Neutralization of Syntax Errors in the Compiler of the Functional-Imperative Language El

Abstract

This article discusses the problem of neutralizing syntax errors in relation to the development of the compiler of the new functional-imperative programming language El. Solving this problem implies the need to choose such a variant for correcting the program being processed, in which it is possible to continue its syntactic analysis in order to detect the maximum possible number of non-induced errors. A neutralization method is proposed that effectively uses the features of the deterministic syntactic parsing algorithm of the compiler, implemented as a stack automaton controlled by an input token and a symbol from the top of the stack. The method is based on a complete enumeration of all valid tokens for modifying one error token – alternately inserting, replacing, and deleting to select that parser restart parameters at which the next syntax error is detected at the maximum distance from the neutralized one. If no modification of a single erroneous token leads to successful neutralization, then one character is alternately discarded from the top of the stored stack or the position of the current token in the input stream advances. After that, all attempts to modify the current token are repeated. The absence of a panic mode can reduce the number of unparsed tokens. This neutralization algorithm combines with “rules for typical errors”, which are included in the grammar for such possible situations when, in order to successfully neutralize the error, you need to insert (replace) not one, but two or three tokens. Such errors are possible in programs in the El language, their effective neutralization by other means seems to be very difficult.