Dependency handling in an on-demand network code execution system

Abstract

Systems and methods are described for managing dependencies among code in an on-demand code execution system in a manner that enables efficient execution of the code. A user may generate a task on the system by submitting code, which may reference additional code on the system. The system can determine a dependency graph for the submitted code, and generate multiple representations of the dependency graph, storing dependency code objects in different numbers of files. On receiving a request to provision an execution environment for the task, the system can select an appropriate representation based on an expected reuse of the files within different representations. For example, a representation of fewer files can be selected where expected future use is low, minimizing retrieval time but also potentially limiting reuse of code objects within the representation.