Why bison is becoming extinct

Abstract

At some point in your career, you're going to implement a computer language. You probably won't be implementing Java or C++. You may not even recognize it as a language. Truth be told, there are an awful lot of domain-specific languages, or "little languages" [7] in common use: configuration files, HTML/XML documents, shell scripts, network protocols, mail headers, command-line arguments. The list goes on. A number of programs allow you to write scripts to control their operation; infact, just the other day I downloaded a neural network simulator which provided a little programming language to steer the simulation.How will you implement your language? There's the ad hoc approach, of course, but it's not well suited to languages whose design is complex or frequently changing. You also end up writing code to perform tasks which can be effectively automated.You might also consider using existing languages like Tcl [18] and Python [6]. These languages are designed to either be embedded in an existing application, or easily extended. This is a good solution when it can be used, saving a lot of time and effort. However, there may be concerns about tying your language to one which is itself changing, or the syntax and semantics of your language may not match those of such a "host" language.A third approach is to use compiler tools to implement your language. Most were designed for the implementation of large programming languages, but the same principles and techniques apply equally well to little languages. This article is the story of one such tool -- a parser generator tool -- and more importantly, what sort of tool is going to replace it, and why.

Generalized Lr Parsing Algorithm for Boolean Grammars

“…is a solution of this system if a substitution of L k for A k for all k turns each equation in (2) into an equality.…”

“…These four choices, as well as all possible continuations of the sequence, are shown in Figure 1. The majority of these sequences converge to the unique solution modulo {ε}, L = (∅, {ε}, {ε}, {ε}, {ε}), in four steps, and these sequences differ only in the order in which ε is added to A, C, D and E. However, the sequence can proceed differently: if ε is added to A and D before E and C, such as in the vector L (2) = (∅, {ε}, ∅, {ε}, ∅), then ε ∈ (AD ∩ EC)(L (2) ), and it becomes possible to choose S at this step. The next vector will be L (3) = ({ε}, {ε}, ∅, {ε}, ∅), but, as one can see from Figure 1, the computation will eventually repair itself and converge to the unique solution modulo {ε}.…”

“…Consider the extension from M = {ε} to M ∪ {a}: the sequence starts from L (0) = (∅, {ε}, {ε}, {ε}, {ε}), only A can be chosen at the first step, which gives L (1) = (∅, {ε, a}, {ε}, {ε}, {ε}). At the second step, only S can be chosen, and the resulting vector L (2) = ({a}, {ε, a}, {ε}, {ε}, {ε}) is the solution modulo {ε, a}.…”

“…Now, according to the transition table, one S-arc and one A-arc from each of the three nodes can possibly be added by reduction: these arcs are shown in Figure 5(c). The arc A [2] can be set by doing a reduction by the rule A → ε, and it does not depend upon the rest of the arcs. The arc A [1] is set by a reduction by A → aA, which can only be done if the arc A [2] is present, and hence there is the required path aA.…”

“…Initially, the algorithm was proposed with linguistic applications in view [18,19], and in the recent years there has been a growing interest in the use of this method for software engineering [2]. In particular, efficient implementation techniques are being researched [3,8], application-oriented parser generators are being implemented [8], and some extensions to the algorithm motivated by applications are considered [4].…”