64 { let key = format!("x-hello-world-{:02}", i); let res = encode(&mut encoder, vec![header(&key, &key)]); assert_eq!(0x80, res[0] & 0x80); } // Insert a new header let key = "x-hello-world-64"; let res = encode(&mut encoder, vec![header(key, key)]); assert_eq!(&[0b01000000, 0x80 | 12], &res[0..2]); assert_eq!(key, huff_decode(&res[2..14])); assert_eq!(0x80 | 12, res[14]); assert_eq!(key, huff_decode(&res[15..])); assert_eq!(27, res.len()); assert_eq!(64, encoder.table.len()); // Now try encoding entries that should exist in the table for i in 1..65 { let key = format!("x-hello-world-{:02}", i); let res = encode(&mut encoder, vec![header(&key, &key)]); assert_eq!(0x80 | (61 + (65 - i)), res[0]); } } #[test] fn test_large_headers_are_not_indexed() { let mut encoder = Encoder::new(128, 0); let key = "hello-world-hello-world-HELLO-zzz"; let res = encode(&mut encoder, vec![header(key, key)]); assert_eq!(&[0, 0x80 | 25], &res[..2]); assert_eq!(0, encoder.table.len()); assert_eq!(0, encoder.table.size()); } #[test] fn test_sensitive_headers_are_never_indexed() { use http::header::HeaderValue; let name = "my-password".parse().unwrap(); let mut value = HeaderValue::from_bytes(b"12345").unwrap(); value.set_sensitive(true); let header = Header::Field { name: Some(name), value, }; // Now, try to encode the sensitive header let mut encoder = Encoder::default(); let res = encode(&mut encoder, vec![header]); assert_eq!(&[0b10000, 0x80 | 8], &res[..2]); assert_eq!("my-password", huff_decode(&res[2..10])); assert_eq!(0x80 | 4, res[10]); assert_eq!("12345", huff_decode(&res[11..])); // Now, try to encode a sensitive header w/ a name in the static table let name = "authorization".parse().unwrap(); let mut value = HeaderValue::from_bytes(b"12345").unwrap(); value.set_sensitive(true); let header = Header::Field { name: Some(name), value, }; let mut encoder = Encoder::default(); let res = encode(&mut encoder, vec![header]); assert_eq!(&[0b11111, 8], &res[..2]); assert_eq!(0x80 | 4, res[2]); assert_eq!("12345", huff_decode(&res[3..])); // Using the name component of a previously indexed header (without // sensitive flag set) let _ = encode( &mut encoder, vec![self::header("my-password", "not-so-secret")], ); let name = "my-password".parse().unwrap(); let mut value = HeaderValue::from_bytes(b"12345").unwrap(); value.set_sensitive(true); let header = Header::Field { name: Some(name), value, }; let res = encode(&mut encoder, vec![header]); assert_eq!(&[0b11111, 47], &res[..2]); assert_eq!(0x80 | 4, res[2]); assert_eq!("12345", huff_decode(&res[3..])); } #[test] fn test_content_length_value_not_indexed() { let mut encoder = Encoder::default(); let res = encode(&mut encoder, vec![header("content-length", "1234")]); assert_eq!(&[15, 13, 0x80 | 3], &res[0..3]); assert_eq!("1234", huff_decode(&res[3..])); assert_eq!(6, res.len()); } #[test] fn test_encoding_headers_with_same_name() { let mut encoder = Encoder::default(); let name = "hello"; // Encode first one let _ = encode(&mut encoder, vec![header(name, "one")]); // Encode second one let res = encode(&mut encoder, vec![header(name, "two")]); assert_eq!(&[0x40 | 62, 0x80 | 3], &res[0..2]); assert_eq!("two", huff_decode(&res[2..])); assert_eq!(5, res.len()); // Encode the first one again let res = encode(&mut encoder, vec![header(name, "one")]); assert_eq!(&[0x80 | 63], &res[..]); // Now the second one let res = encode(&mut encoder, vec![header(name, "two")]); assert_eq!(&[0x80 | 62], &res[..]); } #[test] fn test_evicting_headers_when_multiple_of_same_name_are_in_table() { // The encoder only has space for 2 headers let mut encoder = Encoder::new(76, 0); let _ = encode(&mut encoder, vec![header("foo", "bar")]); assert_eq!(1, encoder.table.len()); let _ = encode(&mut encoder, vec![header("bar", "foo")]); assert_eq!(2, encoder.table.len()); // This will evict the first header, while still referencing the header // name let res = encode(&mut encoder, vec![header("foo", "baz")]); assert_eq!(&[0x40 | 63, 0, 0x80 | 3], &res[..3]); assert_eq!(2, encoder.table.len()); // Try adding the same header again let res = encode(&mut encoder, vec![header("foo", "baz")]); assert_eq!(&[0x80 | 62], &res[..]); assert_eq!(2, encoder.table.len()); } #[test] fn test_max_size_zero() { // Static table only let mut encoder = Encoder::new(0, 0); let res = encode(&mut encoder, vec![method("GET")]); assert_eq!(*res, [0x80 | 2]); assert_eq!(encoder.table.len(), 0); let res = encode(&mut encoder, vec![header("foo", "bar")]); assert_eq!(&[0, 0x80 | 2], &res[..2]); assert_eq!("foo", huff_decode(&res[2..4])); assert_eq!(0x80 | 3, res[4]); assert_eq!("bar", huff_decode(&res[5..8])); assert_eq!(0, encoder.table.len()); // Encode a custom value let res = encode(&mut encoder, vec![header("transfer-encoding", "chunked")]); assert_eq!(&[15, 42, 0x80 | 6], &res[..3]); assert_eq!("chunked", huff_decode(&res[3..])); } #[test] fn test_update_max_size_combos() { let mut encoder = Encoder::default(); assert!(encoder.size_update.is_none()); assert_eq!(4096, encoder.table.max_size()); encoder.update_max_size(4096); // Default size assert!(encoder.size_update.is_none()); encoder.update_max_size(0); assert_eq!(Some(SizeUpdate::One(0)), encoder.size_update); encoder.update_max_size(100); assert_eq!(Some(SizeUpdate::Two(0, 100)), encoder.size_update); let mut encoder = Encoder::default(); encoder.update_max_size(8000); assert_eq!(Some(SizeUpdate::One(8000)), encoder.size_update); encoder.update_max_size(100); assert_eq!(Some(SizeUpdate::One(100)), encoder.size_update); encoder.update_max_size(8000); assert_eq!(Some(SizeUpdate::Two(100, 8000)), encoder.size_update); encoder.update_max_size(4000); assert_eq!(Some(SizeUpdate::Two(100, 4000)), encoder.size_update); encoder.update_max_size(50); assert_eq!(Some(SizeUpdate::One(50)), encoder.size_update); } #[test] fn test_resizing_table() { let mut encoder = Encoder::default(); // Add a header let _ = encode(&mut encoder, vec![header("foo", "bar")]); encoder.update_max_size(1); assert_eq!(1, encoder.table.len()); let res = encode(&mut encoder, vec![method("GET")]); assert_eq!(&[32 | 1, 0x80 | 2], &res[..]); assert_eq!(0, encoder.table.len()); let res = encode(&mut encoder, vec![header("foo", "bar")]); assert_eq!(0, res[0]); encoder.update_max_size(100); let res = encode(&mut encoder, vec![header("foo", "bar")]); assert_eq!(&[32 | 31, 69, 64], &res[..3]); encoder.update_max_size(0); let res = encode(&mut encoder, vec![header("foo", "bar")]); assert_eq!(&[32, 0], &res[..2]); } #[test] fn test_decreasing_table_size_without_eviction() { let mut encoder = Encoder::default(); // Add a header let _ = encode(&mut encoder, vec![header("foo", "bar")]); encoder.update_max_size(100); assert_eq!(1, encoder.table.len()); let res = encode(&mut encoder, vec![header("foo", "bar")]); assert_eq!(&[32 | 31, 69, 0x80 | 62], &res[..]); } #[test] fn test_nameless_header() { let mut encoder = Encoder::default(); let res = encode( &mut encoder, vec![ Header::Field { name: Some("hello".parse().unwrap()), value: HeaderValue::from_bytes(b"world").unwrap(), }, Header::Field { name: None, value: HeaderValue::from_bytes(b"zomg").unwrap(), }, ], ); assert_eq!(&[0x40, 0x80 | 4], &res[0..2]); assert_eq!("hello", huff_decode(&res[2..6])); assert_eq!(0x80 | 4, res[6]); assert_eq!("world", huff_decode(&res[7..11])); // Next is not indexed assert_eq!(&[15, 47, 0x80 | 3], &res[11..14]); assert_eq!("zomg", huff_decode(&res[14..])); } #[test] fn test_large_size_update() { let mut encoder = Encoder::default(); encoder.update_max_size(1912930560); assert_eq!(Some(SizeUpdate::One(1912930560)), encoder.size_update); let mut dst = BytesMut::with_capacity(6); encoder.encode_size_updates(&mut dst); assert_eq!([63, 225, 129, 148, 144, 7], &dst[..]); } #[test] #[ignore] fn test_evicted_overflow() { // Not sure what the best way to do this is. } fn encode(e: &mut Encoder, hdrs: Vec>>) -> BytesMut { let mut dst = BytesMut::with_capacity(1024); e.encode(&mut hdrs.into_iter(), &mut dst); dst } fn method(s: &str) -> Header> { Header::Method(Method::from_bytes(s.as_bytes()).unwrap()) } fn header(name: &str, val: &str) -> Header> { let name = HeaderName::from_bytes(name.as_bytes()).unwrap(); let value = HeaderValue::from_bytes(val.as_bytes()).unwrap(); Header::Field { name: Some(name), value, } } fn huff_decode(src: &[u8]) -> BytesMut { let mut buf = BytesMut::new(); huffman::decode(src, &mut buf).unwrap() } }