wrenpp/README.md

274 lines
9.2 KiB
Markdown

# Wrenpp #
Wrenpp is a lightweight C++ wrapper for the official [Wren](http://wren.io/) C
library. It aims to package all the features accessible through the C API in a
modern C++ interface that is easy to use, while keeping the overhead to a
minimum.
* [Current state](#current-state)
* [Philosophy](#philosophy)
* [Building](#building)
* [Wrenpp](#wrenpp)
* [Wren only](#wren-only)
* [Usage](#usage)
* [Meson](#meson)
* [C++](#c)
* [Header files](#header-files)
* [Initialisation](#initialisation)
* [Foreign functions](#foreign-functions)
* [Implementation](#implementation)
## Current state ##
The library is functional and provides access to most of the features of Wren.
Some are better tested than others. There are no known memory leaks. Please
keep in mind this is an early version and more testing and work will be needed.
## Philosophy ##
The overall principle behind this wrapper is to be feature rich while also
being lightweight and allowing fine-grained control for users who need it.
By design, wrenpp hides the Wren C API entirely, but all functions (considered
"low level" relative to this wrapper) are still accessible through the `VM`
class. If the function you need is not available there it's probably just
missing and adding it should be a trivial task.
Wrenpp provides you with convenience functions in `vm_fun.hpp` that are built
on top of the `VM` class and make basic use of Wren simpler. You're still free
to "mix & match" higher level functions with lower level ones, and you can
decide to not use portions of the wrapper entirely.
## Building ##
### Wrenpp ###
You can make a release build like this:
```bash
mkdir build
cd build
meson setup -Dbuildtype=release ..
ninja
```
The build scripts understand the following options (double check in
`meson_options.txt`):
1. `build_testing` set to true to compile Wren's unit tests too
2. `build_examples` set to true to include examples from the `examples/`
directory in the build
3. `wren_with_cli` set to false to exclude the command line tool from the Wren
subproject build
4. `wren_with_rand` set to true to compile Wren with
[random](http://wren.io/modules/random/) support
5. `wren_with_meta` set to true to compile Wren with
[meta](http://wren.io/modules/meta/) support
6. `wrenpp_with_name_guessing` enable class name guessing in c++ code. It
currently affects only example code and header files, not the library
itself. It controls the `WRENPP_WITH_NAME_GUESSING` preprocessor symbol.
When it's defined, you will have access to better error reporting and extra
functions, such as a `make_wren_object()` overload that doesn't require you
to specify a string class name.
Note that some example projects require random support in Wren, without which
they will crash. If you want to run examples or you are getting "Address
boundary error" segfaults recompile Wrenpp after enabling the example and/or
meta options:
```
mkdir build
cd build
meson setup -Dbuildtype=debug -Dbuild_examples=true -Dwren_with_rand=true -Dwren_with_meta=true ..
ninja
```
### Wren only ###
If you only want to compile the C library of Wren, of course you can do that:
```
mkdir build
cd build
meson setup -Dbuildtype=release -Dwren_with_rand=true -Dwren_with_meta=true ../subprojects/wren
ninja
```
This will build Wren as a shared object. If you want a static library instead
you can add the `-Ddefault_library=static` parameter to your Meson command, or
`-Ddefault_library=both` if you want tboth the shared and the static version of
Wren.
## Usage ##
### Meson ###
You can use Wrenpp as a Meson
[subproject](https://mesonbuild.com/Reference-manual.html#subproject) like you
would do normally. For example:
```
wrenpp = subproject('wrenpp', default_options: ['wren_with_rand=true'])
wrenpp_dep = wrenpp.get_variable('wrenpp_dep')
executable('my_project',
'main.cpp',
dependencies: wrenpp_dep,
)
```
This will make Wrenpp build as part of your project with Wren's optional random
module enabled.
Alternatively you can use the subproject as a fallback if there is no Wrenpp
installed in your system already:
```
wrenpp_dep = dependency('wrenpp', version: '>=0.1.0',
fallback: ['wrenpp', 'wrenpp_dep'],
default_options: ['wren_with_rand=true'],
)
```
Note that when you use Wrenpp as a subproject, Wrenpp's subproject Wren will
become a sub-subproject of your project. This is how Meson works and it simply
means that in your top level source directory you will have to run this command
before you will be able to compile (Meson should detect and print this as
well):
```
meson wrap promote subprojects/wrenpp/subprojects/wren
```
### C++ ###
For working examples refer to the source files in the `examples/` directory.
"Greet" is the simplest example, while "math_vector" is showing off more
features.
#### Header files ####
In order to use Wrenpp in your project you will need to include `wrenpp/vm.hpp`
at least. A description of all the relevant header files follows:
* `vm.hpp` The `VM` class provides simple wrappers around the C API functions.
It wraps and hides the `WrenVM` pointer and provides RAII around it.
Available methods aim to be as close to their C counterparts as possible. You
can use objects of this class in a way that mirrors very closely the way you
would use the C API.
* `vm_fun.hpp` This header provides functions on top of the `VM` class. Their
goal is to make common operations easier and more compact in your code.
* `def_configuration.hpp` The `DefConfiguration` struct extends the bare
`Configuration` struct by providing an implementation for `write_fn`,
`error_fn` and `reallocate_fn`. They are implemented in terms of `std::cout`,
`std::cerr` and `new[]`/`delete[]` respectively. You can safely ignore this
header if this is not suitable for your project.
Other header files are intended for Wrenpp's use and you shouldn't need to use
them. You are free to use them in your project if you think the code inside is
useful to you, however keep in mind that those are not providing core Wrenpp
functionalities and may be removed or modified in the future.
#### Initialisation ####
Wrenpp aims to make your code as clear and concise as possible. To instantiate
a `VM` (which wraps `WrenVM`) and run a simple script you can do the following:
```
#include <wrenpp/vm.hpp>
#include <wrenpp/def_configuration.hpp>
int main() {
wren::DefConfiguration config;
wren::VM vm(&config, nullptr);
vm.interpret("main", "System.print(\"Hello world!\")");
return 0;
}
```
#### Foreign functions ####
You have two options to register foreign functions with wrenpp: the easiest one
is to just use the callback manager available through the `VM` object:
```
#include <wrenpp/vm.hpp>
#include <wrenpp/def_configuration.hpp>
#include <wrenpp/callback_manager.hpp> //necessary if you want to use the callback manager
int main() {
wren::DefConfiguration config;
wren::VM vm(&config, nullptr);
vm.callback_manager()
.add_callback(true, "your_module", "YourClass", "your_method(_,_)", &your_method);
vm.interpret("main", your_fun_script_here);
return 0;
}
```
The current implementation available in `DefConfiguration` registers a callback
to Wren that looks up a `std::unordered_map` for the correct c++ function and
then forwards the call to it. This should be good enough for most cases,
however you are free to provide your own callback for wren by writing your own
Configuration class. Wrenpp tries to not impose you any particular helper code,
in fact if you don't wish to use the `CallbackManager` returned by
`vm.callback_manager()` then you don't even need to include the relative header
file! There will still be a `std::unordered_map` in memory (because that's
owned by `vm`) but it will never be used.
The second method requires you to provide the full Wren foreign functions
callback:
```
#include <wrenpp/vm.hpp>
#include <wrenpp/def_configuration.hpp>
class MyConf : public wren::DefConfiguration {
public:
//This is the callback that Wren will call, refer to the C API documentation
//for more details.
wren::foreign_method_t foreign_method_fn(
wren::VM* vm,
std::string_view module,
std::string_view class_name,
bool is_static,
std::string_view signature
) {
if (module == "your_module" and class_name == "YourClass") {
if (is_static and signature == "your_method(_,_)")
return &your_method;
}
return nullptr;
}
}
int main() {
MyConf config;
wren::VM vm(&config, nullptr);
vm.interpret("main", your_fun_script_here);
return 0;
}
```
The above code is very close to what you'd do with the C API. If you need, you
can even pass a user data pointer to the `VM` constructor (`nullptr` in this
snippet) and retrieve it later from within `foreign_method_fn()` via
`vm.void_user_data()` or `vm.user_data()`. You're unlikely to need it as, as
you can see, `foreign_method_fn()` is not static and so you can have some state
stored in your `MyConf` object, however you still get the chance to use the
C-like user data mechanism if you wish to.
Note that for the above to work the `foreign_method_fn()` method in `MyConf`
must be named exactly like that, and the signature must match (except for the
`static` and `const` keywords). Ideally the code above should alse be
`noexcept` as there will be C code in the call stack when that method is
invoked.
## Implementation ##
*Work in progress*