2 package org.z3950.zing.cql;
3 import java.io.InputStream;
5 import java.io.StreamTokenizer;
6 import java.io.StringReader;
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 public final static int TT_LE = 1000; // The "<=" relation
20 public final static int TT_GE = 1001; // The ">=" relation
21 public final static int TT_NE = 1002; // The "<>" relation
22 public final static int TT_EQEQ = 1003; // The "==" relation
23 public final static int TT_AND = 1004; // The "and" boolean
24 public final static int TT_OR = 1005; // The "or" boolean
25 public final static int TT_NOT = 1006; // The "not" boolean
26 public final static int TT_PROX = 1007; // The "prox" boolean
27 public final 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 this(new StringReader(cql), lexdebug);
64 CQLLexer(Reader cql, boolean lexdebug) {
66 wordChars('!', '?'); // ASCII-dependency!
67 wordChars('[', '`'); // ASCII-dependency!
75 wordChars('\'', '\''); // prevent this from introducing strings
79 ordinaryChars('0', '9');
84 private static void debug(String str) {
86 System.err.println("LEXDEBUG: " + str);
89 // I don't honestly understand why we need this, but the
90 // documentation for java.io.StreamTokenizer.pushBack() is pretty
91 // vague about its semantics, and it seems to me that they could
92 // be summed up as "it doesn't work". This version has the very
93 // clear semantics "pretend I didn't call nextToken() just then".
95 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());
141 } else if (ttype == '=') {
142 debug("token starts with '=' ...");
143 underlyingNextToken();
145 debug("token continues with '=' - it's '=='");
148 debug("next token is " + render() + " (pushed back)");
149 halfDecentPushBack();
151 debug("AFTER: ttype is now " + ttype + " - " + render());
155 debug("done nextToken(): ttype=" + ttype + ", " +
156 "nval=" + nval + ", " + "sval='" + sval + "'" +
157 " (" + render() + ")");
162 // It's important to do keyword recognition here at the lowest
163 // level, otherwise when one of these words follows "<" or ">"
164 // (which can be the beginning of multi-character tokens) it gets
165 // pushed back as a string, and its keywordiness is not
168 public int underlyingNextToken() throws java.io.IOException {
170 if (ttype == TT_WORD)
171 for (int i = 0; i < keywords.length; i++)
172 if (sval.equalsIgnoreCase(keywords[i].keyword))
173 ttype = keywords[i].token;
178 // Simpler interface for the usual case: current token with quoting
180 return render(ttype, true);
183 String render(int token, boolean quoteChars) {
184 if (token == TT_EOF) {
186 } else if (token == TT_NUMBER) {
187 if ((double) nval == (int) nval) {
188 return new Integer((int) nval).toString();
190 return new Double((double) nval).toString();
192 } else if (token == TT_WORD) {
193 return "word: " + sval;
194 } else if (token == '"') {
195 return "string: \"" + sval + "\"";
196 } else if (token == TT_LE) {
198 } else if (token == TT_GE) {
200 } else if (token == TT_NE) {
202 } else if (token == TT_EQEQ) {
206 // Check whether its associated with one of the keywords
207 for (int i = 0; i < keywords.length; i++)
208 if (token == keywords[i].token)
209 return keywords[i].keyword;
211 // Otherwise it must be a single character, such as '(' or '/'.
212 String res = String.valueOf((char) token);
213 if (quoteChars) res = "'" + res + "'";
217 public static void main(String[] args) throws Exception {
218 if (args.length > 1) {
219 System.err.println("Usage: CQLLexer [<CQL-query>]");
220 System.err.println("If unspecified, query is read from stdin");
225 if (args.length == 1) {
228 byte[] bytes = new byte[10000];
230 // Read in the whole of standard input in one go
231 int nbytes = System.in.read(bytes);
232 } catch (java.io.IOException ex) {
233 System.err.println("Can't read query: " + ex.getMessage());
236 cql = new String(bytes);
239 CQLLexer lexer = new CQLLexer(cql, true);
241 while ((token = lexer.nextToken()) != TT_EOF) {
242 // Nothing to do: debug() statements render tokens for us