Parsers in modern compilers

Modern compilers and interpreters are based on different approaches to implement parsers. 

Brief overview of current state

• GCC 4.8.1 - hand-written top-down recursive descent parser
• Clang 3.3 - hand-written top-down recursive descent parser
• PathScale EkoPath 4 - hand-written top-down recursive descent parser (based on GCC frontend)
• Open64 5.0 - hand-written top-down recursive descent parser (based on GCC frontend)
• Cray Chapel 1.7.0 - bottom-up LALR(1) parser (generated by Bison), left-recursive grammar
• IBM X10 2.3.1 - bottom-up LALR(1) parser (generated by LPG)
• Ruby 2.0.0 - bottom-up LALR(1) parser (generated by Bison), left-recursive grammar
• PHP 5.5.0 - bottom-up LALR(1) parser (generated by Bison)
• Open JDK 7 - default: hand-written LALR parser; subproject: ANTLR based LL(*) parser
• Python 2.7.5 & 3.3.2 - hand-written top-down recursive descent parser
• Groovy 2.1.6 - LL(*) parser (generated by ANTLR)
• Go 1.1.1 - bottom-up LALR(1) parser (generated by Bison)
• Google V8 JavaScript Engine - hand-written top-down recursive descent parser
• Apple JavaScriptCore - hand-written top-down recursive descent parser

Lexers in modern compilers

In modern compilers two approaches are widely used to implement a lexer/scanner/tokenizer: the hand-written approach and the automatic lexer generator approach.

Compiler	Lexer/scanner/tokenizer implementation	Source
Cray Chapel 1.7.0	Flex	chaplel/compiler/parser/chapel.lex
IBM X10 2.3.1	LPG (LALR parser generator)	x10/x10.compiler/src/x10/parser/X10Lexer.java
GCC 4.8.1	handwritten	gcc/gcc/c-family/c-lex.c
Clang 3.3	handwritten	cfe/lib/Lex/Lexer.cpp
PathScale EkoPath 4	handwritten - base on GCC frontend
Open64 5.0	handwritten - based on GCC frontend