/*
 * Copyright (c) 2012, 2016, Oracle and/or its affiliates. All rights reserved.
 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 *
 * This code is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License version 2 only, as
 * published by the Free Software Foundation.  Oracle designates this
 * particular file as subject to the "Classpath" exception as provided
 * by Oracle in the LICENSE file that accompanied this code.
 *
 * This code is distributed in the hope that it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 * version 2 for more details (a copy is included in the LICENSE file that
 * accompanied this code).
 *
 * You should have received a copy of the GNU General Public License version
 * 2 along with this work; if not, write to the Free Software Foundation,
 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
 *
 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 * or visit www.oracle.com if you need additional information or have any
 * questions.
 */

package com.sun.tools.javac.parser;

import java.text.BreakIterator;
import java.util.HashMap;
import java.util.Map;

import com.sun.source.doctree.AttributeTree.ValueKind;
import com.sun.tools.javac.parser.DocCommentParser.TagParser.Kind;
import com.sun.tools.javac.parser.Tokens.Comment;
import com.sun.tools.javac.parser.Tokens.TokenKind;
import com.sun.tools.javac.tree.DCTree;
import com.sun.tools.javac.tree.DCTree.DCAttribute;
import com.sun.tools.javac.tree.DCTree.DCDocComment;
import com.sun.tools.javac.tree.DCTree.DCEndPosTree;
import com.sun.tools.javac.tree.DCTree.DCErroneous;
import com.sun.tools.javac.tree.DCTree.DCIdentifier;
import com.sun.tools.javac.tree.DCTree.DCReference;
import com.sun.tools.javac.tree.DCTree.DCText;
import com.sun.tools.javac.tree.DocTreeMaker;
import com.sun.tools.javac.tree.JCTree;
import com.sun.tools.javac.util.DiagnosticSource;
import com.sun.tools.javac.util.List;
import com.sun.tools.javac.util.ListBuffer;
import com.sun.tools.javac.util.Log;
import com.sun.tools.javac.util.Name;
import com.sun.tools.javac.util.Names;
import com.sun.tools.javac.util.Position;

import static com.sun.tools.javac.util.LayoutCharacters.*;

This is NOT part of any supported API. If you write code that depends on this, you do so at your own risk. This code and its internal interfaces are subject to change or deletion without notice.

/** * * <p><b>This is NOT part of any supported API. * If you write code that depends on this, you do so at your own risk. * This code and its internal interfaces are subject to change or * deletion without notice.</b> */
public class DocCommentParser { static class ParseException extends Exception { private static final long serialVersionUID = 0; ParseException(String key) { super(key); } } final ParserFactory fac; final DiagnosticSource diagSource; final Comment comment; final DocTreeMaker m; final Names names; BreakIterator sentenceBreaker;
The input buffer, index of most recent character read, index of one past last character in buffer.
/** The input buffer, index of most recent character read, * index of one past last character in buffer. */
protected char[] buf; protected int bp; protected int buflen;
The current character.
/** The current character. */
protected char ch; int textStart = -1; int lastNonWhite = -1; boolean newline = true; Map<Name, TagParser> tagParsers; public DocCommentParser(ParserFactory fac, DiagnosticSource diagSource, Comment comment) { this.fac = fac; this.diagSource = diagSource; this.comment = comment; names = fac.names; m = fac.docTreeMaker; initTagParsers(); } public DocCommentParser(ParserFactory fac) { this(fac, null, null); } public DCDocComment parse() { String c = comment.getText(); buf = new char[c.length() + 1]; c.getChars(0, c.length(), buf, 0); buf[buf.length - 1] = EOI; buflen = buf.length - 1; bp = -1; nextChar(); List<DCTree> body = blockContent(); List<DCTree> tags = blockTags(); int pos = !body.isEmpty() ? body.head.pos : !tags.isEmpty() ? tags.head.pos : Position.NOPOS; DCDocComment dc = m.at(pos).newDocCommentTree(comment, body, tags); return dc; } void nextChar() { ch = buf[bp < buflen ? ++bp : buflen]; switch (ch) { case '\f': case '\n': case '\r': newline = true; } }
Read block content, consisting of text, html and inline tags. Terminated by the end of input, or the beginning of the next block tag: i.e. @ as the first non-whitespace character on a line.
/** * Read block content, consisting of text, html and inline tags. * Terminated by the end of input, or the beginning of the next block tag: * i.e. @ as the first non-whitespace character on a line. */
@SuppressWarnings("fallthrough") protected List<DCTree> blockContent() { ListBuffer<DCTree> trees = new ListBuffer<>(); textStart = -1; loop: while (bp < buflen) { switch (ch) { case '\n': case '\r': case '\f': newline = true; // fallthrough case ' ': case '\t': nextChar(); break; case '&': entity(trees); break; case '<': newline = false; addPendingText(trees, bp - 1); trees.add(html()); if (textStart == -1) { textStart = bp; lastNonWhite = -1; } break; case '>': newline = false; addPendingText(trees, bp - 1); trees.add(m.at(bp).newErroneousTree(newString(bp, bp + 1), diagSource, "dc.bad.gt")); nextChar(); if (textStart == -1) { textStart = bp; lastNonWhite = -1; } break; case '{': inlineTag(trees); break; case '@': if (newline) { addPendingText(trees, lastNonWhite); break loop; } // fallthrough default: newline = false; if (textStart == -1) textStart = bp; lastNonWhite = bp; nextChar(); } } if (lastNonWhite != -1) addPendingText(trees, lastNonWhite); return trees.toList(); }
Read a series of block tags, including their content. Standard tags parse their content appropriately. Non-standard tags are represented by UnknownBlockTag.
/** * Read a series of block tags, including their content. * Standard tags parse their content appropriately. * Non-standard tags are represented by {@link UnknownBlockTag}. */
protected List<DCTree> blockTags() { ListBuffer<DCTree> tags = new ListBuffer<>(); while (ch == '@') tags.add(blockTag()); return tags.toList(); }
Read a single block tag, including its content. Standard tags parse their content appropriately. Non-standard tags are represented by UnknownBlockTag.
/** * Read a single block tag, including its content. * Standard tags parse their content appropriately. * Non-standard tags are represented by {@link UnknownBlockTag}. */
protected DCTree blockTag() { int p = bp; try { nextChar(); if (isIdentifierStart(ch)) { Name name = readTagName(); TagParser tp = tagParsers.get(name); if (tp == null) { List<DCTree> content = blockContent(); return m.at(p).newUnknownBlockTagTree(name, content); } else { switch (tp.getKind()) { case BLOCK: return tp.parse(p); case INLINE: return erroneous("dc.bad.inline.tag", p); } } } blockContent(); return erroneous("dc.no.tag.name", p); } catch (ParseException e) { blockContent(); return erroneous(e.getMessage(), p); } } protected void inlineTag(ListBuffer<DCTree> list) { newline = false; nextChar(); if (ch == '@') { addPendingText(list, bp - 2); list.add(inlineTag()); textStart = bp; lastNonWhite = -1; } else { if (textStart == -1) textStart = bp - 1; lastNonWhite = bp; } }
Read a single inline tag, including its content. Standard tags parse their content appropriately. Non-standard tags are represented by UnknownBlockTag. Malformed tags may be returned as Erroneous.
/** * Read a single inline tag, including its content. * Standard tags parse their content appropriately. * Non-standard tags are represented by {@link UnknownBlockTag}. * Malformed tags may be returned as {@link Erroneous}. */
protected DCTree inlineTag() { int p = bp - 1; try { nextChar(); if (isIdentifierStart(ch)) { Name name = readTagName(); TagParser tp = tagParsers.get(name); if (tp == null) { skipWhitespace(); DCTree text = inlineText(WhitespaceRetentionPolicy.REMOVE_ALL); if (text != null) { nextChar(); return m.at(p).newUnknownInlineTagTree(name, List.of(text)).setEndPos(bp); } } else { if (!tp.retainWhiteSpace) { skipWhitespace(); } if (tp.getKind() == TagParser.Kind.INLINE) { DCEndPosTree<?> tree = (DCEndPosTree<?>) tp.parse(p); if (tree != null) { return tree.setEndPos(bp); } } else { // handle block tags (ex: @see) in inline content inlineText(WhitespaceRetentionPolicy.REMOVE_ALL); // skip content nextChar(); } } } return erroneous("dc.no.tag.name", p); } catch (ParseException e) { return erroneous(e.getMessage(), p); } } private static enum WhitespaceRetentionPolicy { RETAIN_ALL, REMOVE_FIRST_SPACE, REMOVE_ALL }
Read plain text content of an inline tag. Matching pairs of { } are skipped; the text is terminated by the first unmatched }. It is an error if the beginning of the next tag is detected.
/** * Read plain text content of an inline tag. * Matching pairs of { } are skipped; the text is terminated by the first * unmatched }. It is an error if the beginning of the next tag is detected. */
private DCTree inlineText(WhitespaceRetentionPolicy whitespacePolicy) throws ParseException { switch (whitespacePolicy) { case REMOVE_ALL: skipWhitespace(); break; case REMOVE_FIRST_SPACE: if (ch == ' ') nextChar(); break; case RETAIN_ALL: default: // do nothing break; } int pos = bp; int depth = 1; loop: while (bp < buflen) { switch (ch) { case '\n': case '\r': case '\f': newline = true; break; case ' ': case '\t': break; case '{': newline = false; lastNonWhite = bp; depth++; break; case '}': if (--depth == 0) { return m.at(pos).newTextTree(newString(pos, bp)); } newline = false; lastNonWhite = bp; break; case '@': if (newline) break loop; newline = false; lastNonWhite = bp; break; default: newline = false; lastNonWhite = bp; break; } nextChar(); } throw new ParseException("dc.unterminated.inline.tag"); }
Read Java class name, possibly followed by member Matching pairs of < > are skipped. The text is terminated by the first unmatched }. It is an error if the beginning of the next tag is detected.
/** * Read Java class name, possibly followed by member * Matching pairs of {@literal < >} are skipped. The text is terminated by the first * unmatched }. It is an error if the beginning of the next tag is detected. */
// TODO: boolean allowMember should be enum FORBID, ALLOW, REQUIRE // TODO: improve quality of parse to forbid bad constructions. // TODO: update to use ReferenceParser @SuppressWarnings("fallthrough") protected DCReference reference(boolean allowMember) throws ParseException { int pos = bp; int depth = 0; // scan to find the end of the signature, by looking for the first // whitespace not enclosed in () or <>, or the end of the tag loop: while (bp < buflen) { switch (ch) { case '\n': case '\r': case '\f': newline = true; // fallthrough case ' ': case '\t': if (depth == 0) break loop; break; case '(': case '<': newline = false; depth++; break; case ')': case '>': newline = false; --depth; break; case '}': if (bp == pos) return null; newline = false; break loop; case '@': if (newline) break loop; // fallthrough default: newline = false; } nextChar(); } if (depth != 0) throw new ParseException("dc.unterminated.signature"); String sig = newString(pos, bp); // Break sig apart into qualifiedExpr member paramTypes. JCTree qualExpr; Name member; List<JCTree> paramTypes; Log.DeferredDiagnosticHandler deferredDiagnosticHandler = new Log.DeferredDiagnosticHandler(fac.log); try { int hash = sig.indexOf("#"); int lparen = sig.indexOf("(", hash + 1); if (hash == -1) { if (lparen == -1) { qualExpr = parseType(sig); member = null; } else { qualExpr = null; member = parseMember(sig.substring(0, lparen)); } } else { qualExpr = (hash == 0) ? null : parseType(sig.substring(0, hash)); if (lparen == -1) member = parseMember(sig.substring(hash + 1)); else member = parseMember(sig.substring(hash + 1, lparen)); } if (lparen < 0) { paramTypes = null; } else { int rparen = sig.indexOf(")", lparen); if (rparen != sig.length() - 1) throw new ParseException("dc.ref.bad.parens"); paramTypes = parseParams(sig.substring(lparen + 1, rparen)); } if (!deferredDiagnosticHandler.getDiagnostics().isEmpty()) throw new ParseException("dc.ref.syntax.error"); } finally { fac.log.popDiagnosticHandler(deferredDiagnosticHandler); } return m.at(pos).newReferenceTree(sig, qualExpr, member, paramTypes).setEndPos(bp); } JCTree parseType(String s) throws ParseException { JavacParser p = fac.newParser(s, false, false, false); JCTree tree = p.parseType(); if (p.token().kind != TokenKind.EOF) throw new ParseException("dc.ref.unexpected.input"); return tree; } Name parseMember(String s) throws ParseException { JavacParser p = fac.newParser(s, false, false, false); Name name = p.ident(); if (p.token().kind != TokenKind.EOF) throw new ParseException("dc.ref.unexpected.input"); return name; } List<JCTree> parseParams(String s) throws ParseException { if (s.trim().isEmpty()) return List.nil(); JavacParser p = fac.newParser(s.replace("...", "[]"), false, false, false); ListBuffer<JCTree> paramTypes = new ListBuffer<>(); paramTypes.add(p.parseType()); if (p.token().kind == TokenKind.IDENTIFIER) p.nextToken(); while (p.token().kind == TokenKind.COMMA) { p.nextToken(); paramTypes.add(p.parseType()); if (p.token().kind == TokenKind.IDENTIFIER) p.nextToken(); } if (p.token().kind != TokenKind.EOF) throw new ParseException("dc.ref.unexpected.input"); return paramTypes.toList(); }
Read Java identifier Matching pairs of { } are skipped; the text is terminated by the first unmatched }. It is an error if the beginning of the next tag is detected.
/** * Read Java identifier * Matching pairs of { } are skipped; the text is terminated by the first * unmatched }. It is an error if the beginning of the next tag is detected. */
@SuppressWarnings("fallthrough") protected DCIdentifier identifier() throws ParseException { skipWhitespace(); int pos = bp; if (isJavaIdentifierStart(ch)) { Name name = readJavaIdentifier(); return m.at(pos).newIdentifierTree(name); } throw new ParseException("dc.identifier.expected"); }
Read a quoted string. It is an error if the beginning of the next tag is detected.
/** * Read a quoted string. * It is an error if the beginning of the next tag is detected. */
@SuppressWarnings("fallthrough") protected DCText quotedString() { int pos = bp; nextChar(); loop: while (bp < buflen) { switch (ch) { case '\n': case '\r': case '\f': newline = true; break; case ' ': case '\t': break; case '"': nextChar(); // trim trailing white-space? return m.at(pos).newTextTree(newString(pos, bp)); case '@': if (newline) break loop; } nextChar(); } return null; }
Read a term ie. one word. It is an error if the beginning of the next tag is detected.
/** * Read a term ie. one word. * It is an error if the beginning of the next tag is detected. */
@SuppressWarnings("fallthrough") protected DCText inlineWord() { int pos = bp; int depth = 0; loop: while (bp < buflen) { switch (ch) { case '\n': newline = true; // fallthrough case '\r': case '\f': case ' ': case '\t': return m.at(pos).newTextTree(newString(pos, bp)); case '@': if (newline) break loop; case '{': depth++; break; case '}': if (depth == 0 || --depth == 0) return m.at(pos).newTextTree(newString(pos, bp)); break; } newline = false; nextChar(); } return null; }
Read general text content of an inline tag, including HTML entities and elements. Matching pairs of { } are skipped; the text is terminated by the first unmatched }. It is an error if the beginning of the next tag is detected.
/** * Read general text content of an inline tag, including HTML entities and elements. * Matching pairs of { } are skipped; the text is terminated by the first * unmatched }. It is an error if the beginning of the next tag is detected. */
@SuppressWarnings("fallthrough") private List<DCTree> inlineContent() { ListBuffer<DCTree> trees = new ListBuffer<>(); skipWhitespace(); int pos = bp; int depth = 1; textStart = -1; loop: while (bp < buflen) { switch (ch) { case '\n': case '\r': case '\f': newline = true; // fall through case ' ': case '\t': nextChar(); break; case '&': entity(trees); break; case '<': newline = false; addPendingText(trees, bp - 1); trees.add(html()); break; case '{': if (textStart == -1) textStart = bp; newline = false; depth++; nextChar(); break; case '}': newline = false; if (--depth == 0) { addPendingText(trees, bp - 1); nextChar(); return trees.toList(); } nextChar(); break; case '@': if (newline) break loop; // fallthrough default: if (textStart == -1) textStart = bp; nextChar(); break; } } return List.of(erroneous("dc.unterminated.inline.tag", pos)); } protected void entity(ListBuffer<DCTree> list) { newline = false; addPendingText(list, bp - 1); list.add(entity()); if (textStart == -1) { textStart = bp; lastNonWhite = -1; } }
Read an HTML entity. &identifier; or &#digits; or &#xhex-digits;
/** * Read an HTML entity. * {@literal &identifier; } or {@literal &#digits; } or {@literal &#xhex-digits; } */
protected DCTree entity() { int p = bp; nextChar(); Name name = null; if (ch == '#') { int namep = bp; nextChar(); if (isDecimalDigit(ch)) { nextChar(); while (isDecimalDigit(ch)) nextChar(); name = names.fromChars(buf, namep, bp - namep); } else if (ch == 'x' || ch == 'X') { nextChar(); if (isHexDigit(ch)) { nextChar(); while (isHexDigit(ch)) nextChar(); name = names.fromChars(buf, namep, bp - namep); } } } else if (isIdentifierStart(ch)) { name = readIdentifier(); } if (name == null) return erroneous("dc.bad.entity", p); else { if (ch != ';') return erroneous("dc.missing.semicolon", p); nextChar(); return m.at(p).newEntityTree(name); } }
Read the start or end of an HTML tag, or an HTML comment <identifier attrs> or </identifier>
/** * Read the start or end of an HTML tag, or an HTML comment * {@literal <identifier attrs> } or {@literal </identifier> } */
protected DCTree html() { int p = bp; nextChar(); if (isIdentifierStart(ch)) { Name name = readIdentifier(); List<DCTree> attrs = htmlAttrs(); if (attrs != null) { boolean selfClosing = false; if (ch == '/') { nextChar(); selfClosing = true; } if (ch == '>') { nextChar(); DCTree dctree = m.at(p).newStartElementTree(name, attrs, selfClosing).setEndPos(bp); return dctree; } } } else if (ch == '/') { nextChar(); if (isIdentifierStart(ch)) { Name name = readIdentifier(); skipWhitespace(); if (ch == '>') { nextChar(); return m.at(p).newEndElementTree(name); } } } else if (ch == '!') { nextChar(); if (ch == '-') { nextChar(); if (ch == '-') { nextChar(); while (bp < buflen) { int dash = 0; while (ch == '-') { dash++; nextChar(); } // Strictly speaking, a comment should not contain "--" // so dash > 2 is an error, dash == 2 implies ch == '>' // See http://www.w3.org/TR/html-markup/syntax.html#syntax-comments // for more details. if (dash >= 2 && ch == '>') { nextChar(); return m.at(p).newCommentTree(newString(p, bp)); } nextChar(); } } } } bp = p + 1; ch = buf[bp]; return erroneous("dc.malformed.html", p); }
Read a series of HTML attributes, terminated by > . Each attribute is of the form identifier[=value] . "value" may be unquoted, single-quoted, or double-quoted.
/** * Read a series of HTML attributes, terminated by {@literal > }. * Each attribute is of the form {@literal identifier[=value] }. * "value" may be unquoted, single-quoted, or double-quoted. */
protected List<DCTree> htmlAttrs() { ListBuffer<DCTree> attrs = new ListBuffer<>(); skipWhitespace(); loop: while (isIdentifierStart(ch)) { int namePos = bp; Name name = readAttributeName(); skipWhitespace(); List<DCTree> value = null; ValueKind vkind = ValueKind.EMPTY; if (ch == '=') { ListBuffer<DCTree> v = new ListBuffer<>(); nextChar(); skipWhitespace(); if (ch == '\'' || ch == '"') { vkind = (ch == '\'') ? ValueKind.SINGLE : ValueKind.DOUBLE; char quote = ch; nextChar(); textStart = bp; while (bp < buflen && ch != quote) { if (newline && ch == '@') { attrs.add(erroneous("dc.unterminated.string", namePos)); // No point trying to read more. // In fact, all attrs get discarded by the caller // and superseded by a malformed.html node because // the html tag itself is not terminated correctly. break loop; } attrValueChar(v); } addPendingText(v, bp - 1); nextChar(); } else { vkind = ValueKind.UNQUOTED; textStart = bp; while (bp < buflen && !isUnquotedAttrValueTerminator(ch)) { attrValueChar(v); } addPendingText(v, bp - 1); } skipWhitespace(); value = v.toList(); } DCAttribute attr = m.at(namePos).newAttributeTree(name, vkind, value); attrs.add(attr); } return attrs.toList(); } protected void attrValueChar(ListBuffer<DCTree> list) { switch (ch) { case '&': entity(list); break; case '{': inlineTag(list); break; default: nextChar(); } } protected void addPendingText(ListBuffer<DCTree> list, int textEnd) { if (textStart != -1) { if (textStart <= textEnd) { list.add(m.at(textStart).newTextTree(newString(textStart, textEnd + 1))); } textStart = -1; } } protected DCErroneous erroneous(String code, int pos) { int i = bp - 1; loop: while (i > pos) { switch (buf[i]) { case '\f': case '\n': case '\r': newline = true; break; case '\t': case ' ': break; default: break loop; } i--; } textStart = -1; return m.at(pos).newErroneousTree(newString(pos, i + 1), diagSource, code); } protected boolean isIdentifierStart(char ch) { return Character.isUnicodeIdentifierStart(ch); } protected Name readIdentifier() { int start = bp; nextChar(); while (bp < buflen && Character.isUnicodeIdentifierPart(ch)) nextChar(); return names.fromChars(buf, start, bp - start); } protected Name readAttributeName() { int start = bp; nextChar(); while (bp < buflen && (Character.isUnicodeIdentifierPart(ch) || ch == '-')) nextChar(); return names.fromChars(buf, start, bp - start); } protected Name readTagName() { int start = bp; nextChar(); while (bp < buflen && (Character.isUnicodeIdentifierPart(ch) || ch == '.' || ch == '-' || ch == ':')) { nextChar(); } return names.fromChars(buf, start, bp - start); } protected boolean isJavaIdentifierStart(char ch) { return Character.isJavaIdentifierStart(ch); } protected Name readJavaIdentifier() { int start = bp; nextChar(); while (bp < buflen && Character.isJavaIdentifierPart(ch)) nextChar(); return names.fromChars(buf, start, bp - start); } protected boolean isDecimalDigit(char ch) { return ('0' <= ch && ch <= '9'); } protected boolean isHexDigit(char ch) { return ('0' <= ch && ch <= '9') || ('a' <= ch && ch <= 'f') || ('A' <= ch && ch <= 'F'); } protected boolean isUnquotedAttrValueTerminator(char ch) { switch (ch) { case '\f': case '\n': case '\r': case '\t': case ' ': case '"': case '\'': case '`': case '=': case '<': case '>': return true; default: return false; } } protected boolean isWhitespace(char ch) { return Character.isWhitespace(ch); } protected void skipWhitespace() { while (isWhitespace(ch)) { nextChar(); } }
Params:
  • start – position of first character of string
  • end – position of character beyond last character to be included
/** * @param start position of first character of string * @param end position of character beyond last character to be included */
String newString(int start, int end) { return new String(buf, start, end - start); } static abstract class TagParser { enum Kind { INLINE, BLOCK } final Kind kind; final DCTree.Kind treeKind; final boolean retainWhiteSpace; TagParser(Kind k, DCTree.Kind tk) { kind = k; treeKind = tk; retainWhiteSpace = false; } TagParser(Kind k, DCTree.Kind tk, boolean retainWhiteSpace) { kind = k; treeKind = tk; this.retainWhiteSpace = retainWhiteSpace; } Kind getKind() { return kind; } DCTree.Kind getTreeKind() { return treeKind; } abstract DCTree parse(int pos) throws ParseException; }
See Also:
/** * @see <a href="http://docs.oracle.com/javase/8/docs/technotes/tools/unix/javadoc.html#CHDJGIJB">Javadoc Tags</a> */
private void initTagParsers() { TagParser[] parsers = { // @author name-text new TagParser(Kind.BLOCK, DCTree.Kind.AUTHOR) { public DCTree parse(int pos) { List<DCTree> name = blockContent(); return m.at(pos).newAuthorTree(name); } }, // {@code text} new TagParser(Kind.INLINE, DCTree.Kind.CODE, true) { public DCTree parse(int pos) throws ParseException { DCTree text = inlineText(WhitespaceRetentionPolicy.REMOVE_FIRST_SPACE); nextChar(); return m.at(pos).newCodeTree((DCText) text); } }, // @deprecated deprecated-text new TagParser(Kind.BLOCK, DCTree.Kind.DEPRECATED) { public DCTree parse(int pos) { List<DCTree> reason = blockContent(); return m.at(pos).newDeprecatedTree(reason); } }, // {@docRoot} new TagParser(Kind.INLINE, DCTree.Kind.DOC_ROOT) { public DCTree parse(int pos) throws ParseException { if (ch == '}') { nextChar(); return m.at(pos).newDocRootTree(); } inlineText(WhitespaceRetentionPolicy.REMOVE_ALL); // skip unexpected content nextChar(); throw new ParseException("dc.unexpected.content"); } }, // @exception class-name description new TagParser(Kind.BLOCK, DCTree.Kind.EXCEPTION) { public DCTree parse(int pos) throws ParseException { skipWhitespace(); DCReference ref = reference(false); List<DCTree> description = blockContent(); return m.at(pos).newExceptionTree(ref, description); } }, // @hidden hidden-text new TagParser(Kind.BLOCK, DCTree.Kind.HIDDEN) { public DCTree parse(int pos) { List<DCTree> reason = blockContent(); return m.at(pos).newHiddenTree(reason); } }, // @index search-term options-description new TagParser(Kind.INLINE, DCTree.Kind.INDEX) { public DCTree parse(int pos) throws ParseException { skipWhitespace(); if (ch == '}') { throw new ParseException("dc.no.content"); } DCTree term = ch == '"' ? quotedString() : inlineWord(); if (term == null) { throw new ParseException("dc.no.content"); } skipWhitespace(); List<DCTree> description = List.nil(); if (ch != '}') { description = inlineContent(); } else { nextChar(); } return m.at(pos).newIndexTree(term, description); } }, // {@inheritDoc} new TagParser(Kind.INLINE, DCTree.Kind.INHERIT_DOC) { public DCTree parse(int pos) throws ParseException { if (ch == '}') { nextChar(); return m.at(pos).newInheritDocTree(); } inlineText(WhitespaceRetentionPolicy.REMOVE_ALL); // skip unexpected content nextChar(); throw new ParseException("dc.unexpected.content"); } }, // {@link package.class#member label} new TagParser(Kind.INLINE, DCTree.Kind.LINK) { public DCTree parse(int pos) throws ParseException { DCReference ref = reference(true); List<DCTree> label = inlineContent(); return m.at(pos).newLinkTree(ref, label); } }, // {@linkplain package.class#member label} new TagParser(Kind.INLINE, DCTree.Kind.LINK_PLAIN) { public DCTree parse(int pos) throws ParseException { DCReference ref = reference(true); List<DCTree> label = inlineContent(); return m.at(pos).newLinkPlainTree(ref, label); } }, // {@literal text} new TagParser(Kind.INLINE, DCTree.Kind.LITERAL, true) { public DCTree parse(int pos) throws ParseException { DCTree text = inlineText(WhitespaceRetentionPolicy.REMOVE_FIRST_SPACE); nextChar(); return m.at(pos).newLiteralTree((DCText) text); } }, // @param parameter-name description new TagParser(Kind.BLOCK, DCTree.Kind.PARAM) { public DCTree parse(int pos) throws ParseException { skipWhitespace(); boolean typaram = false; if (ch == '<') { typaram = true; nextChar(); } DCIdentifier id = identifier(); if (typaram) { if (ch != '>') throw new ParseException("dc.gt.expected"); nextChar(); } skipWhitespace(); List<DCTree> desc = blockContent(); return m.at(pos).newParamTree(typaram, id, desc); } }, // @provides service-name description new TagParser(Kind.BLOCK, DCTree.Kind.PROVIDES) { public DCTree parse(int pos) throws ParseException { skipWhitespace(); DCReference ref = reference(true); List<DCTree> description = blockContent(); return m.at(pos).newProvidesTree(ref, description); } }, // @return description new TagParser(Kind.BLOCK, DCTree.Kind.RETURN) { public DCTree parse(int pos) { List<DCTree> description = blockContent(); return m.at(pos).newReturnTree(description); } }, // @see reference | quoted-string | HTML new TagParser(Kind.BLOCK, DCTree.Kind.SEE) { public DCTree parse(int pos) throws ParseException { skipWhitespace(); switch (ch) { case '"': DCText string = quotedString(); if (string != null) { skipWhitespace(); if (ch == '@' || ch == EOI && bp == buf.length - 1) { return m.at(pos).newSeeTree(List.<DCTree>of(string)); } } break; case '<': List<DCTree> html = blockContent(); if (html != null) return m.at(pos).newSeeTree(html); break; case '@': if (newline) throw new ParseException("dc.no.content"); break; case EOI: if (bp == buf.length - 1) throw new ParseException("dc.no.content"); break; default: if (isJavaIdentifierStart(ch) || ch == '#') { DCReference ref = reference(true); List<DCTree> description = blockContent(); return m.at(pos).newSeeTree(description.prepend(ref)); } } throw new ParseException("dc.unexpected.content"); } }, // @serialData data-description new TagParser(Kind.BLOCK, DCTree.Kind.SERIAL_DATA) { public DCTree parse(int pos) { List<DCTree> description = blockContent(); return m.at(pos).newSerialDataTree(description); } }, // @serialField field-name field-type description new TagParser(Kind.BLOCK, DCTree.Kind.SERIAL_FIELD) { public DCTree parse(int pos) throws ParseException { skipWhitespace(); DCIdentifier name = identifier(); skipWhitespace(); DCReference type = reference(false); List<DCTree> description = null; if (isWhitespace(ch)) { skipWhitespace(); description = blockContent(); } return m.at(pos).newSerialFieldTree(name, type, description); } }, // @serial field-description | include | exclude new TagParser(Kind.BLOCK, DCTree.Kind.SERIAL) { public DCTree parse(int pos) { List<DCTree> description = blockContent(); return m.at(pos).newSerialTree(description); } }, // @since since-text new TagParser(Kind.BLOCK, DCTree.Kind.SINCE) { public DCTree parse(int pos) { List<DCTree> description = blockContent(); return m.at(pos).newSinceTree(description); } }, // @throws class-name description new TagParser(Kind.BLOCK, DCTree.Kind.THROWS) { public DCTree parse(int pos) throws ParseException { skipWhitespace(); DCReference ref = reference(false); List<DCTree> description = blockContent(); return m.at(pos).newThrowsTree(ref, description); } }, // @uses service-name description new TagParser(Kind.BLOCK, DCTree.Kind.USES) { public DCTree parse(int pos) throws ParseException { skipWhitespace(); DCReference ref = reference(true); List<DCTree> description = blockContent(); return m.at(pos).newUsesTree(ref, description); } }, // {@value package.class#field} new TagParser(Kind.INLINE, DCTree.Kind.VALUE) { public DCTree parse(int pos) throws ParseException { DCReference ref = reference(true); skipWhitespace(); if (ch == '}') { nextChar(); return m.at(pos).newValueTree(ref); } nextChar(); throw new ParseException("dc.unexpected.content"); } }, // @version version-text new TagParser(Kind.BLOCK, DCTree.Kind.VERSION) { public DCTree parse(int pos) { List<DCTree> description = blockContent(); return m.at(pos).newVersionTree(description); } }, }; tagParsers = new HashMap<>(); for (TagParser p: parsers) tagParsers.put(names.fromString(p.getTreeKind().tagName), p); } }