*/
package org.opendaylight.yangtools.yang.parser.rfc7950.repo;
+import static com.google.common.base.Verify.verify;
+
import com.google.common.annotations.VisibleForTesting;
import com.google.common.base.CharMatcher;
-import java.util.Collections;
-import java.util.List;
-import java.util.regex.Pattern;
+import com.google.common.base.VerifyException;
+import org.antlr.v4.runtime.Token;
+import org.antlr.v4.runtime.tree.ParseTree;
import org.antlr.v4.runtime.tree.TerminalNode;
-import org.opendaylight.yangtools.antlrv4.code.gen.YangStatementParser.ArgumentContext;
+import org.eclipse.jdt.annotation.NonNull;
import org.opendaylight.yangtools.yang.common.YangVersion;
+import org.opendaylight.yangtools.yang.parser.antlr.YangStatementParser;
+import org.opendaylight.yangtools.yang.parser.antlr.YangStatementParser.ArgumentContext;
+import org.opendaylight.yangtools.yang.parser.antlr.YangStatementParser.QuotedStringContext;
+import org.opendaylight.yangtools.yang.parser.antlr.YangStatementParser.UnquotedStringContext;
import org.opendaylight.yangtools.yang.parser.spi.source.SourceException;
import org.opendaylight.yangtools.yang.parser.spi.source.StatementSourceReference;
-final class ArgumentContextUtils {
- private static final CharMatcher WHITESPACE_MATCHER = CharMatcher.whitespace();
- private static final CharMatcher ANYQUOTE_MATCHER = CharMatcher.anyOf("'\"");
- private static final Pattern ESCAPED_DQUOT = Pattern.compile("\\\"", Pattern.LITERAL);
- private static final Pattern ESCAPED_BACKSLASH = Pattern.compile("\\\\", Pattern.LITERAL);
- private static final Pattern ESCAPED_LF = Pattern.compile("\\n", Pattern.LITERAL);
- private static final Pattern ESCAPED_TAB = Pattern.compile("\\t", Pattern.LITERAL);
-
- private ArgumentContextUtils() {
- throw new UnsupportedOperationException();
- }
+/**
+ * Utilities for dealing with YANG statement argument strings, encapsulated in ANTLR grammar's ArgumentContext.
+ */
+abstract class ArgumentContextUtils {
+ /**
+ * YANG 1.0 version of strings, which were not completely clarified in
+ * <a href="https://tools.ietf.org/html/rfc6020#section-6.1.3">RFC6020</a>.
+ */
+ private static final class RFC6020 extends ArgumentContextUtils {
+ private static final @NonNull RFC6020 INSTANCE = new RFC6020();
- static String stringFromStringContext(final ArgumentContext context, final YangVersion yangVersion,
- final StatementSourceReference ref) {
- final StringBuilder sb = new StringBuilder();
- List<TerminalNode> strings = context.STRING();
- if (strings.isEmpty()) {
- strings = Collections.singletonList(context.IDENTIFIER());
- }
- for (final TerminalNode stringNode : strings) {
- final String str = stringNode.getText();
- final char firstChar = str.charAt(0);
- final char lastChar = str.charAt(str.length() - 1);
- if (firstChar == '"' && lastChar == '"') {
- final String innerStr = str.substring(1, str.length() - 1);
- /*
- * Unescape escaped double quotes, tabs, new line and backslash
- * in the inner string and trim the result.
- */
- checkDoubleQuotedString(innerStr, yangVersion, ref);
-
- sb.append(ESCAPED_TAB.matcher(
- ESCAPED_LF.matcher(
- ESCAPED_BACKSLASH.matcher(
- ESCAPED_DQUOT.matcher(
- trimWhitespace(innerStr, stringNode.getSymbol().getCharPositionInLine()))
- .replaceAll("\\\""))
- .replaceAll("\\\\"))
- .replaceAll("\\\n"))
- .replaceAll("\\\t"));
- } else if (firstChar == '\'' && lastChar == '\'') {
- /*
- * According to RFC6020 a single quote character cannot occur in
- * a single-quoted string, even when preceded by a backslash.
- */
- sb.append(str, 1, str.length() - 1);
- } else {
- checkUnquotedString(str, yangVersion, ref);
- sb.append(str);
- }
+ @Override
+ void checkDoubleQuoted(final String str, final StatementSourceReference ref, final int backslash) {
+ // No-op
}
- return sb.toString();
- }
- private static void checkUnquotedString(final String str, final YangVersion yangVersion,
- final StatementSourceReference ref) {
- if (yangVersion == YangVersion.VERSION_1_1) {
- SourceException.throwIf(ANYQUOTE_MATCHER.matchesAnyOf(str), ref,
- "YANG 1.1: unquoted string (%s) contains illegal characters", str);
+ @Override
+ void checkUnquoted(final String str, final StatementSourceReference ref) {
+ // No-op
}
}
- private static void checkDoubleQuotedString(final String str, final YangVersion yangVersion,
- final StatementSourceReference ref) {
- if (yangVersion == YangVersion.VERSION_1_1) {
- for (int i = 0; i < str.length() - 1; i++) {
- if (str.charAt(i) == '\\') {
- switch (str.charAt(i + 1)) {
+ /**
+ * YANG 1.1 version of strings, which were clarified in
+ * <a href="https://tools.ietf.org/html/rfc7950#section-6.1.3">RFC7950</a>.
+ */
+ // NOTE: the differences clarified lead to a proper ability to delegate this to ANTLR lexer, but that does not
+ // understand versions and needs to work with both.
+ private static final class RFC7950 extends ArgumentContextUtils {
+ private static final CharMatcher ANYQUOTE_MATCHER = CharMatcher.anyOf("'\"");
+ private static final @NonNull RFC7950 INSTANCE = new RFC7950();
+
+ @Override
+ void checkDoubleQuoted(final String str, final StatementSourceReference ref, final int backslash) {
+ if (backslash < str.length() - 1) {
+ int index = backslash;
+ while (index != -1) {
+ switch (str.charAt(index + 1)) {
case 'n':
case 't':
case '\\':
case '\"':
- i++;
+ index = str.indexOf('\\', index + 2);
break;
default:
throw new SourceException(ref, "YANG 1.1: illegal double quoted string (%s). In double "
- + "quoted string the backslash must be followed by one of the following character "
- + "[n,t,\",\\], but was '%s'.", str, str.charAt(i + 1));
+ + "quoted string the backslash must be followed by one of the following character "
+ + "[n,t,\",\\], but was '%s'.", str, str.charAt(index + 1));
}
}
}
}
+
+ @Override
+ void checkUnquoted(final String str, final StatementSourceReference ref) {
+ SourceException.throwIf(ANYQUOTE_MATCHER.matchesAnyOf(str), ref,
+ "YANG 1.1: unquoted string (%s) contains illegal characters", str);
+ }
+ }
+
+ private static final CharMatcher WHITESPACE_MATCHER = CharMatcher.whitespace();
+
+ private ArgumentContextUtils() {
+ // Hidden on purpose
+ }
+
+ static @NonNull ArgumentContextUtils forVersion(final YangVersion version) {
+ switch (version) {
+ case VERSION_1:
+ return RFC6020.INSTANCE;
+ case VERSION_1_1:
+ return RFC7950.INSTANCE;
+ default:
+ throw new IllegalStateException("Unhandled version " + version);
+ }
+ }
+
+ // TODO: teach the only caller about versions, or provide common-enough idioms for its use case
+ static @NonNull ArgumentContextUtils rfc6020() {
+ return RFC6020.INSTANCE;
+ }
+
+ /*
+ * NOTE: this method we do not use convenience methods provided by generated parser code, but instead are making
+ * based on the grammar assumptions. While this is more verbose, it cuts out a number of unnecessary code,
+ * such as intermediate List allocation et al.
+ */
+ final @NonNull String stringFromStringContext(final ArgumentContext context, final StatementSourceReference ref) {
+ // Get first child, which we fully expect to exist and be a lexer token
+ final ParseTree firstChild = context.getChild(0);
+ if (firstChild instanceof UnquotedStringContext) {
+ // Simple case, just grab the text, as ANTLR has done all the heavy lifting
+ final String str = firstChild.getText();
+ checkUnquoted(str, ref);
+ return str;
+ }
+
+ verify(firstChild instanceof QuotedStringContext, "Unexpected shape of %s", context);
+ if (context.getChildCount() == 1) {
+ // No concatenation needed, special-case
+ return unquoteString((QuotedStringContext) firstChild, ref);
+ }
+
+ // Potentially-complex case of string quoting, escaping and concatenation.
+ return concatStrings(context, ref);
+ }
+
+ private String unquoteString(final QuotedStringContext context, final StatementSourceReference ref) {
+ final ParseTree secondChild = context.getChild(1);
+ verify(secondChild instanceof TerminalNode, "Unexpected shape of %s", context);
+ final Token secondToken = ((TerminalNode) secondChild).getSymbol();
+ final int type = secondToken.getType();
+ switch (type) {
+ case YangStatementParser.DQUOT_END:
+ case YangStatementParser.SQUOT_END:
+ // We are missing actual body, hence this is an empty string
+ return "";
+ case YangStatementParser.SQUOT_STRING:
+ return secondChild.getText();
+ case YangStatementParser.DQUOT_STRING:
+ // We should be looking at the first token, which is DQUOT_START, but since it is a single-character
+ // token, let's not bother.
+ return normalizeDoubleQuoted(secondChild.getText(), secondToken.getCharPositionInLine() - 1, ref);
+ default:
+ throw new VerifyException("Unhandled token type " + type);
+ }
+ }
+
+ private String concatStrings(final ArgumentContext context, final StatementSourceReference ref) {
+ /*
+ * We have multiple fragments. Just search the tree. This code is equivalent to
+ *
+ * context.quotedString().forEach(stringNode -> sb.append(unquoteString(stringNode, ref))
+ *
+ * except we minimize allocations which that would do.
+ */
+ final StringBuilder sb = new StringBuilder();
+ for (ParseTree child : context.children) {
+ if (child instanceof TerminalNode) {
+ final TerminalNode childNode = (TerminalNode) child;
+ switch (childNode.getSymbol().getType()) {
+ case YangStatementParser.SEP:
+ case YangStatementParser.PLUS:
+ // Operator, which we are handling by concat
+ break;
+ default:
+ throw new VerifyException("Unexpected symbol in " + childNode);
+ }
+ } else {
+ verify(child instanceof QuotedStringContext, "Unexpected fragment component %s", child);
+ sb.append(unquoteString((QuotedStringContext) child, ref));
+ continue;
+ }
+ }
+ return sb.toString();
+ }
+
+ private String normalizeDoubleQuoted(final String str, final int dquot, final StatementSourceReference ref) {
+ // Whitespace normalization happens irrespective of further handling and has no effect on the result
+ final String stripped = trimWhitespace(str, dquot);
+
+ // Now we need to perform some amount of unescaping. This serves as a pre-check before we dispatch
+ // validation and processing (which will reuse the work we have done)
+ final int backslash = stripped.indexOf('\\');
+ return backslash == -1 ? stripped : unescape(ref, stripped, backslash);
+ }
+
+ /*
+ * NOTE: Enforcement and transformation logic done by these methods should logically reside in the lexer and ANTLR
+ * account the for it with lexer modes. We do not want to force a re-lexing phase in the parser just because
+ * we decided to let ANTLR do the work.
+ */
+ abstract void checkDoubleQuoted(String str, StatementSourceReference ref, int backslash);
+
+ abstract void checkUnquoted(String str, StatementSourceReference ref);
+
+ /*
+ * Unescape escaped double quotes, tabs, new line and backslash in the inner string and trim the result.
+ */
+ private String unescape(final StatementSourceReference ref, final String str, final int backslash) {
+ checkDoubleQuoted(str, ref, backslash);
+ StringBuilder sb = new StringBuilder(str.length());
+ unescapeBackslash(sb, str, backslash);
+ return sb.toString();
+ }
+
+ @VisibleForTesting
+ static void unescapeBackslash(final StringBuilder sb, final String str, final int backslash) {
+ String substring = str;
+ int backslashIndex = backslash;
+ while (true) {
+ int nextIndex = backslashIndex + 1;
+ if (backslashIndex != -1 && nextIndex < substring.length()) {
+ replaceBackslash(sb, substring, nextIndex);
+ substring = substring.substring(nextIndex + 1);
+ if (substring.length() > 0) {
+ backslashIndex = substring.indexOf('\\');
+ } else {
+ break;
+ }
+ } else {
+ sb.append(substring);
+ break;
+ }
+ }
+ }
+
+ private static void replaceBackslash(final StringBuilder sb, final String str, final int nextAfterBackslash) {
+ int backslash = nextAfterBackslash - 1;
+ sb.append(str, 0, backslash);
+ final char c = str.charAt(nextAfterBackslash);
+ switch (c) {
+ case '\\':
+ case '"':
+ sb.append(c);
+ break;
+ case 't':
+ sb.append('\t');
+ break;
+ case 'n':
+ sb.append('\n');
+ break;
+ default:
+ sb.append(str, backslash, nextAfterBackslash + 1);
+ }
}
@VisibleForTesting
static String trimWhitespace(final String str, final int dquot) {
- int brk = str.indexOf('\n');
- if (brk == -1) {
- // No need to trim whitespace
+ final int firstBrk = str.indexOf('\n');
+ if (firstBrk == -1) {
return str;
}
final StringBuilder sb = new StringBuilder(length);
// Append first segment, which needs only tail-trimming
- sb.append(str, 0, trimTrailing(str, 0, brk)).append('\n');
+ sb.append(str, 0, trimTrailing(str, 0, firstBrk)).append('\n');
// With that out of the way, setup our iteration state. The string segment we are looking at is
// str.substring(start, end), which is guaranteed not to include any line breaks, i.e. end <= brk unless we are
// at the last segment.
- int start = brk + 1;
- brk = str.indexOf('\n', start);
+ int start = firstBrk + 1;
+ int brk = str.indexOf('\n', start);
// Loop over inner strings
while (brk != -1) {