A Type-and-Effect System for Asynchronous, Typed Events
By: Yuheng Long and Hridesh Rajan
Download PaperAbstract
Implicit concurrency between handlers is important for event driven systems because it helps simultaneously promote modularity and scalability. Knowing the side-effect of the handlers is critical in these systems to avoid concurrency hazards such as data races. As event systems are dynamic, because statically known and unknown handlers can register almost at anytime during program execution, static effect analyses have to reconcile over competing goals such as precision, soundness and modularity. We recently developed asynchronous, typed events, a system that can analyze the effects of the handlers at runtime. This mechanism utilizes runtime information to enable precise effect computation and greatly improves concurrency between handlers. In this paper, we present the formal underpinnings of asynchronous, typed events, and examine the concurrency safety properties it provides. The technical innovations of our system include a novel effect system to soundly approximate the dynamism introduced by runtime handlers registration, a static analysis to precompute the effects and a dynamic analysis that uses the precomputed effects to improve concurrency. Our design simplifies modular concurrency reasoning and avoids concurrency hazards.
ACM Reference
Long, Y. and Rajan, H. 2016. A Type-and-Effect System for Asynchronous, Typed Events. Modularity’16: 15th International Conference on Modularity (Mar. 2016).
BibTeX Reference
@inproceedings{long2016type,
author = {Yuheng Long and Hridesh Rajan},
title = {A Type-and-Effect System for Asynchronous, Typed Events},
booktitle = {Modularity'16: 15th International Conference on Modularity},
series = {Modularity'16},
location = {Malaga, Spain},
month = {March},
year = {2016},
entrysubtype = {conference},
abstract = {
Implicit concurrency between handlers is important for event driven systems
because it helps simultaneously promote modularity and scalability. Knowing
the side-effect of the handlers is critical in these systems to avoid
concurrency hazards such as data races. As event systems are dynamic, because
statically known and unknown handlers can register almost at anytime during
program execution, static effect analyses have to reconcile over competing
goals such as precision, soundness and modularity. We recently developed
asynchronous, typed events, a system that can analyze the effects of the
handlers at runtime. This mechanism utilizes runtime information to enable
precise effect computation and greatly improves concurrency between handlers.
In this paper, we present the formal underpinnings of asynchronous, typed
events, and examine the concurrency safety properties it provides. The
technical innovations of our system include a novel effect system to soundly
approximate the dynamism introduced by runtime handlers registration, a static
analysis to precompute the effects and a dynamic analysis that uses the
precomputed effects to improve concurrency. Our design simplifies modular
concurrency reasoning and avoids concurrency hazards.
}
}