Conditional Execution of Transpiler Passes Based on Per-Script Feature Detection

A new research paper details a significant architectural upgrade to Google's production JavaScript compiler, the Closure Compiler. Authored by Rishipal Singh Bhatia, the work tackles a core inefficiency in how compilers handle evolving ECMAScript standards. Traditionally, compilers run a fixed, monolithic pipeline of transpilation passes based on a target language level, forcing expensive Abstract Syntax Tree (AST) traversals to lower features that may not even exist in the input source code. This results in substantial computational waste.

The proposed improvement implements conditional execution of transpiler passes. It works by accurately tracking and dynamically maintaining the exact set of language features present in each JavaScript script throughout the entire compilation process. This 'FeatureSet' acts as a filter, allowing the compiler to dynamically skip running unnecessary lowering passes. The paper details critical architectural safeguards, including strategic pass ordering and dynamic validation, to ensure the transpiled code remains feature-correct.

Evaluation on large-scale production monorepos demonstrated the real-world impact: a considerable reduction in compilation time alongside savings in compute and memory usage. This optimization is particularly valuable for industrial-scale development, where build times directly impact developer productivity and infrastructure costs. The implementation is already available in the Google Closure Compiler, marking a move from theoretical research to practical engineering gain.