1 // $Id: CQLLexer.java,v 1.10 2007-06-28 00:00:31 mike Exp $
3 package org.z3950.zing.cql;
4 import java.io.StreamTokenizer;
5 import java.io.StringReader;
6 import java.util.Hashtable;
9 // This is a semi-trivial subclass for java.io.StreamTokenizer that:
10 // * Has a halfDecentPushBack() method that actually works
11 // * Includes a render() method
12 // * Knows about the multi-character tokens "<=", ">=" and "<>"
13 // * Recognises a set of keywords as tokens in their own right
14 // * Includes some primitive debugging-output facilities
15 // It's used only by CQLParser.
17 class CQLLexer extends StreamTokenizer {
18 // New publicly visible token-types
19 static int TT_LE = 1000; // The "<=" relation
20 static int TT_GE = 1001; // The ">=" relation
21 static int TT_NE = 1002; // The "<>" relation
22 static int TT_AND = 1003; // The "and" boolean
23 static int TT_OR = 1004; // The "or" boolean
24 static int TT_NOT = 1005; // The "not" boolean
25 static int TT_PROX = 1006; // The "prox" boolean
26 static int TT_pWORD = 1010; // The "word" proximity unit
27 static int TT_SENTENCE = 1011; // The "sentence" proximity unit
28 static int TT_PARAGRAPH = 1012; // The "paragraph" proximity unit
29 static int TT_ELEMENT = 1013; // The "element" proximity unit
30 static int TT_ORDERED = 1014; // The "ordered" proximity ordering
31 static int TT_UNORDERED = 1015; // The "unordered" proximity ordering
33 // Support for keywords. It would be nice to compile this linear
34 // list into a Hashtable, but it's hard to store ints as hash
35 // values, and next to impossible to use them as hash keys. So
36 // we'll just scan the (very short) list every time we need to do
38 private class Keyword {
41 Keyword(int token, String keyword) {
43 this.keyword = keyword;
46 // This should logically be static, but Java won't allow it :-P
47 private Keyword[] keywords = {
48 new Keyword(TT_AND, "and"),
49 new Keyword(TT_OR, "or"),
50 new Keyword(TT_NOT, "not"),
51 new Keyword(TT_PROX, "prox"),
52 new Keyword(TT_pWORD, "word"),
53 new Keyword(TT_SENTENCE, "sentence"),
54 new Keyword(TT_PARAGRAPH, "paragraph"),
55 new Keyword(TT_ELEMENT, "element"),
56 new Keyword(TT_ORDERED, "ordered"),
57 new Keyword(TT_UNORDERED, "unordered"),
60 // For halfDecentPushBack() and the code at the top of nextToken()
61 private static int TT_UNDEFINED = -1000;
62 private int saved_ttype = TT_UNDEFINED;
63 private double saved_nval;
64 private String saved_sval;
66 // Controls debugging output
67 private static boolean DEBUG;
69 CQLLexer(String cql, boolean lexdebug) {
70 super(new StringReader(cql));
71 wordChars('!', '?'); // ASCII-dependency!
72 wordChars('[', '`'); // ASCII-dependency!
80 wordChars('\'', '\''); // prevent this from introducing strings
85 private static void debug(String str) {
87 System.err.println("LEXDEBUG: " + str);
90 // I don't honestly understand why we need this, but the
91 // documentation for java.io.StreamTokenizer.pushBack() is pretty
92 // vague about its semantics, and it seems to me that they could
93 // be summed up as "it doesn't work". This version has the very
94 // clear semantics "pretend I didn't call nextToken() just then".
96 private void halfDecentPushBack() {
102 public int nextToken() throws java.io.IOException {
103 if (saved_ttype != TT_UNDEFINED) {
107 saved_ttype = TT_UNDEFINED;
108 debug("using saved ttype=" + ttype + ", " +
109 "nval=" + nval + ", sval='" + sval + "'");
113 underlyingNextToken();
115 debug("token starts with '<' ...");
116 underlyingNextToken();
118 debug("token continues with '=' - it's '<='");
120 } else if (ttype == '>') {
121 debug("token continues with '>' - it's '<>'");
124 debug("next token is " + render() + " (pushed back)");
125 halfDecentPushBack();
127 debug("AFTER: ttype is now " + ttype + " - " + render());
129 } else if (ttype == '>') {
130 debug("token starts with '>' ...");
131 underlyingNextToken();
133 debug("token continues with '=' - it's '>='");
136 debug("next token is " + render() + " (pushed back)");
137 halfDecentPushBack();
139 debug("AFTER: ttype is now " + ttype + " - " + render());
143 debug("done nextToken(): ttype=" + ttype + ", " +
144 "nval=" + nval + ", " + "sval='" + sval + "'" +
145 " (" + render() + ")");
150 // It's important to do keyword recognition here at the lowest
151 // level, otherwise when one of these words follows "<" or ">"
152 // (which can be the beginning of multi-character tokens) it gets
153 // pushed back as a string, and its keywordiness is not
156 public int underlyingNextToken() throws java.io.IOException {
158 if (ttype == TT_WORD)
159 for (int i = 0; i < keywords.length; i++)
160 if (sval.equalsIgnoreCase(keywords[i].keyword))
161 ttype = keywords[i].token;
166 // Simpler interface for the usual case: current token with quoting
168 return render(ttype, true);
171 String render(int token, boolean quoteChars) {
172 if (token == TT_EOF) {
174 } else if (token == TT_NUMBER) {
175 if ((double) nval == (int) nval) {
176 return new Integer((int) nval).toString();
178 return new Double((double) nval).toString();
180 } else if (token == TT_WORD) {
181 return "word: " + sval;
182 } else if (token == '"') {
183 return "string: \"" + sval + "\"";
184 } else if (token == TT_LE) {
186 } else if (token == TT_GE) {
188 } else if (token == TT_NE) {
192 // Check whether its associated with one of the keywords
193 for (int i = 0; i < keywords.length; i++)
194 if (token == keywords[i].token)
195 return keywords[i].keyword;
197 // Otherwise it must be a single character, such as '(' or '/'.
198 String res = String.valueOf((char) token);
199 if (quoteChars) res = "'" + res + "'";
203 public static void main(String[] args) throws Exception {
204 if (args.length > 1) {
205 System.err.println("Usage: CQLLexer [<CQL-query>]");
206 System.err.println("If unspecified, query is read from stdin");
211 if (args.length == 1) {
214 byte[] bytes = new byte[10000];
216 // Read in the whole of standard input in one go
217 int nbytes = System.in.read(bytes);
218 } catch (java.io.IOException ex) {
219 System.err.println("Can't read query: " + ex.getMessage());
222 cql = new String(bytes);
225 CQLLexer lexer = new CQLLexer(cql, true);
227 while ((token = lexer.nextToken()) != TT_EOF) {
228 // Nothing to do: debug() statements render tokens for us