4.7.6 Compaction of LR Parsing Tables
4.7.6 Compaction of LR Parsing Tables
A typical programming language grammar with 50 to 100 terminals and 100 productions may have an LALR parsing table with several hundred states. The action function may easily have 20,000 entries, each requiring at least 8 bits to encode. On small devices, a more efficient encoding than a two-dimensional array may be important. We shall mention briefly a few techniques that have been used to compress the ACTION and GOTO fields of an LR parsing table.
One useful technique for compacting the action field is to recognize that usually many rows of the action table are identical. For example, in Fig. 4.42, states 0 and 3 have identical action entries, and so do 2 and 6. We can therefore save considerable space, at little cost in time, if we create a pointer for each state into a one-dimensional array. Pointers for states with the same actions point to the same location. To access information from this array, we assign each terminal a number from zero to one less than the number of terminals, and we use this integer as an offset from the pointer value for each state. In a given state, the parsing action for the ith terminal will be found i locations past the pointer value for that state.
Further space efficiency can be achieved at the expense of a somewhat slower parser by creating a list for the actions of each state. The list consists of (terminal-symbol, action) pairs. The most frequent action for a state can be placed at the end of the list, and in place of a terminal we may use the notation “any,” meaning that if the current input symbol has not been found so far on the list, we should do that action no matter what the input is. Moreover, error entries can safely be replaced by reduce actions, for further uniformity along a row. The errors will be detected later, before a shift move.
Example 4.65: Consider the parsing table of Fig. 4.37. First, note that the actions for states 0, 4, 6, and 7 agree. We can represent them all by the list
SYMBOL |
ACTION |
id |
s5 |
( |
s4 |
any |
error |
State 1 has a similar list:
SYMBOL |
ACTION |
+ |
s6 |
$ |
acc |
any |
error |
In state 2, we can replace the error entries by r2, so reduction by production 2 will occur on any input but *. Thus the list for state 2 is
SYMBOL |
ACTION |
* |
s7 |
any |
r2 |
State 3 has only error and r4 entries. We can replace the former by the latter, so the list for state 3 consists of only the pair (any, r4). States 5, 10, and 11 can be treated similarly. The list for state 8 is
SYMBOL |
ACTION |
+ |
s6 |
) |
s11 |
any |
error |
and for state 9
SYMBOL |
ACTION |
* |
S7 |
any |
R1 |
□
We can also encode the GOTO table by a list, but here it app ears more efficient to make a list of pairs for each nonterminal A. Each pair on the list for A is of the form (currentState, nextState), indicating
GOTO [currentState, A] = nextState
This technique is useful because there tend to be rather few states in any one column of the GOTO table. The reason is that the GOTO on nonterminal A can only be a state derivable from a set of items in which some items have A immediately to the left of a dot. No set has items with X and Y immediately to the left of a dot if X ≠ Y. Thus, each state app ears in at most one GOTO column.
For more space reduction, we note that the error entries in the goto table are never consulted. We can therefore replace each error entry by the most common non-error entry in its column. This entry becomes the default; it is represented in the list for each column by one pair with any in place of currentState.
Example 4.66: Consider Fig. 4.37 again. The column for F has entry 10 for state 7, and all other entries are either 3 or error. We may replace error by 3 and create for column F the list
CURRENTSTATE |
NEXTSTATE |
7 |
10 |
any |
3 |
Similarly, a suitable list for column T is
CURRENTSTATE |
NEXTSTATE |
6 |
9 |
any |
2 |
For column E we may choose either 1 or 8 to be the default; two entries are necessary in either case. For example, we might create for column E the list
CURRENTSTATE |
NEXTSTATE |
4 |
8 |
any |
1 |
□
This space savings in these small examples may be misleading, because the total number of entries in the lists created in this example and the previous one together with the pointers from states to action lists and from nonterminals to next-state lists, result in unimpressive space savings over the matrix implementation of Fig. 4.37. For practical grammars, the space needed for the list representation is typically less than ten percent of that needed for the matrix representation. The table-compression methods for finite automata that were discussed in Section 3.9.8 can also be used to represent LR parsing tables.
4.7.6 Compaction of LR Parsing Tables的更多相关文章
- 4.7.3 Canonical LR(1) Parsing Tables
4.7.3 Canonical LR(1) Parsing Tables We now give the rules for constructing the LR(1) ACTION and GOT ...
- 4.7.5 Efficient Construction of LALR Parsing Tables
4.7.5 Efficient Construction of LALR Parsing Tables There are several modifications we can make to A ...
- 4.7.4 Constructing LALR Parsing Tables
4.7.4 Constructing LALR Parsing Tables We now introduce our last parser construction method, the LAL ...
- 4.4 Top-Down Parsing
4.4 Top-Down Parsing Top-down parsing can be viewed as the problem of constructing a parse tree for ...
- (转)Understanding C parsers generated by GNU Bison
原文链接:https://www.cs.uic.edu/~spopuri/cparser.html Satya Kiran PopuriGraduate StudentUniversity of Il ...
- 基于虎书实现LALR(1)分析并生成GLSL编译器前端代码(C#)
基于虎书实现LALR(1)分析并生成GLSL编译器前端代码(C#) 为了完美解析GLSL源码,获取其中的信息(都有哪些in/out/uniform等),我决定做个GLSL编译器的前端(以后简称编译器或 ...
- Donald Ervin Knuth:最年轻的图灵奖高德纳
高德纳(Donald Ervin Knuth,1938年),美国著名计算机科学家,斯坦福大学电脑系荣誉教授.高德纳教授被誉为现代计算机科学的鼻祖,在计算机科学及数学领域发表了多部 具广泛影响的论文和著 ...
- 4.8 Using Ambiguous Grammars
4.8 Using Ambiguous Grammars It is a fact that every ambiguous grammar fails to be LR and thus is no ...
- Lexer and parser generators (ocamllex, ocamlyacc)
Chapter 12 Lexer and parser generators (ocamllex, ocamlyacc) This chapter describes two program gene ...
随机推荐
- 零基础入门学习Python(18)--函数:灵活即强大
前言 上一节课我们基本介绍Python函数的用法,这一节课我们主要针对函数的参数进行进一步的深入学习. 知识点 形参(parameter)和实参(argument) >>> def ...
- 201621123079《Java程序设计》第1周学习总结
第1周-Java基本概念 1.本周学习总结 第一次上课接触java,了解了java的由来和历史,还有JCP,JSP的概念,并学会如何建立一个java文件和运行过程.感觉java比之前学习的数据结构更高 ...
- c++基础_特殊回文数
#include <iostream> using namespace std; int main(){ int n; cin>>n; ;i<;i++){ int tem ...
- LeetCode(56)Merge Intervals
题目 Given a collection of intervals, merge all overlapping intervals. For example, Given [1,3],[2,6], ...
- Vue如何在data中正常引入图片路径
在Vue项目中通过data设置图片路径,然后在template中引入后页面无法显示图片,浏览器控制台报错: 刚开始以为是路径出问题了,于是绝对路径.相对路 ...
- matplotlib多种绘图方式
目录 散点图 条形图 直方图 总结 散点图 假设通过爬虫你获取到了北京2016年3,10月份每天白天的最高气温(分别位于列表a,b),那么此时如何寻找出气温和随时间(天)变化的某种规律? a = [1 ...
- 关于meta标签的使用,属性的说明
原文转自:http://blog.csdn.net/gavid0124/article/details/46826127 网页源代码中有时候会遇到这样的一段代码: <metaname=" ...
- 九度oj 题目1180:对称矩阵
题目1180:对称矩阵 时间限制:1 秒 内存限制:32 兆 特殊判题:否 提交:3092 解决:1607 题目描述: 输入一个N维矩阵,判断是否对称. 输入: 输入第一行包括一个数:N(1<= ...
- Android SwipeSelector
Android SwipeSelector Android SwipeSelector是github上一个第三方开源的项目,其项目主页:https://github.com/roughike/S ...
- 【最大流】Escape
https://www.bnuoj.com/v3/contest_show.php?cid=9149#problem/F [题意] 给定n个人和m个星球,每个人可以匹配某些星球,每个星球有一定的容量限 ...