Abstract:
Although scripting languages have become very popular, even mature scripting language
implementations remain interpreted. Several compilers and reimplementations
have been attempted, generally focusing on performance.
Based on our survey of these reimplementations, we determine that there are three
important features of scripting languages that are difficult to compile or reimplement.
Since scripting languages are defined primarily through the semantics of their original
implementations, they often change semantics between releases. They provide C APIs,
used both for foreign-function interfaces and to write third-party extensions. These
APIs typically have tight integration with the original implementation, and are used to
provide large standard libraries, which are difficult to re-use, and costly to reimplement.
Finally, they support run-time code generation. These features make it difficult to
design a fully compatible compiler.
We present a technique to support these features in an ahead-of-time compiler for
PHP. Our technique uses the original PHP implementation through the provided C
API, both in our compiler and in our generated code. We support all of these important
scripting language features. Additionally, our approach allows us to automatically
support limited future language changes. We present a discussion and performance
evaluation of this technique.