Manipulating Object and Value Types


The inject-code node inserts the given code into the generated code for the given type or function, and it is a child of the object-type, value-type, modify-function and add-function nodes.

The code can be embedded into XML (be careful to use the correct XML entities for characters like ‘<’, ‘>’, ‘&’):

    <inject-code class="native | target | target-declaration"
        position="beginning | end" since="...">
        // the code

or obtained from an external file:

    <inject-code class="native | target | target-declaration"
        position="beginning | end" since="..."

The class attribute specifies which module of the generated code that will be affected by the code injection. The class attribute accepts the following values:

  • native: The c++ code

  • target: The binding code

  • target-declaration: The code will be injected into the generated header file containing the c++ wrapper class definition.

  • file: The file name

  • snippet: The snippet label (optional)

If the position attribute is set to beginning (the default), the code is inserted at the beginning of the function. If it is set to end, the code is inserted at the end of the function.

The since attribute specify the API version where this code was injected.

If a snippet label is given, the code between annotations of the form

// @snippet label
// @snippet label

will be extracted.


The modify-field node allows you to alter the access privileges for a given C++ field when mapping it onto the target language, and it is a child of an object-type or a value-type node.

    <modify-field name="..."
        write="true | false"
        read="true | false" />

The name attribute is the name of the field, the optional write and read attributes specify the field’s access privileges in the target language API (both are set to true by default). The remove attribute is an optional attribute, which can mark the field to be discarded on generation; it has the same purpose of the deprecated tag remove.


The modify-function node allows you to modify a given C++ function when mapping it onto the target language, and it is a child of an object-type or a value-type node. Use the modify-argument node to specify which argument the modification affects.

    <modify-function signature="..."
                     remove="all | c++"
                     access="public | private | protected"
                     allow-thread="true | auto | false"
                     exception-handling="off | auto-off | auto-on | on"
                     rename="..." />

The signature attribute is a normalized C++ signature, excluding return values but including potential const declarations.

The since attribute specify the API version when this function was modified.

The allow-thread attribute specifies whether a function should be wrapped into Py_BEGIN_ALLOW_THREADS and Py_END_ALLOW_THREADS, that is, temporarily release the GIL (global interpreter lock). Doing so is required for any thread-related function (wait operations), functions that might call a virtual function (potentially reimplemented in Python), and recommended for lengthy I/O operations or similar. It has performance costs, though. The value auto means that it will be turned off for functions for which it is deemed to be safe, for example, simple getters. The attribute defaults to false.

The exception-handling attribute specifies whether to generate exception handling code (nest the function call into try / catch statements). It accepts the following values:

  • no, false: Do not generate exception handling code

  • auto-off: Generate exception handling code for functions declaring a non-empty throw list

  • auto-on: Generate exception handling code unless function declares noexcept

  • yes, true: Always generate exception handling code

The remove, access and rename attributes are optional attributes for added convenience; they serve the same purpose as the deprecated tags remove, access and rename.


The add-function node allows you to add a given function onto the target language, and it is a child of an object-type or value-type nodes if the function is supposed to be a method, or namespace-type and Including Snippets if the function is supposed to be a function inside a namespace or a global function.

Typically when adding a function some code must be injected to provide the function logic. This can be done using the inject-code node.

    <add-function signature="..." return-type="..." access="public | protected" static="yes | no" since="..."/>

The return-type attribute defaults to void, the access to public and the static one to no.

The since attribute specify the API version when this function was added.

Within the signature, names for the function parameters can be specified by enclosing them within the delimiter @:

void foo(int @parameter1@,float)


The conversion-rule node allows you to write customized code to convert the given argument between the target language and C++, and is a child of the value-type, object-type, primitive-type and container-type nodes.

The code pointed by the file attribute is very tied to the generator using APIExtractor, so it don’t follow any rules, but the generator rules..

<value-type name="Foo">
    <convertion-rule file="my_converter_implementation.h" since="..."/>

The since attribute specify the API version when this conversion rule became valid.


You can also use the conversion-rule node to specify how the conversion of a single function argument should be done in a function.

The file and snippet attributes are also supported (see inject-code nodes).