autoloading

Autoloading is a software technique that defers loading of code modules, libraries, or symbols until they are actually needed. Rather than loading all dependencies at startup, an autoloader maintains a mapping from identifiers (such as class names or function names) to the files that define them. When the program first references an unloaded symbol, the autoloader loads the corresponding file and defines the symbol.

Mechanisms vary by language, but common approaches include a registry of autoload entries, unresolved symbol hooks, or import hooks. The loader may also check caches to avoid repeated loads or use on-demand compilation. Some systems intercept unresolved references, while others expose lazy-loading through explicit calls.

Benefits include reduced startup time and memory usage, and sometimes simpler dependency management. Drawbacks include unpredictable latency on first use, potential exposure to I/O or error handling costs, debugging complexity, and the need for careful version and namespace management to avoid loading inconsistent components.

Examples appear in various ecosystems. In Lisp-family languages, autoload declarations defer loading until a symbol is invoked. PHP provides autoloaders via spl_autoload_register. Ruby offers autoload to defer defining constants. Python supports lazy imports via import hooks and PEP proposals. Java relies on the classloader mechanism for lazy class loading; Node.js and other ecosystems can implement lazy requires via dynamic loading.

Design considerations include error handling, caching, thread-safety, and security. Autoloading can simplify deployment but requires robust reporting and predictable behavior to avoid surprises during development and production.