It used typestate analysis to check variables transitions errors, to rule out some semantically non meaningful transitions from one state to another (i.e. starting from a value, some sequences of operations on a variable are nonsensical), of which reading an uninitialized variable is a special case. In this role of compile-time checking of data initialization is similar to definite assignment analysis performed by Java, Cyclone and C#.
Hermes and its predecessor, NIL (Network Implementation Language), were the earliest programming languages supporting this form of initialization checking.[6]
Typestate was actually used more extensively, to generate compiler-inserted "delete" operations.
References
^Strom, Robert E.; Bacon, David F; Goldberg, Arthur P.; Lowry, Andy; Yellin, Daniel M.; Yemini, Shaula (1991). Hermes - A Language for Distributed Computing. Englewood Cliffs, NJ, USA: Prentice-Hall. ISBN978-0-13-389537-7.
^Strom, R.E. (1990). "Hermes: an integrated language and system for distributed programming". IEEE Workshop on Experimental Distributed Systems. pp. 75–82. doi:10.1109/EDS.1990.138054. S2CID10223422.
^Bacon, David F.; Lowry, Andy (1990). A Portable Run-time System for the Hermes Distributed Programming Language. USENIX Summer Technical Conference. Anaheim, California, USA: Usenix Association. pp. 39–50.
^Strom, Robert E. (1983). "Mechanisms for compile-time enforcement of security". Proceedings of the 10th ACM SIGACT-SIGPLAN symposium on Principles of programming languages - POPL '83. pp. 276–284. doi:10.1145/567067.567093. ISBN978-0897910903. S2CID6630704.
