1 // $Id: CQLLexer.java,v 1.14 2007-07-03 13:30:42 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_EQEQ = 1003; // The "==" relation
23 static int TT_AND = 1004; // The "and" boolean
24 static int TT_OR = 1005; // The "or" boolean
25 static int TT_NOT = 1006; // The "not" boolean
26 static int TT_PROX = 1007; // The "prox" boolean
27 static int TT_SORTBY = 1008; // The "sortby" operator
29 // Support for keywords. It would be nice to compile this linear
30 // list into a Hashtable, but it's hard to store ints as hash
31 // values, and next to impossible to use them as hash keys. So
32 // we'll just scan the (very short) list every time we need to do
34 private class Keyword {
37 Keyword(int token, String keyword) {
39 this.keyword = keyword;
42 // This should logically be static, but Java won't allow it :-P
43 private Keyword[] keywords = {
44 new Keyword(TT_AND, "and"),
45 new Keyword(TT_OR, "or"),
46 new Keyword(TT_NOT, "not"),
47 new Keyword(TT_PROX, "prox"),
48 new Keyword(TT_SORTBY, "sortby"),
51 // For halfDecentPushBack() and the code at the top of nextToken()
52 private static int TT_UNDEFINED = -1000;
53 private int saved_ttype = TT_UNDEFINED;
54 private double saved_nval;
55 private String saved_sval;
57 // Controls debugging output
58 private static boolean DEBUG;
60 CQLLexer(String cql, boolean lexdebug) {
61 super(new StringReader(cql));
62 wordChars('!', '?'); // ASCII-dependency!
63 wordChars('[', '`'); // ASCII-dependency!
71 wordChars('\'', '\''); // prevent this from introducing strings
76 private static void debug(String str) {
78 System.err.println("LEXDEBUG: " + str);
81 // I don't honestly understand why we need this, but the
82 // documentation for java.io.StreamTokenizer.pushBack() is pretty
83 // vague about its semantics, and it seems to me that they could
84 // be summed up as "it doesn't work". This version has the very
85 // clear semantics "pretend I didn't call nextToken() just then".
87 private void halfDecentPushBack() {
93 public int nextToken() throws java.io.IOException {
94 if (saved_ttype != TT_UNDEFINED) {
98 saved_ttype = TT_UNDEFINED;
99 debug("using saved ttype=" + ttype + ", " +
100 "nval=" + nval + ", sval='" + sval + "'");
104 underlyingNextToken();
106 debug("token starts with '<' ...");
107 underlyingNextToken();
109 debug("token continues with '=' - it's '<='");
111 } else if (ttype == '>') {
112 debug("token continues with '>' - it's '<>'");
115 debug("next token is " + render() + " (pushed back)");
116 halfDecentPushBack();
118 debug("AFTER: ttype is now " + ttype + " - " + render());
120 } else if (ttype == '>') {
121 debug("token starts with '>' ...");
122 underlyingNextToken();
124 debug("token continues with '=' - it's '>='");
127 debug("next token is " + render() + " (pushed back)");
128 halfDecentPushBack();
130 debug("AFTER: ttype is now " + ttype + " - " + render());
132 } else if (ttype == '=') {
133 debug("token starts with '=' ...");
134 underlyingNextToken();
136 debug("token continues with '=' - it's '=='");
139 debug("next token is " + render() + " (pushed back)");
140 halfDecentPushBack();
142 debug("AFTER: ttype is now " + ttype + " - " + render());
146 debug("done nextToken(): ttype=" + ttype + ", " +
147 "nval=" + nval + ", " + "sval='" + sval + "'" +
148 " (" + render() + ")");
153 // It's important to do keyword recognition here at the lowest
154 // level, otherwise when one of these words follows "<" or ">"
155 // (which can be the beginning of multi-character tokens) it gets
156 // pushed back as a string, and its keywordiness is not
159 public int underlyingNextToken() throws java.io.IOException {
161 if (ttype == TT_WORD)
162 for (int i = 0; i < keywords.length; i++)
163 if (sval.equalsIgnoreCase(keywords[i].keyword))
164 ttype = keywords[i].token;
169 // Simpler interface for the usual case: current token with quoting
171 return render(ttype, true);
174 String render(int token, boolean quoteChars) {
175 if (token == TT_EOF) {
177 } else if (token == TT_NUMBER) {
178 if ((double) nval == (int) nval) {
179 return new Integer((int) nval).toString();
181 return new Double((double) nval).toString();
183 } else if (token == TT_WORD) {
184 return "word: " + sval;
185 } else if (token == '"') {
186 return "string: \"" + sval + "\"";
187 } else if (token == TT_LE) {
189 } else if (token == TT_GE) {
191 } else if (token == TT_NE) {
193 } else if (token == TT_EQEQ) {
197 // Check whether its associated with one of the keywords
198 for (int i = 0; i < keywords.length; i++)
199 if (token == keywords[i].token)
200 return keywords[i].keyword;
202 // Otherwise it must be a single character, such as '(' or '/'.
203 String res = String.valueOf((char) token);
204 if (quoteChars) res = "'" + res + "'";
208 public static void main(String[] args) throws Exception {
209 if (args.length > 1) {
210 System.err.println("Usage: CQLLexer [<CQL-query>]");
211 System.err.println("If unspecified, query is read from stdin");
216 if (args.length == 1) {
219 byte[] bytes = new byte[10000];
221 // Read in the whole of standard input in one go
222 int nbytes = System.in.read(bytes);
223 } catch (java.io.IOException ex) {
224 System.err.println("Can't read query: " + ex.getMessage());
227 cql = new String(bytes);
230 CQLLexer lexer = new CQLLexer(cql, true);
232 while ((token = lexer.nextToken()) != TT_EOF) {
233 // Nothing to do: debug() statements render tokens for us