Differences in design and implementation between C and Python

[toaction]

LLPkgStore was originally designed for the C language, but has now implemented preliminary support for the Python language. What are the differences in design and implementation logic for different languages?

[toaction]

Following the Open-Closed Principle, when adding support for Python packages we largely followed the original design and processing logic. For simplification and due to Python's characteristics, we made changes to version management and dependency management:

Version Management

When managing LLPkg packages for C, there is an llpkgstore.json file that maps LLPkg package versions to native C library versions. The purpose is to let users automatically obtain the corresponding LLPkg package via the llgo get command by specifying the C library version. Usage:

llgo get clib@cversion

llgo get LLPkg@LLPkg_version

However, this feature turned out to be of limited value, so it was removed for Python package management.

Dependency Management

For C packages, we leverage Conan to obtain the dependency information of the C library and generate a dependency declaration file [clib]_autogen_link.go:

package libxml2

import (
    _ "github.com/goplus/lib/c"
    _ "github.com/goplus/lib/c/os"
)

const LLGoPackage string = "link: $(pkg-config --libs libxml-2.0);"

By using imports with blank identifiers, the dependencies of the C library are delegated to Go for management.

There are two reasons why dependency management is needed for C:

1. When generating LLGo bindings for a C library, all types are mapped. If one library uses types from another library, you must first obtain LLGo bindings for that other library.
2. For user convenience, the LLPkg corresponding to a C library includes .pc files that are published in GitHub Releases. To use them, the dependent .pc files must also be downloaded automatically.

In Python, however, all object types can be mapped to py.Object, so extensive type conversion is generally unnecessary. In addition, if users want to use the LLPkg for Python, a Python environment must be present, and dependency management can be fully handled by pip.