This library is a simple C++ wrapper for the C library http-parser [1] (This code was derived from the HTTP parser code in NGINX). http-parser is a simple HTTP streaming parser (for those of you familiar with XML, it works much like a SAX parser). It knows nothing of sockets or streams. You feed it data and it invokes registered callbacks to notify of available data. Because http-parser operates at the very lowest level, it does not buffer data or allocate memory dynamically. This library attempts to make that library easily usable by C++ programs by interpreting those callbacks and buffering data where needed.

The API for defined classes is documented using Doxygen [2]. You will need to run Doxygen from the project source folder to generate the output HTML.

Compiled HTML documentation for official releases is available online. Check the project page.

This project does not distribute the code to http-parser directly. To fetch the entire source code, make sure you fetch submodules [3] too:

$ git clone ...
$ cd httpxx
$ git submodule init
$ git submodule update

http-parser itself has no dependencies and compiles with C++ compilers. httpcxx uses only standard library facilities (namely std::string and std::map) and introduces no additional dependencies.

The code should compile as is under a standard-compliant C++03 implementation.

A good memory allocation policy is important in server programs, which typically run for a long time and suffer from memory fragmentation. httpcxx does its best to avoid repeated allocation, but it needs a little help on your part.

http::Request and http::Response parser object allocate memory as required because they buffer different parts of the incoming HTTP request/response in std::string instances. However, they are implemented carefully as to use the growing property of std::string [4] to their advantage. In particular, you may re-use http::Request and http::Response parser objects for parsing multiple request/response objects using their .clear() method. This method marks all header lengths as 0 but keeps the instances as well as the map. All this ensures that parsers avoid repeated memory allocation.

Check out the sample programs in the demo/ subfolder.