The Very Slow Jython Project

A project to re-think implementation choices in the Jython core, through the gradual, narrated evolution of a toy implementation.

Contents

• 1. Background to the Very Slow Jython Project
  • 1.1. Background to the project
  • 1.2. Organisation of Sources
• 2. A Tree-Python Interpreter
  • 2.1. Some Help from the Reference Interpreter
    • 2.1.1. Orientation
    • 2.1.2. Python Compilation Example
  • 2.2. An Interpreter in Java for the Python AST
    • 2.2.1. Representing the AST in Java
    • 2.2.2. Generating a Java Tree Literal
    • 2.2.3. A Visit from the Evaluator
  • 2.3. Type and Operation Dispatch
    • 2.3.1. Operation Dispatch in CPython
    • 2.3.2. Operation Dispatch in Jython 2.7.1
    • 2.3.3. A Java Object Approach
    • 2.3.4. Dispatch via overloading in Java
    • 2.3.5. Dispatch via a Java MethodHandle
    • 2.3.6. Caching Type Decisions
    • 2.3.7. Extension to Unary Operators
    • 2.3.8. Dispatch with Multiple Implementations
    • 2.3.9. Progress so far
  • 2.4. Simple Statements and Assignment
    • 2.4.1. Aims of this Section
    • 2.4.2. Additions to the AST
    • 2.4.3. A Look at Variable Scope in CPython
    • 2.4.4. Generating the Layout
    • 2.4.5. Extending the AST Interpreter
    • 2.4.6. Efficient load and store
    • 2.4.7. Optimisation of Function Calls
• 3. A Generated Code Interpreter
  • 3.1. Introduction
    • 3.1.1. Concepts Banked
    • 3.1.2. Concepts to be Explored
  • 3.2. An Interpreter for CPython Byte Code
    • 3.2.1. PyObject as an Interface
    • 3.2.2. Basic types int, float and str
    • 3.2.3. Objects for bytes and tuple
    • 3.2.4. Implementing dict
    • 3.2.5. A Python exception as a Java Throwable
    • 3.2.6. Runtime Support: Py
    • 3.2.7. Implementing code
    • 3.2.8. Implementing frame
    • 3.2.9. The frame as an Interpreter
    • 3.2.10. Program Examples
  • 3.3. Type and Arithmetic Operations
    • 3.3.1. Representing a Python Class
    • 3.3.2. A Unary Operation UNARY_NEGATIVE
    • 3.3.3. A Binary Operation BINARY_ADD
    • 3.3.4. How we Fill the Slots
  • 3.4. Sequences and Indexing
    • 3.4.1. Tuple Creation and Indexing
    • 3.4.2. List Creation and Indexing
  • 3.5. Built-in Inheritance
    • 3.5.1. The Example of bool
    • 3.5.2. Inheritance will not be Free
    • 3.5.3. A Simplified MRO
    • 3.5.4. Hints from the C Implementation
    • 3.5.5. A bool Implementation
    • 3.5.6. Exceptions and Inheritance
  • 3.6. Comparison and Loops
    • 3.6.1. Motivating Example
    • 3.6.2. Branch instructions
    • 3.6.3. Rich Comparison
  • 3.7. Refactoring for evo3
    • 3.7.1. Static Factory Methods in Py
    • 3.7.2. Some Re-naming
    • 3.7.3. Re-working PyType
    • 3.7.4. Lightweight EmptyException
    • 3.7.5. Type Cast in the Method Handle
  • 3.8. Function Definition and Call
    • 3.8.1. Calling a Function
    • 3.8.2. Defining a Function in Java
    • 3.8.3. Defining a Function in Python
  • 3.9. Refactoring for evo4
    • 3.9.1. Another Re-work of Core Types
    • 3.9.2. Slot Functions
    • 3.9.3. Java Signatures of Slots
  • 3.10. Access to Attributes
    • 3.10.1. Exploration
    • 3.10.2. The Mechanism of Attribute Access
    • 3.10.3. A Glance up the Mountain
  • 3.11. Attributes Defined in Java
    • 3.11.1. Access to Attributes: A Trivial Case
    • 3.11.2. Class and Instance Improvements
    • 3.11.3. Descriptors in Class Creation (Java)
    • 3.11.4. Integrating the Parts
  • 3.12. Attributes Defined in Python
    • 3.12.1. Access to Attributes: A Trivial Case
    • 3.12.2. Class and Instance Improvements
    • 3.12.3. Descriptors in Class Creation (Python)
    • 3.12.4. Integrating the Parts
• 4. A Plain Java Object Interpreter
  • 4.1. Introduction
    • 4.1.1. Motivations
    • 4.1.2. Design Concept
    • 4.1.3. Feasibility
  • 4.2. Operations on Built-in Types
    • 4.2.1. Direction of Travel
    • 4.2.2. PyType and Operations for float
    • 4.2.3. PyType and Operations for int and bool
    • 4.2.4. Continuing on Course
    • 4.2.5. Some Dots on the RADAR
  • 4.3. Hash, Lookup and Dictionary
    • 4.3.1. The Type str
    • 4.3.2. Plain Object Keys and dict
    • 4.3.3. Attribute Names
    • 4.3.4. The Dictionary of a PyType
  • 4.4. Methods on Built-in Types
    • 4.4.1. The Visible Classes
    • 4.4.2. Relationships amongst the Classes
    • 4.4.3. Design Features
    • 4.4.4. Annotations identifying Python Methods
  • 4.5. Python Modules in Java
    • 4.5.1. Objectives
    • 4.5.2. Module Data Structures
    • 4.5.3. An Attempt at builtins
  • 4.6. Creation and Initialisation
    • 4.6.1. type.__call__
    • 4.6.2. __new__
• 5. The Plain Java Object Model Extended
  • 5.1. Introduction
    • 5.1.1. Java Classes as Python Types
    • 5.1.2. Special Methods
    • 5.1.3. A Jython API using Java modules
  • 5.2. Object and PyType Basic Patterns
    • 5.2.1. Concepts
    • 5.2.2. Class Models for the Concepts
  • 5.3. Instance Models of object and type
    • 5.3.1. Representing list
    • 5.3.2. Object, object and Python class
    • 5.3.3. Type Objects for type
    • 5.3.4. Representing float
    • 5.3.5. Possibility of Caching on the Representation
    • 5.3.6. Representing int and bool
    • 5.3.7. Summary Examples
  • 5.4. Subclasses in Python
    • 5.4.1. Implementation to Pull from VSJ3
    • 5.4.2. Exceptions
    • 5.4.3. Subclasses Defined as in Python
  • 5.5. Instance Models for Java Objects
    • 5.5.1. The Challenge of Subclasses
  • 5.6. Type System Initialisation
    • 5.6.1. A Thread-Safe Type System
    • 5.6.2. Specifying a Python Type
• 6. Performance
  • 6.1. Unary Operations
    • 6.1.1. VSJ 2 evo 4
    • 6.1.2. Jython 2.7.2
    • 6.1.3. VSJ 2 evo 4 with invokedynamic
    • 6.1.4. VSJ 3 evo 1
    • 6.1.5. VSJ 3 evo 1 with invokedynamic
    • 6.1.6. Analysis
  • 6.2. Binary Operations
    • 6.2.1. VSJ 2 evo 4
    • 6.2.2. Jython 2.7.2
    • 6.2.3. VSJ 2 evo 4 with invokedynamic
    • 6.2.4. VSJ 3 evo 1
    • 6.2.5. VSJ 3 evo 1 with invokedynamic
    • 6.2.6. Analysis
  • 6.3. On the Validity of the Benchmarks
    • 6.3.1. Reliability of Timings
    • 6.3.2. Vanishing Time in Primitive Operations
• 7. Architecture
  • 7.1. Java Object as Python object
    • 7.1.1. Motivations
    • 7.1.2. Essential Idea
    • 7.1.3. Canonical Implementation
    • 7.1.4. Adopted Implementation
    • 7.1.5. Found Java Type
  • 7.2. Object Implementation
    • 7.2.1. Special Methods in Java
    • 7.2.2. Special Methods in Python
    • 7.2.3. __new__ and Java Constructors
    • 7.2.4. Relation of tp_new to the Java constructor
    • 7.2.5. Examples guiding architectural choices
  • 7.3. Type Slots
    • 7.3.1. Choice of MethodHandle
    • 7.3.2. Learning from CPython
    • 7.3.3. One-to-One Principle
    • 7.3.4. Potentially Problematic Slots in CPython
    • 7.3.5. Initialisation of Slots
  • 7.4. Access to Attributes
    • 7.4.1. A Reminder of the CPython Opcodes
    • 7.4.2. Slots for Attribute Access
    • 7.4.3. Abstract API support for attributes
    • 7.4.4. The Mechanism of Attribute Access
  • 7.5. Interpreter Structure
    • 7.5.1. Introduction
    • 7.5.2. Runtime, Thread and Interpreter (CPython)
    • 7.5.3. Use cases
    • 7.5.4. Proposed Model for Jython [untested]
  • 7.6. Types of code and frame Object
    • 7.6.1. Several Types of Code
    • 7.6.2. Old Notes (needs rewrite)
  • 7.7. Compiled Code (Python to JVM)
  • 7.8. Functions Defined in Java
    • 7.8.1. Introduction
    • 7.8.2. Patterns of Method Definition
    • 7.8.3. Argument and Return Value Coercion
  • 7.9. Descriptors
    • 7.9.1. Introduction
    • 7.9.2. Descriptors and __getattribute__
    • 7.9.3. Methods in Python (PyFunction)
    • 7.9.4. The @classmethod Decorator (PyClassMethod)
    • 7.9.5. The @staticmethod Decorator (PyStaticMethod)
    • 7.9.6. Descriptors for Built-in Types
    • 7.9.7. Members (PyMemberDescr)
    • 7.9.8. Attributes (PyGetSetDescr)
    • 7.9.9. Built-in Methods (PyMethodDescr)
    • 7.9.10. Special Methods (PyWrapperDescr)
    • 7.9.11. Built-in Class Methods (PyClassMethodDescr)
    • 7.9.12. Built-in Static Methods (PyStaticMethod)
• 8. Writing Code
  • 8.1. Coding Standard
    • 8.1.1. Divergence from Jython 2 Standard
  • 8.2. Porting CPython Code to Java
    • 8.2.1. Architectural Drivers
    • 8.2.2. Lessons from Practice

Indices and tables

• Index

• Module Index

• Search Page