lightconf is a lightweight header-only configuration file library for C++11. It is designed to be (almost) compatible with JSON.

• Header-only means it can be copied directly into your project and used without introducing any deployment issues.
• Hierarchical configuration files with a simple path-based interface for accessing nested properties.
• The primary lightconf text format .config is structurally identical to JSON but supports comments. Comments, blank lines, and ordering will be preserved when updating an existing file.
• It can also read and write JSON directly.
• Easily extensible to allow serializing custom types and enums automatically.
• BSD licensed.

Certain features used in lightconf, such as decltype and variadic templates, mean that it is only usable by compilers that support C++11. It has been tested using Clang 3.2 and GCC 4.7.2.

There are various common ways to implement configuration files:

• JSON is a common text format for storing structured data. The drawback of using a JSON library directly is that most JSON libraries do not abstract away the physical structure of a JSON document, and still require explicitly drilling down and extracting values from the document, especially when dealing with custom types. lightconf is not a replacement for JSON, but rather a more specialized interface that is capable of reading and writing both JSON and a slightly more human-friendly format called .config.
• XML suffers from the same problems as using a JSON library with the added problem of being extremely verbose.
• Boost.PropertyTree is a nice library similar to lightconf that supports JSON among other formats. The primary drawback is its dependency on Boost.
• INI is a very simple flat key/value format. It does not support hierarchical or vector data.

The text-based file format that lightconf is designed to use for primary serialization and deserialization is referred to internally as .config (simply to differentiate it from JSON, the file extension is not important). It differs from JSON in the following ways:

• Keys are bare identifiers rather than strings, and are of the form [A-Za-z][A-Za-z0-9_\-]* (if a JSON file has keys which don't match this pattern, an exception will be thrown when trying to deserialize it)
• Keys are separated from values by an = rather than a colon
• Commas (including trailing commas) in lists and groups are optional
• Comments are supported (JSON with comments will read as well, but the comments will not be preserved when saving a JSON file)
• JSON supports documents with arrays as root elements, but .config files always have a group (an object in JSON) as a root, with no braces

A full working example is available in sample/lightconf_sample.cpp. This is a very quick example of how to use lightconf.

The primary object containing configuration information is lightconf::group. It acts like a map from string keys to lightconf::value values, and is analogous to JSON's objects. A value is effectively a union type which can contain the standard JSON types: doubles, strings, bools, arrays (implemented as std::vector<lightconf::value>) and other groups.

Values are accessed via paths, which are period separated names corresponding to nested groups. The path foo.bar.baz corresponds to the value 123 in the configuration layout foo = { bar = { baz = 123 } }.

// Deserialize a configuration file stored in a string
lightconf::group config_group = lightconf::config_format::read(source);

// Retrieve a value from the configuration group (throwing an exception if it doesn't exist)
int foo_count = config_group.get<int>("counts.foo_count");

// Retrieve a value from the configuration group (returning a default value if it doesn't exist)
bool is_bar = config_group.get<bool>("flags.is_bar", false);

// Set a value in the configuration group, creating parent groups
config_group.set<double>("constants.transcendental.pi", 3.14159);

// Typed vectors and tuples can also be stored and retrieved, as long as the inner types can be stored
config_group.set<std::vector<string>>("names", { "Fred", "George", "Stephen" });
std::vector<string> names = config_group.get<std::vector<string>>("names");
config_group.set<std::tuple<int, bool>>("my_tuple", std::make_tuple<int, bool>(10, false));
std::tuple<int, bool> my_tuple = config_group.get<std::tuple<int, bool>>("my_tuple");

// Generate a new configuration file using the original one as a base, and attempting to keep 
// lines under 80 characters long by wrapping lists and groups that exceed that length
std::string new_source = lightconf::config_format::write(config_group, source, 80);

// We can also create a JSON version
std::string json_source = lightconf::json_format::write(config_group);

Custom types and enums can also be serialized directly by adding specializations of the template class lightconf::value_type_info. See sample/lightconf_sample.cpp for an example.