1 // $Id: CQLLexer.java,v 1.5 2002-11-14 22:04:16 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_ANY = 1007; // The "any" relation
27 static int TT_ALL = 1008; // The "all" relation
28 static int TT_EXACT = 1009; // The "exact" relation
29 static int TT_pWORD = 1010; // The "word" proximity unit
30 static int TT_SENTENCE = 1011; // The "sentence" proximity unit
31 static int TT_PARAGRAPH = 1012; // The "paragraph" proximity unit
32 static int TT_ELEMENT = 1013; // The "element" proximity unit
33 static int TT_ORDERED = 1014; // The "ordered" proximity ordering
34 static int TT_UNORDERED = 1015; // The "unordered" proximity ordering
35 static int TT_RELEVANT = 1016; // The "relevant" relation modifier
36 static int TT_FUZZY = 1017; // The "fuzzy" relation modifier
37 static int TT_STEM = 1018; // The "stem" relation modifier
38 static int TT_SCR = 1019; // The server choice relation
40 // Support for keywords. It would be nice to compile this linear
41 // list into a Hashtable, but it's hard to store ints as hash
42 // values, and next to impossible to use them as hash keys. So
43 // we'll just scan the (very short) list every time we need to do
45 private class Keyword {
48 Keyword(int token, String keyword) {
50 this.keyword = keyword;
53 // This should logically be static, but Java won't allow it :-P
54 private Keyword[] keywords = {
55 new Keyword(TT_AND, "and"),
56 new Keyword(TT_OR, "or"),
57 new Keyword(TT_NOT, "not"),
58 new Keyword(TT_PROX, "prox"),
59 new Keyword(TT_ANY, "any"),
60 new Keyword(TT_ALL, "all"),
61 new Keyword(TT_EXACT, "exact"),
62 new Keyword(TT_pWORD, "word"),
63 new Keyword(TT_SENTENCE, "sentence"),
64 new Keyword(TT_PARAGRAPH, "paragraph"),
65 new Keyword(TT_ELEMENT, "element"),
66 new Keyword(TT_ORDERED, "ordered"),
67 new Keyword(TT_UNORDERED, "unordered"),
68 new Keyword(TT_RELEVANT, "relevant"),
69 new Keyword(TT_FUZZY, "fuzzy"),
70 new Keyword(TT_STEM, "stem"),
71 new Keyword(TT_SCR, "scr"),
74 // For halfDecentPushBack() and the code at the top of nextToken()
75 private static int TT_UNDEFINED = -1000;
76 private int saved_ttype = TT_UNDEFINED;
77 private double saved_nval;
78 private String saved_sval;
80 // Controls debugging output
81 private static boolean DEBUG;
83 CQLLexer(String cql, boolean lexdebug) {
84 super(new StringReader(cql));
85 wordChars('!', '?'); // ASCII-dependency!
86 wordChars('[', '`'); // ASCII-dependency!
94 wordChars('\'', '\''); // prevent this from introducing strings
99 private static void debug(String str) {
101 System.err.println("LEXDEBUG: " + str);
104 // I don't honestly understand why we need this, but the
105 // documentation for java.io.StreamTokenizer.pushBack() is pretty
106 // vague about its semantics, and it seems to me that they could
107 // be summed up as "it doesn't work". This version has the very
108 // clear semantics "pretend I didn't call nextToken() just then".
110 private void halfDecentPushBack() {
116 public int nextToken() throws java.io.IOException {
117 if (saved_ttype != TT_UNDEFINED) {
121 saved_ttype = TT_UNDEFINED;
122 debug("using saved ttype=" + ttype + ", " +
123 "nval=" + nval + ", sval='" + sval + "'");
127 underlyingNextToken();
129 debug("token starts with '<' ...");
130 underlyingNextToken();
132 debug("token continues with '=' - it's '<='");
134 } else if (ttype == '>') {
135 debug("token continues with '>' - it's '<>'");
138 debug("next token is " + render() + " (pushed back)");
139 halfDecentPushBack();
141 debug("AFTER: ttype is now " + ttype + " - " + render());
143 } else if (ttype == '>') {
144 debug("token starts with '>' ...");
145 underlyingNextToken();
147 debug("token continues with '=' - it's '>='");
150 debug("next token is " + render() + " (pushed back)");
151 halfDecentPushBack();
153 debug("AFTER: ttype is now " + ttype + " - " + render());
157 debug("done nextToken(): ttype=" + ttype + ", " +
158 "nval=" + nval + ", " + "sval='" + sval + "'" +
159 " (" + render() + ")");
164 // It's important to do keyword recognition here at the lowest
165 // level, otherwise when one of these words follows "<" or ">"
166 // (which can be the beginning of multi-character tokens) it gets
167 // pushed back as a string, and its keywordiness is not
170 public int underlyingNextToken() throws java.io.IOException {
172 if (ttype == TT_WORD)
173 for (int i = 0; i < keywords.length; i++)
174 if (sval.equalsIgnoreCase(keywords[i].keyword))
175 ttype = keywords[i].token;
180 // Simpler interface for the usual case: current token with quoting
182 return render(ttype, true);
185 String render(int token, boolean quoteChars) {
186 if (token == TT_EOF) {
188 } else if (token == TT_NUMBER) {
189 return new Integer((int) nval).toString();
190 } else if (token == TT_WORD) {
191 return "word: " + sval;
192 } else if (token == '"') {
193 return "string: \"" + sval + "\"";
194 } else if (token == TT_LE) {
196 } else if (token == TT_GE) {
198 } else if (token == TT_NE) {
202 // Check whether its associated with one of the keywords
203 for (int i = 0; i < keywords.length; i++)
204 if (token == keywords[i].token)
205 return keywords[i].keyword;
207 // Otherwise it must be a single character, such as '(' or '/'.
208 String res = String.valueOf((char) token);
209 if (quoteChars) res = "'" + res + "'";
213 public static void main(String[] args) throws Exception {
214 if (args.length > 1) {
215 System.err.println("Usage: CQLLexer [<CQL-query>]");
216 System.err.println("If unspecified, query is read from stdin");
221 if (args.length == 1) {
224 byte[] bytes = new byte[10000];
226 // Read in the whole of standard input in one go
227 int nbytes = System.in.read(bytes);
228 } catch (java.io.IOException ex) {
229 System.err.println("Can't read query: " + ex.getMessage());
232 cql = new String(bytes);
235 CQLLexer lexer = new CQLLexer(cql, true);
237 while ((token = lexer.nextToken()) != TT_EOF) {
238 // Nothing to do: debug() statements render tokens for us