People in the programing languages field use fancy buzzwords like translator, transpiler, compiler, interpreter, JIT (just in time) compiler. Let’s figure out what they mean. But before let’s create some sort of programming language hierarchy:
Translator is a general word for transpilers, compilers and interpreters. It’s a tool converting one code to another one (low-level or high level).
Compilers create machine code before runtime, which they save into binary files. Each processor1 requires a special compiler. Plus, compiler depends on a platform. Say, there is a compiler for C++ targeted for the Linux OS and a particular processor.
Benefits of compilers are:
But there is a considerable drawback for a compiled language. Every little change in the program requires recompilation of the whole module.
Languages that are compiled: C, C++, Haskell, Rust, Go
Interpreters, on the other hand, can stumble upon such error that compiler could remove, because they translate a program line by line, during runtime. Due to this translation thing going, a transpiler is initially slower than an already compiled code. But there is a catch, when running a program line by line, there is more information that can be extracted, because interpreter will see variable values that are impossible to analyse statically, before a program is run2. That is where JIT compiler comes into play. It finds runtime optimizations and identifies parts of code which are faster to compile and execute rather than execute line by line (a default way of interpreter).
Briefly, interpreters:
Interpreted languages: Python, Perl, Ruby
It is a tool that converts source code of one language to source code of the other language. Usually referred to as translators between two high level languages. Transpiler is called source to source compiler sometimes.
Also, there is a less frequently used term. Assemblers. Essentially, they are compilers for assembly language. Assemblers convert assembly code into machine code before runtime. I think it is a part of a compiler (compilation is a two-step process: translate language into assembly language, then translate assembly into machine language).
Java’s source code is compiled into a binary byte code. This code runs on JVM (technically not a processor), that usually is a blazingly fast interpreter (and often with a JIT compiler), that maps byte-code commands to processor commands. Java can be considered both compiled and interpreted language, similarly to Kotlin and Scala (JVM languages). Oftentimes though, Java is thought to be compiled.
But I think it is possible to write an interpreter for any compiled language and a compiler for any interpreted language. Therefore, language classification in terms of compiled/interpreted languages is suggestive, not mandatory.
Linker is a program that links compiled files. It resolves imports.
I’m not sure that compiler cannot target multiple processors at once. ISA (Instruction set architecture) of processor is an important thing. Different processors with similar ISAs might be targeted by one compiler.↩︎
And this cannot be fixed by compiling a program twice: first to run it and find optimizations during runtime, and second to create optimized binary executables. This is because optimizations do depend on the input arguments, which may vary vastly.↩︎