|
| 1 | +# cpp-httplib C++20 Streaming API |
| 2 | + |
| 3 | +This document describes the C++20 streaming extensions for cpp-httplib, providing a generator-like API for handling HTTP responses incrementally with **true socket-level streaming**. |
| 4 | + |
| 5 | +> **Important Notes**: |
| 6 | +> |
| 7 | +> - **No Keep-Alive**: Each `stream::Get()` call uses a dedicated connection that is closed after the response is fully read. For connection reuse, use `Client::Get()`. |
| 8 | +> - **Single iteration only**: The `body()` generator can only be iterated once. Calling `body()` again after iteration has no effect. |
| 9 | +> - **Result is not thread-safe**: While `stream::Get()` can be called from multiple threads simultaneously, the returned `stream::Result` must be used from a single thread only. |
| 10 | +
|
| 11 | +## Overview |
| 12 | + |
| 13 | +The C++20 streaming API allows you to process HTTP response bodies chunk by chunk using C++20 coroutines, similar to Python's generators or C++23's `std::generator`. Data is read directly from the network socket, enabling low-memory processing of large responses. This is particularly useful for: |
| 14 | + |
| 15 | +- **LLM/AI streaming responses** (e.g., ChatGPT, Claude, Ollama) |
| 16 | +- **Server-Sent Events (SSE)** |
| 17 | +- **Large file downloads** with progress tracking |
| 18 | +- **Reverse proxy implementations** |
| 19 | + |
| 20 | +## Requirements |
| 21 | + |
| 22 | +- C++20 compiler with coroutine support |
| 23 | +- Include `httplib.h` and compile with `-std=c++20` |
| 24 | + |
| 25 | +## Quick Start |
| 26 | + |
| 27 | +```cpp |
| 28 | +#include "httplib.h" |
| 29 | + |
| 30 | +int main() { |
| 31 | + httplib::Client cli("http://localhost:8080"); |
| 32 | + |
| 33 | + // Get streaming response |
| 34 | + auto result = httplib::stream::Get(cli, "/stream"); |
| 35 | + |
| 36 | + if (result) { |
| 37 | + // Process response body in chunks |
| 38 | + for (auto chunk : result.body(4096)) { |
| 39 | + std::cout << chunk; // Process each chunk as it arrives |
| 40 | + } |
| 41 | + } |
| 42 | + |
| 43 | + return 0; |
| 44 | +} |
| 45 | +``` |
| 46 | + |
| 47 | +## API Layers |
| 48 | + |
| 49 | +cpp-httplib provides multiple API layers for different use cases: |
| 50 | + |
| 51 | +```text |
| 52 | +┌─────────────────────────────────────────────┐ |
| 53 | +│ SSEClient (planned) │ ← SSE-specific, parsed events |
| 54 | +│ - on_message(), on_event() │ |
| 55 | +│ - Auto-reconnect, Last-Event-ID │ |
| 56 | +├─────────────────────────────────────────────┤ |
| 57 | +│ stream::Get() / stream::Result │ ← C++20, Generator-based |
| 58 | +│ - for (auto chunk : result.body()) │ |
| 59 | +├─────────────────────────────────────────────┤ |
| 60 | +│ open_stream() / StreamHandle │ ← General-purpose streaming |
| 61 | +│ - handle.read(buf, len) │ |
| 62 | +├─────────────────────────────────────────────┤ |
| 63 | +│ Client::Get() │ ← Traditional, full buffering |
| 64 | +└─────────────────────────────────────────────┘ |
| 65 | +``` |
| 66 | + |
| 67 | +| Use Case | Recommended API | |
| 68 | +|----------|----------------| |
| 69 | +| SSE with auto-reconnect | SSEClient (planned) or `ssecli-stream.cc` example | |
| 70 | +| LLM streaming (JSON Lines) | `stream::Get()` | |
| 71 | +| Large file download | `stream::Get()` or `open_stream()` | |
| 72 | +| Reverse proxy | `open_stream()` | |
| 73 | +| Small responses with Keep-Alive | `Client::Get()` | |
| 74 | + |
| 75 | +## API Reference |
| 76 | + |
| 77 | +### Low-Level API: `StreamHandle` |
| 78 | + |
| 79 | +The `StreamHandle` struct provides direct control over streaming responses. It takes ownership of the socket connection and reads data directly from the network. |
| 80 | + |
| 81 | +> **Note:** When using `open_stream()`, the connection is dedicated to streaming and **Keep-Alive is not supported**. For Keep-Alive connections, use `client.Get()` instead. |
| 82 | +
|
| 83 | +```cpp |
| 84 | +// Open a stream (takes ownership of socket) |
| 85 | +httplib::Client cli("http://localhost:8080"); |
| 86 | +auto handle = cli.open_stream("/path"); |
| 87 | + |
| 88 | +// Check validity |
| 89 | +if (handle.is_valid()) { |
| 90 | + // Access response headers immediately |
| 91 | + int status = handle.response->status; |
| 92 | + auto content_type = handle.response->get_header_value("Content-Type"); |
| 93 | + |
| 94 | + // Read body incrementally |
| 95 | + char buf[4096]; |
| 96 | + ssize_t n; |
| 97 | + while ((n = handle.read(buf, sizeof(buf))) > 0) { |
| 98 | + process(buf, n); |
| 99 | + } |
| 100 | +} |
| 101 | +``` |
| 102 | +
|
| 103 | +#### StreamHandle Members |
| 104 | +
|
| 105 | +| Member | Type | Description | |
| 106 | +|--------|------|-------------| |
| 107 | +| `response` | `std::unique_ptr<Response>` | HTTP response with headers | |
| 108 | +| `error` | `Error` | Error code if request failed | |
| 109 | +| `is_valid()` | `bool` | Returns true if response is valid | |
| 110 | +| `read(buf, len)` | `ssize_t` | Read up to `len` bytes directly from socket | |
| 111 | +| `get_read_error()` | `Error` | Get the last read error | |
| 112 | +| `has_read_error()` | `bool` | Check if a read error occurred | |
| 113 | +
|
| 114 | +### High-Level API: `stream::Get()` and `stream::Result` |
| 115 | +
|
| 116 | +The `httplib.h` header provides a more ergonomic API using C++20 coroutines (requires C++20). |
| 117 | +
|
| 118 | +```cpp |
| 119 | +#include "httplib.h" |
| 120 | +
|
| 121 | +httplib::Client cli("http://localhost:8080"); |
| 122 | +
|
| 123 | +// Simple GET |
| 124 | +auto result = httplib::stream::Get(cli, "/path"); |
| 125 | +
|
| 126 | +// GET with custom headers |
| 127 | +httplib::Headers headers = {{"Authorization", "Bearer token"}}; |
| 128 | +auto result = httplib::stream::Get(cli, "/path", headers); |
| 129 | +``` |
| 130 | + |
| 131 | +#### stream::Result Members |
| 132 | + |
| 133 | +| Member | Type | Description | |
| 134 | +|--------|------|-------------| |
| 135 | +| `operator bool()` | `bool` | Returns true if response is valid | |
| 136 | +| `status()` | `int` | HTTP status code | |
| 137 | +| `headers()` | `const Headers&` | Response headers | |
| 138 | +| `get_header_value(key, def)` | `std::string` | Get header value (with optional default) | |
| 139 | +| `has_header(key)` | `bool` | Check if header exists | |
| 140 | +| `body(chunk_size)` | `Generator<std::string_view>` | Generator yielding body chunks | |
| 141 | +| `error()` | `Error` | Get the connection/request error | |
| 142 | +| `read_error()` | `Error` | Get the last read error | |
| 143 | +| `has_read_error()` | `bool` | Check if a read error occurred | |
| 144 | + |
| 145 | +### Generator Class |
| 146 | + |
| 147 | +The `httplib::Generator<T>` class is a C++20 coroutine-based generator, similar to `std::generator` (C++23). |
| 148 | + |
| 149 | +```cpp |
| 150 | +// Iterate over chunks |
| 151 | +for (auto chunk : result.body(1024)) { |
| 152 | + // chunk is std::string_view |
| 153 | + process(chunk); |
| 154 | +} |
| 155 | +``` |
| 156 | + |
| 157 | +## Usage Examples |
| 158 | + |
| 159 | +### Example 1: SSE (Server-Sent Events) Client |
| 160 | + |
| 161 | +```cpp |
| 162 | +#include "httplib.h" |
| 163 | +#include <iostream> |
| 164 | + |
| 165 | +int main() { |
| 166 | + httplib::Client cli("http://localhost:1234"); |
| 167 | + |
| 168 | + auto result = httplib::stream::Get(cli, "/events"); |
| 169 | + if (!result) { return 1; } |
| 170 | + |
| 171 | + for (auto chunk : result.body()) { |
| 172 | + std::cout << chunk << std::flush; |
| 173 | + } |
| 174 | + |
| 175 | + return 0; |
| 176 | +} |
| 177 | +``` |
| 178 | + |
| 179 | +For a complete SSE client with auto-reconnection and event parsing, see `example/ssecli-stream.cc`. |
| 180 | + |
| 181 | +### Example 2: LLM Streaming Response |
| 182 | + |
| 183 | +```cpp |
| 184 | +#include "httplib.h" |
| 185 | +#include <iostream> |
| 186 | + |
| 187 | +int main() { |
| 188 | + httplib::Client cli("http://localhost:11434"); // Ollama |
| 189 | + |
| 190 | + auto result = httplib::stream::Get(cli, "/api/generate"); |
| 191 | + |
| 192 | + if (result && result.status() == 200) { |
| 193 | + for (auto chunk : result.body()) { |
| 194 | + std::cout << chunk << std::flush; |
| 195 | + } |
| 196 | + } |
| 197 | + |
| 198 | + // Check for connection errors |
| 199 | + if (result.read_error() != httplib::Error::Success) { |
| 200 | + std::cerr << "Connection lost\n"; |
| 201 | + } |
| 202 | + |
| 203 | + return 0; |
| 204 | +} |
| 205 | +``` |
| 206 | + |
| 207 | +### Example 3: Large File Download with Progress |
| 208 | + |
| 209 | +```cpp |
| 210 | +#include "httplib.h" |
| 211 | +#include <fstream> |
| 212 | +#include <iostream> |
| 213 | + |
| 214 | +int main() { |
| 215 | + httplib::Client cli("http://example.com"); |
| 216 | + auto result = httplib::stream::Get(cli, "/large-file.zip"); |
| 217 | + |
| 218 | + if (!result || result.status() != 200) { |
| 219 | + std::cerr << "Download failed\n"; |
| 220 | + return 1; |
| 221 | + } |
| 222 | + |
| 223 | + std::ofstream file("download.zip", std::ios::binary); |
| 224 | + size_t total = 0; |
| 225 | + |
| 226 | + for (auto chunk : result.body(65536)) { // 64KB chunks |
| 227 | + file.write(chunk.data(), chunk.size()); |
| 228 | + total += chunk.size(); |
| 229 | + std::cout << "\rDownloaded: " << (total / 1024) << " KB" << std::flush; |
| 230 | + } |
| 231 | + |
| 232 | + std::cout << "\nComplete!\n"; |
| 233 | + return 0; |
| 234 | +} |
| 235 | +``` |
| 236 | + |
| 237 | +### Example 4: Reverse Proxy Streaming |
| 238 | + |
| 239 | +```cpp |
| 240 | +#include "httplib.h" |
| 241 | + |
| 242 | +httplib::Server svr; |
| 243 | + |
| 244 | +svr.Get("/proxy/(.*)", [](const httplib::Request& req, httplib::Response& res) { |
| 245 | + httplib::Client upstream("http://backend:8080"); |
| 246 | + auto handle = upstream.open_stream("/" + req.matches[1].str()); |
| 247 | + |
| 248 | + if (!handle.is_valid()) { |
| 249 | + res.status = 502; |
| 250 | + return; |
| 251 | + } |
| 252 | + |
| 253 | + res.status = handle.response->status; |
| 254 | + res.set_chunked_content_provider( |
| 255 | + handle.response->get_header_value("Content-Type"), |
| 256 | + [handle = std::move(handle)](size_t, httplib::DataSink& sink) mutable { |
| 257 | + char buf[8192]; |
| 258 | + auto n = handle.read(buf, sizeof(buf)); |
| 259 | + if (n > 0) { |
| 260 | + sink.write(buf, static_cast<size_t>(n)); |
| 261 | + return true; |
| 262 | + } |
| 263 | + sink.done(); |
| 264 | + return true; |
| 265 | + } |
| 266 | + ); |
| 267 | +}); |
| 268 | + |
| 269 | +svr.listen("0.0.0.0", 3000); |
| 270 | +``` |
| 271 | + |
| 272 | +## Comparison with Existing APIs |
| 273 | + |
| 274 | +| Feature | `Client::Get()` | `open_stream()` | `stream::Get()` | |
| 275 | +|---------|----------------|-----------------|----------------| |
| 276 | +| Headers available | After complete | Immediately | Immediately | |
| 277 | +| Body reading | All at once | Direct from socket | Generator-based | |
| 278 | +| Memory usage | Full body in RAM | Minimal (controlled) | Minimal (controlled) | |
| 279 | +| Keep-Alive support | ✅ Yes | ❌ No | ❌ No | |
| 280 | +| Compression | Auto-handled | Auto-handled | Auto-handled | |
| 281 | +| C++ standard | C++11 | C++11 | C++20 | |
| 282 | +| Best for | Small responses, Keep-Alive | Low-level streaming | Modern streaming | |
| 283 | + |
| 284 | +## Features |
| 285 | + |
| 286 | +- **True socket-level streaming**: Data is read directly from the network socket |
| 287 | +- **Low memory footprint**: Only the current chunk is held in memory |
| 288 | +- **Compression support**: Automatic decompression for gzip, brotli, and zstd |
| 289 | +- **Chunked transfer**: Full support for chunked transfer encoding |
| 290 | +- **SSL/TLS support**: Works with HTTPS connections |
| 291 | +- **C++23 ready**: `Generator<T>` is compatible with `std::generator` interface |
| 292 | + |
| 293 | +## Important Notes |
| 294 | + |
| 295 | +### Keep-Alive Behavior |
| 296 | + |
| 297 | +The streaming API (`stream::Get()` / `open_stream()`) takes ownership of the socket connection for the duration of the stream. This means: |
| 298 | + |
| 299 | +- **Keep-Alive is not supported** for streaming connections |
| 300 | +- The socket is closed when `StreamHandle` is destroyed |
| 301 | +- For Keep-Alive scenarios, use the standard `client.Get()` API instead |
| 302 | + |
| 303 | +```cpp |
| 304 | +// Use for streaming (no Keep-Alive) |
| 305 | +auto stream = httplib::stream::Get(cli, "/large-stream"); |
| 306 | + |
| 307 | +// Use for Keep-Alive connections |
| 308 | +auto result = cli.Get("/api/data"); // Connection can be reused |
| 309 | +``` |
| 310 | + |
| 311 | +## Building |
| 312 | + |
| 313 | +Compile with C++20 support: |
| 314 | + |
| 315 | +```bash |
| 316 | +# GCC |
| 317 | +g++ -std=c++20 -o myapp myapp.cpp -lpthread -lssl -lcrypto |
| 318 | + |
| 319 | +# Clang |
| 320 | +clang++ -std=c++20 -o myapp myapp.cpp -lpthread -lssl -lcrypto |
| 321 | +``` |
| 322 | + |
| 323 | +## Related |
| 324 | + |
| 325 | +- [Issue #2269](https://github.com/yhirose/cpp-httplib/issues/2269) - Original feature request |
| 326 | +- [httplib.h](./httplib.h) - Main library (includes C++20 streaming API) |
| 327 | +- [example/ssecli-stream.cc](./example/ssecli-stream.cc) - SSE client with auto-reconnection |
0 commit comments