2 * Copyright (c) 2015 Cisco Systems, Inc. and others. All rights reserved.
4 * This program and the accompanying materials are made available under the
5 * terms of the Eclipse Public License v1.0 which accompanies this distribution,
6 * and is available at http://www.eclipse.org/legal/epl-v10.html
8 package org.opendaylight.yangtools.yang.parser.rfc7950.repo;
10 import static com.google.common.base.Verify.verify;
12 import com.google.common.annotations.VisibleForTesting;
13 import com.google.common.base.CharMatcher;
14 import com.google.common.base.VerifyException;
15 import org.antlr.v4.runtime.tree.ParseTree;
16 import org.antlr.v4.runtime.tree.TerminalNode;
17 import org.eclipse.jdt.annotation.NonNull;
18 import org.opendaylight.yangtools.yang.common.YangVersion;
19 import org.opendaylight.yangtools.yang.parser.antlr.YangStatementParser;
20 import org.opendaylight.yangtools.yang.parser.antlr.YangStatementParser.ArgumentContext;
21 import org.opendaylight.yangtools.yang.parser.spi.source.SourceException;
22 import org.opendaylight.yangtools.yang.parser.spi.source.StatementSourceReference;
25 * Utilities for dealing with YANG statement argument strings, encapsulated in ANTLR grammar's ArgumentContext.
27 abstract class ArgumentContextUtils {
29 * YANG 1.0 version of strings, which were not completely clarified in
30 * <a href="https://tools.ietf.org/html/rfc6020#section-6.1.3">RFC6020</a>.
32 private static final class RFC6020 extends ArgumentContextUtils {
33 private static final @NonNull RFC6020 INSTANCE = new RFC6020();
36 void checkDoubleQuoted(final String str, final StatementSourceReference ref, final int backslash) {
41 void checkUnquoted(final String str, final StatementSourceReference ref) {
47 * YANG 1.1 version of strings, which were clarified in
48 * <a href="https://tools.ietf.org/html/rfc7950#section-6.1.3">RFC7950</a>.
50 // NOTE: the differences clarified lead to a proper ability to delegate this to ANTLR lexer, but that does not
51 // understand versions and needs to work with both.
52 private static final class RFC7950 extends ArgumentContextUtils {
53 private static final CharMatcher ANYQUOTE_MATCHER = CharMatcher.anyOf("'\"");
54 private static final @NonNull RFC7950 INSTANCE = new RFC7950();
57 void checkDoubleQuoted(final String str, final StatementSourceReference ref, final int backslash) {
58 if (backslash < str.length() - 1) {
59 int index = backslash;
61 switch (str.charAt(index + 1)) {
66 index = str.indexOf('\\', index + 2);
69 throw new SourceException(ref, "YANG 1.1: illegal double quoted string (%s). In double "
70 + "quoted string the backslash must be followed by one of the following character "
71 + "[n,t,\",\\], but was '%s'.", str, str.charAt(index + 1));
78 void checkUnquoted(final String str, final StatementSourceReference ref) {
79 SourceException.throwIf(ANYQUOTE_MATCHER.matchesAnyOf(str), ref,
80 "YANG 1.1: unquoted string (%s) contains illegal characters", str);
84 private static final CharMatcher WHITESPACE_MATCHER = CharMatcher.whitespace();
86 private ArgumentContextUtils() {
90 static @NonNull ArgumentContextUtils forVersion(final YangVersion version) {
93 return RFC6020.INSTANCE;
95 return RFC7950.INSTANCE;
97 throw new IllegalStateException("Unhandled version " + version);
101 // TODO: teach the only caller about versions, or provide common-enough idioms for its use case
102 static @NonNull ArgumentContextUtils rfc6020() {
103 return RFC6020.INSTANCE;
107 * NOTE: this method we do not use convenience methods provided by generated parser code, but instead are making
108 * based on the grammar assumptions. While this is more verbose, it cuts out a number of unnecessary code,
109 * such as intermediate List allocation et al.
111 final @NonNull String stringFromStringContext(final ArgumentContext context, final StatementSourceReference ref) {
112 // Get first child, which we fully expect to exist and be a lexer token
113 final ParseTree firstChild = context.getChild(0);
114 verify(firstChild instanceof TerminalNode, "Unexpected shape of %s", context);
115 final TerminalNode firstNode = (TerminalNode) firstChild;
116 final int firstType = firstNode.getSymbol().getType();
118 case YangStatementParser.IDENTIFIER:
119 // Simple case, there is a simple string, which cannot contain anything that we would need to process.
120 return firstNode.getText();
121 case YangStatementParser.STRING:
122 // Complex case, defer to a separate method
123 return concatStrings(context, ref);
125 throw new VerifyException("Unexpected first symbol in " + context);
129 private String concatStrings(final ArgumentContext context, final StatementSourceReference ref) {
131 * We have multiple fragments. Just search the tree. This code is equivalent to
133 * context.STRING().forEach(stringNode -> appendString(sb, stringNode, ref))
135 * except we minimize allocations which that would do.
137 final StringBuilder sb = new StringBuilder();
138 for (ParseTree child : context.children) {
139 verify(child instanceof TerminalNode, "Unexpected fragment component %s", child);
140 final TerminalNode childNode = (TerminalNode) child;
141 switch (childNode.getSymbol().getType()) {
142 case YangStatementParser.SEP:
145 case YangStatementParser.PLUS:
146 // Operator, which we are handling by concat
148 case YangStatementParser.STRING:
149 // a lexer string, could be pretty much anything
150 // TODO: appendString() is a dispatch based on quotes, which we should be able to defer to lexer for
151 // a dedicated type. That would expand the switch table here, but since we have it anyway, it
152 // would be nice to have the quoting distinction already taken care of. The performance
153 // difference will need to be benchmarked, though.
154 appendString(sb, childNode, ref);
157 throw new VerifyException("Unexpected symbol in " + childNode);
160 return sb.toString();
163 private void appendString(final StringBuilder sb, final TerminalNode stringNode,
164 final StatementSourceReference ref) {
165 final String str = stringNode.getText();
166 final char firstChar = str.charAt(0);
167 final char lastChar = str.charAt(str.length() - 1);
168 if (firstChar == '"' && lastChar == '"') {
169 sb.append(normalizeDoubleQuoted(str.substring(1, str.length() - 1),
170 stringNode.getSymbol().getCharPositionInLine(), ref));
171 } else if (firstChar == '\'' && lastChar == '\'') {
173 * According to RFC6020 a single quote character cannot occur in a single-quoted string, even when preceded
176 sb.append(str, 1, str.length() - 1);
178 checkUnquoted(str, ref);
183 private String normalizeDoubleQuoted(final String str, final int dquot, final StatementSourceReference ref) {
184 // Whitespace normalization happens irrespective of further handling and has no effect on the result
185 final String stripped = trimWhitespace(str, dquot);
187 // Now we need to perform some amount of unescaping. This serves as a pre-check before we dispatch
188 // validation and processing (which will reuse the work we have done)
189 final int backslash = stripped.indexOf('\\');
190 return backslash == -1 ? stripped : unescape(ref, stripped, backslash);
194 * NOTE: Enforcement and transformation logic done by these methods should logically reside in the lexer and ANTLR
195 * account the for it with lexer modes. We do not want to force a re-lexing phase in the parser just because
196 * we decided to let ANTLR do the work.
198 abstract void checkDoubleQuoted(String str, StatementSourceReference ref, int backslash);
200 abstract void checkUnquoted(String str, StatementSourceReference ref);
203 * Unescape escaped double quotes, tabs, new line and backslash in the inner string and trim the result.
205 private String unescape(final StatementSourceReference ref, final String str, final int backslash) {
206 checkDoubleQuoted(str, ref, backslash);
207 StringBuilder sb = new StringBuilder(str.length());
208 unescapeBackslash(sb, str, backslash);
209 return sb.toString();
213 static void unescapeBackslash(final StringBuilder sb, final String str, final int backslash) {
214 String substring = str;
215 int backslashIndex = backslash;
217 int nextIndex = backslashIndex + 1;
218 if (backslashIndex != -1 && nextIndex < substring.length()) {
219 replaceBackslash(sb, substring, nextIndex);
220 substring = substring.substring(nextIndex + 1);
221 if (substring.length() > 0) {
222 backslashIndex = substring.indexOf('\\');
227 sb.append(substring);
233 private static void replaceBackslash(final StringBuilder sb, final String str, final int nextAfterBackslash) {
234 int backslash = nextAfterBackslash - 1;
235 sb.append(str, 0, backslash);
236 final char c = str.charAt(nextAfterBackslash);
249 sb.append(str, backslash, nextAfterBackslash + 1);
254 static String trimWhitespace(final String str, final int dquot) {
255 final int firstBrk = str.indexOf('\n');
256 if (firstBrk == -1) {
260 // Okay, we may need to do some trimming, set up a builder and append the first segment
261 final int length = str.length();
262 final StringBuilder sb = new StringBuilder(length);
264 // Append first segment, which needs only tail-trimming
265 sb.append(str, 0, trimTrailing(str, 0, firstBrk)).append('\n');
267 // With that out of the way, setup our iteration state. The string segment we are looking at is
268 // str.substring(start, end), which is guaranteed not to include any line breaks, i.e. end <= brk unless we are
269 // at the last segment.
270 int start = firstBrk + 1;
271 int brk = str.indexOf('\n', start);
273 // Loop over inner strings
275 trimLeadingAndAppend(sb, dquot, str, start, trimTrailing(str, start, brk)).append('\n');
277 brk = str.indexOf('\n', start);
280 return trimLeadingAndAppend(sb, dquot, str, start, length).toString();
283 private static StringBuilder trimLeadingAndAppend(final StringBuilder sb, final int dquot, final String str,
284 final int start, final int end) {
288 while (pos <= dquot) {
290 // We ran out of data, nothing to append
294 final char ch = str.charAt(offset);
296 // tabs are to be treated as 8 spaces
298 } else if (WHITESPACE_MATCHER.matches(ch)) {
307 // We have expanded beyond double quotes, push equivalent spaces
308 while (pos - 1 > dquot) {
313 return sb.append(str, offset, end);
316 private static int trimTrailing(final String str, final int start, final int end) {
318 while (ret > start) {
319 final int prev = ret - 1;
320 if (!WHITESPACE_MATCHER.matches(str.charAt(prev))) {