D5 · Translation

Spec reference: Section D - Programming Languages
Key idea: Understand why and how high-level code is translated into machine code.

---

Why do programs need to be translated?

Computers only understand machine code - binary instructions (0s and 1s) specific to a processor's instruction set. Programmers write code in high-level languages (Python, Java, C++) that are readable by humans but cannot be directly executed by a CPU.

Translation converts source code into machine code so the computer can run it.

---

Types of programming language

Level	Description	Example
Machine code	Binary instructions (0s and 1s) directly executed by CPU	10110000 01100001
Assembly language	Human-readable mnemonics representing machine code instructions	MOV AL, 61h
Low-level language	Close to hardware - assembly	Assembly
High-level language	Human-readable, abstracted from hardware	Python, Java, C#, JavaScript

---

Three types of translator

1. Assembler

An assembler translates assembly language into machine code.

Assembly: MOV AX, 5    →    Assembler    →    Machine code: 10111000 00000101

• One assembly instruction → one (or a few) machine code instructions.
• Used for low-level systems programming, device drivers, embedded systems.

---

2. Compiler

A compiler translates an entire high-level language program into machine code all at once, before execution.

Source code (Python/C++)  →  Compiler  →  Executable file (.exe)
                                              ↓
                                    Run later without re-translating

How compilation works:

1. Lexical analysis - tokenises the source code (breaks into keywords, identifiers, operators).
2. Syntax analysis - checks grammar rules (produces a parse tree).
3. Semantic analysis - checks meaning (are variables declared? correct types?).
4. Code generation - produces machine code or intermediate code.
5. Optimisation - improves efficiency of the generated code.

Languages that use compilation: C, C++, Rust, Go.

---

3. Interpreter

An interpreter translates and executes one line at a time, converting each instruction to machine code immediately before running it.

Source code → [Interpreter reads line 1] → [Executes line 1]
           → [Interpreter reads line 2] → [Executes line 2]
           → ... and so on

Languages that use interpretation: Python, JavaScript (in browser), Ruby, PHP.

---

Compiler vs Interpreter - comparison

	Compiler	Interpreter
Translates	Entire program at once	One line at a time
Output	Executable file (.exe)	No separate output file
Execution speed	Fast (translated once, run many times)	Slower (translated during every run)
Error reporting	All errors reported at once after analysis	Stops at the first error encountered
Portability	Compiled code is platform-specific	Source code can run anywhere with interpreter
Development	Slower edit-compile-run cycle	Faster to test - run immediately
Examples	C, C++, Rust	Python, JavaScript, Ruby

---

Reasons for translation

Programs need to be translated because:

1. CPUs only understand machine code - binary instructions.
2. High-level languages are hardware-independent - the same Python code runs on Windows, Mac, Linux.
3. Human productivity - writing in Python is vastly more productive than writing in binary.
4. Portability - write once, translate for different platforms.

---

Benefits of using high-level languages

Benefit	Detail
Readability	Code resembles English - easier to write, read and maintain
Portability	Same source code can be translated for different hardware
Abstraction	Hardware details are hidden - focus on solving the problem
Productivity	One high-level statement may replace many machine code instructions
Rich libraries	Access to huge collections of pre-written functions

---

Drawbacks of translation

Drawback	Detail
Compilation overhead	Compilation step adds time before execution
Interpreted programs are slower	Translation on-the-fly is slower than pre-compiled code
Platform dependency	Compiled code may not run on a different CPU architecture
Source code exposure	Interpreted programs need the source code present - easier to copy

---

Just-In-Time (JIT) compilation

Some languages (Java, C#, JavaScript in V8) use JIT compilation - a hybrid approach:

1. Source code is compiled to bytecode (intermediate, platform-independent).
2. At runtime, a JIT compiler converts bytecode to machine code just before it is needed.
3. Frequently-run code is compiled once and cached, giving near-native speed.

Java source → Java compiler → Bytecode (.class)
                                    ↓
                            JVM + JIT compiler
                                    ↓
                            Machine code (at runtime)

---

Summary

Translator	Input	Output	When?
Assembler	Assembly language	Machine code	Before execution
Compiler	High-level source	Executable (machine code)	Before execution
Interpreter	High-level source	Executes directly	During execution
JIT compiler	Bytecode	Machine code	During execution (optimised)

---

Test Yourself

Question 1 of 5

Why must high-level programs be translated?