PyHP

PyHP is an (incomplete!) implementation of the PHP language using the RPython technology. It uses JIT transformer provided by the RPython library to achieve good performance.

PyHP stands for Python + PHP, as RRpython is a subset of the Python language.

Highly unstable and work in progress.

HippyVM is too built on top of RPython, and is a complete, working implementation.

Features

• all variable types like int, float, string, array, iterator
• if, while, for, foreach statements
• functions
• pass by value or by reference to a function
• global variables in function blocks using global
• global constants using define()
• (incomplete) standard library functions

Project structure

• main.py - main entry point
• sourceparcer.py - parses a given PHP file using the grammar.txt definition, and produces an AST tree consisting of operations.py nodes
• bytecode.py - turns the AST tree produced by sourceparcer.py into bytecode by calling compile() on the tree. Produces a ByteCode instance consisting of opcodes.py nodes
• frame.py - execution frame. Contains the stack and the variables/functions map for the function or the global program. A frame instance is passed to execute method of a ByteCode instance as the only parameter
• grammar.txt - EBNF PHP grammar used by sourceparcer.py

Additional files:

• operations.py - AST tree nodes
• opcodes.py - class per each opcode. Each opcode has a eval(frame) method which gets called by the Bytecode.execute() method
• symbols.py - contains an optimized Map class used for symbols map in scopes.py
• stdlib.py - various PHP standard library methods like strlen
• functions.py - wraps a ByteCode instance or a native function from stdlib.py into an object which then gets used to run a function or the main program
• datatypes.py - all datatypes' box classes used to store the variables, like int, float, array, etc.

Building

Starting the VM

vagrant up
vagrant ssh
cd /var/www/pyhp

Building the interpreter

rpython -Ojit targetpyhp.py

Emit the -Ojit to compile an interpreter without JIT support. Speeds up the build process by 5x, but the produced interpreter runs much slower.

Running the interpreter

./pyhp-c fibonacci.php

Debugging the interpreter

PYPYLOG=jit-log-opt:jit.txt ./pyhp-c fibonacci.php

Plot the trace as a graph

PYTHONPATH=$PYTHONPATH:/home/vagrant/pypy-src/ python ~/pypy-src/rpython/tool/logparser.py draw-time jit.txt --mainwidth=8000 filename.png

Running tests

PYTHONPATH=$PYTHONPATH:/home/vagrant/pypy-src/ py.test tests

Attributions and inspirations

JavaScript on top of RPython

Example interpreter

Pascal on top of RPython

HippyVM - php on top of RPython