jaroslav@1646: /*
jaroslav@1646: * Copyright (c) 2007, 2011, Oracle and/or its affiliates. All rights reserved.
jaroslav@1646: * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
jaroslav@1646: *
jaroslav@1646: * This code is free software; you can redistribute it and/or modify it
jaroslav@1646: * under the terms of the GNU General Public License version 2 only, as
jaroslav@1646: * published by the Free Software Foundation. Oracle designates this
jaroslav@1646: * particular file as subject to the "Classpath" exception as provided
jaroslav@1646: * by Oracle in the LICENSE file that accompanied this code.
jaroslav@1646: *
jaroslav@1646: * This code is distributed in the hope that it will be useful, but WITHOUT
jaroslav@1646: * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
jaroslav@1646: * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
jaroslav@1646: * version 2 for more details (a copy is included in the LICENSE file that
jaroslav@1646: * accompanied this code).
jaroslav@1646: *
jaroslav@1646: * You should have received a copy of the GNU General Public License version
jaroslav@1646: * 2 along with this work; if not, write to the Free Software Foundation,
jaroslav@1646: * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
jaroslav@1646: *
jaroslav@1646: * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
jaroslav@1646: * or visit www.oracle.com if you need additional information or have any
jaroslav@1646: * questions.
jaroslav@1646: */
jaroslav@1646:
jaroslav@1646: package sun.invoke.util;
jaroslav@1646:
jaroslav@1646: /**
jaroslav@1646: * Utility routines for dealing with bytecode-level names.
jaroslav@1646: * Includes universal mangling rules for the JVM.
jaroslav@1646: *
jaroslav@1646: *
Avoiding Dangerous Characters
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * The JVM defines a very small set of characters which are illegal
jaroslav@1646: * in name spellings. We will slightly extend and regularize this set
jaroslav@1646: * into a group of dangerous characters.
jaroslav@1646: * These characters will then be replaced, in mangled names, by escape sequences.
jaroslav@1646: * In addition, accidental escape sequences must be further escaped.
jaroslav@1646: * Finally, a special prefix will be applied if and only if
jaroslav@1646: * the mangling would otherwise fail to begin with the escape character.
jaroslav@1646: * This happens to cover the corner case of the null string,
jaroslav@1646: * and also clearly marks symbols which need demangling.
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * Dangerous characters are the union of all characters forbidden
jaroslav@1646: * or otherwise restricted by the JVM specification,
jaroslav@1646: * plus their mates, if they are brackets
jaroslav@1646: * ([
and ]
,
jaroslav@1646: * <
and >
),
jaroslav@1646: * plus, arbitrarily, the colon character :
.
jaroslav@1646: * There is no distinction between type, method, and field names.
jaroslav@1646: * This makes it easier to convert between mangled names of different
jaroslav@1646: * types, since they do not need to be decoded (demangled).
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * The escape character is backslash \
jaroslav@1646: * (also known as reverse solidus).
jaroslav@1646: * This character is, until now, unheard of in bytecode names,
jaroslav@1646: * but traditional in the proposed role.
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * Replacement Characters
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * Every escape sequence is two characters
jaroslav@1646: * (in fact, two UTF8 bytes) beginning with
jaroslav@1646: * the escape character and followed by a
jaroslav@1646: * replacement character.
jaroslav@1646: * (Since the replacement character is never a backslash,
jaroslav@1646: * iterated manglings do not double in size.)
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * Each dangerous character has some rough visual similarity
jaroslav@1646: * to its corresponding replacement character.
jaroslav@1646: * This makes mangled symbols easier to recognize by sight.
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * The dangerous characters are
jaroslav@1646: * /
(forward slash, used to delimit package components),
jaroslav@1646: * .
(dot, also a package delimiter),
jaroslav@1646: * ;
(semicolon, used in signatures),
jaroslav@1646: * $
(dollar, used in inner classes and synthetic members),
jaroslav@1646: * <
(left angle),
jaroslav@1646: * >
(right angle),
jaroslav@1646: * [
(left square bracket, used in array types),
jaroslav@1646: * ]
(right square bracket, reserved in this scheme for language use),
jaroslav@1646: * and :
(colon, reserved in this scheme for language use).
jaroslav@1646: * Their replacements are, respectively,
jaroslav@1646: * |
(vertical bar),
jaroslav@1646: * ,
(comma),
jaroslav@1646: * ?
(question mark),
jaroslav@1646: * %
(percent),
jaroslav@1646: * ^
(caret),
jaroslav@1646: * _
(underscore), and
jaroslav@1646: * {
(left curly bracket),
jaroslav@1646: * }
(right curly bracket),
jaroslav@1646: * !
(exclamation mark).
jaroslav@1646: * In addition, the replacement character for the escape character itself is
jaroslav@1646: * -
(hyphen),
jaroslav@1646: * and the replacement character for the null prefix is
jaroslav@1646: * =
(equal sign).
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * An escape character \
jaroslav@1646: * followed by any of these replacement characters
jaroslav@1646: * is an escape sequence, and there are no other escape sequences.
jaroslav@1646: * An equal sign is only part of an escape sequence
jaroslav@1646: * if it is the second character in the whole string, following a backslash.
jaroslav@1646: * Two consecutive backslashes do not form an escape sequence.
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * Each escape sequence replaces a so-called original character
jaroslav@1646: * which is either one of the dangerous characters or the escape character.
jaroslav@1646: * A null prefix replaces an initial null string, not a character.
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * All this implies that escape sequences cannot overlap and may be
jaroslav@1646: * determined all at once for a whole string. Note that a spelling
jaroslav@1646: * string can contain accidental escapes, apparent escape
jaroslav@1646: * sequences which must not be interpreted as manglings.
jaroslav@1646: * These are disabled by replacing their leading backslash with an
jaroslav@1646: * escape sequence (\-
). To mangle a string, three logical steps
jaroslav@1646: * are required, though they may be carried out in one pass:
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * - In each accidental escape, replace the backslash with an escape sequence
jaroslav@1646: * (
\-
).
jaroslav@1646: * - Replace each dangerous character with an escape sequence
jaroslav@1646: * (
\|
for /
, etc.).
jaroslav@1646: * - If the first two steps introduced any change, and
jaroslav@1646: * if the string does not already begin with a backslash, prepend a null prefix (
\=
).
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * To demangle a mangled string that begins with an escape,
jaroslav@1646: * remove any null prefix, and then replace (in parallel)
jaroslav@1646: * each escape sequence by its original character.
jaroslav@1646: * Spelling strings which contain accidental
jaroslav@1646: * escapes must have them replaced, even if those
jaroslav@1646: * strings do not contain dangerous characters.
jaroslav@1646: * This restriction means that mangling a string always
jaroslav@1646: * requires a scan of the string for escapes.
jaroslav@1646: * But then, a scan would be required anyway,
jaroslav@1646: * to check for dangerous characters.
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * Nice Properties
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * If a bytecode name does not contain any escape sequence,
jaroslav@1646: * demangling is a no-op: The string demangles to itself.
jaroslav@1646: * Such a string is called self-mangling.
jaroslav@1646: * Almost all strings are self-mangling.
jaroslav@1646: * In practice, to demangle almost any name “found in nature”,
jaroslav@1646: * simply verify that it does not begin with a backslash.
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * Mangling is a one-to-one function, while demangling
jaroslav@1646: * is a many-to-one function.
jaroslav@1646: * A mangled string is defined as validly mangled if
jaroslav@1646: * it is in fact the unique mangling of its spelling string.
jaroslav@1646: * Three examples of invalidly mangled strings are \=foo
,
jaroslav@1646: * \-bar
, and baz\!
, which demangle to foo
, \bar
, and
jaroslav@1646: * baz\!
, but then remangle to foo
, \bar
, and \=baz\-!
.
jaroslav@1646: * If a language back-end or runtime is using mangled names,
jaroslav@1646: * it should never present an invalidly mangled bytecode
jaroslav@1646: * name to the JVM. If the runtime encounters one,
jaroslav@1646: * it should also report an error, since such an occurrence
jaroslav@1646: * probably indicates a bug in name encoding which
jaroslav@1646: * will lead to errors in linkage.
jaroslav@1646: * However, this note does not propose that the JVM verifier
jaroslav@1646: * detect invalidly mangled names.
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * As a result of these rules, it is a simple matter to
jaroslav@1646: * compute validly mangled substrings and concatenations
jaroslav@1646: * of validly mangled strings, and (with a little care)
jaroslav@1646: * these correspond to corresponding operations on their
jaroslav@1646: * spelling strings.
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * - Any prefix of a validly mangled string is also validly mangled,
jaroslav@1646: * although a null prefix may need to be removed.
jaroslav@1646: * - Any suffix of a validly mangled string is also validly mangled,
jaroslav@1646: * although a null prefix may need to be added.
jaroslav@1646: * - Two validly mangled strings, when concatenated,
jaroslav@1646: * are also validly mangled, although any null prefix
jaroslav@1646: * must be removed from the second string,
jaroslav@1646: * and a trailing backslash on the first string may need escaping,
jaroslav@1646: * if it would participate in an accidental escape when followed
jaroslav@1646: * by the first character of the second string.
jaroslav@1646: *
jaroslav@1646: * If languages that include non-Java symbol spellings use this
jaroslav@1646: * mangling convention, they will enjoy the following advantages:
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * - They can interoperate via symbols they share in common.
jaroslav@1646: * - Low-level tools, such as backtrace printers, will have readable displays.
jaroslav@1646: * - Future JVM and language extensions can safely use the dangerous characters
jaroslav@1646: * for structuring symbols, but will never interfere with valid spellings.
jaroslav@1646: * - Runtimes and compilers can use standard libraries for mangling and demangling.
jaroslav@1646: * - Occasional transliterations and name composition will be simple and regular,
jaroslav@1646: * for classes, methods, and fields.
jaroslav@1646: * - Bytecode names will continue to be compact.
jaroslav@1646: * When mangled, spellings will at most double in length, either in
jaroslav@1646: * UTF8 or UTF16 format, and most will not change at all.
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * Suggestions for Human Readable Presentations
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * For human readable displays of symbols,
jaroslav@1646: * it will be better to present a string-like quoted
jaroslav@1646: * representation of the spelling, because JVM users
jaroslav@1646: * are generally familiar with such tokens.
jaroslav@1646: * We suggest using single or double quotes before and after
jaroslav@1646: * mangled symbols which are not valid Java identifiers,
jaroslav@1646: * with quotes, backslashes, and non-printing characters
jaroslav@1646: * escaped as if for literals in the Java language.
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * For example, an HTML-like spelling
jaroslav@1646: * <pre>
mangles to
jaroslav@1646: * \^pre\_
and could
jaroslav@1646: * display more cleanly as
jaroslav@1646: * '<pre>'
,
jaroslav@1646: * with the quotes included.
jaroslav@1646: * Such string-like conventions are not suitable
jaroslav@1646: * for mangled bytecode names, in part because
jaroslav@1646: * dangerous characters must be eliminated, rather
jaroslav@1646: * than just quoted. Otherwise internally structured
jaroslav@1646: * strings like package prefixes and method signatures
jaroslav@1646: * could not be reliably parsed.
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * In such human-readable displays, invalidly mangled
jaroslav@1646: * names should not be demangled and quoted,
jaroslav@1646: * for this would be misleading. Likewise, JVM symbols
jaroslav@1646: * which contain dangerous characters (like dots in field
jaroslav@1646: * names or brackets in method names) should not be
jaroslav@1646: * simply quoted. The bytecode names
jaroslav@1646: * \=phase\,1
and
jaroslav@1646: * phase.1
are distinct,
jaroslav@1646: * and in demangled displays they should be presented as
jaroslav@1646: * 'phase.1'
and something like
jaroslav@1646: * 'phase'.1
, respectively.
jaroslav@1646: *
jaroslav@1646: *
jaroslav@1646: * @author John Rose
jaroslav@1646: * @version 1.2, 02/06/2008
jaroslav@1646: * @see http://blogs.sun.com/jrose/entry/symbolic_freedom_in_the_vm
jaroslav@1646: */
jaroslav@1646: public class BytecodeName {
jaroslav@1646: private BytecodeName() { } // static only class
jaroslav@1646:
jaroslav@1646: /** Given a source name, produce the corresponding bytecode name.
jaroslav@1646: * The source name should not be qualified, because any syntactic
jaroslav@1646: * markers (dots, slashes, dollar signs, colons, etc.) will be mangled.
jaroslav@1646: * @param s the source name
jaroslav@1646: * @return a valid bytecode name which represents the source name
jaroslav@1646: */
jaroslav@1646: public static String toBytecodeName(String s) {
jaroslav@1646: String bn = mangle(s);
jaroslav@1646: assert((Object)bn == s || looksMangled(bn)) : bn;
jaroslav@1646: assert(s.equals(toSourceName(bn))) : s;
jaroslav@1646: return bn;
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: /** Given an unqualified bytecode name, produce the corresponding source name.
jaroslav@1646: * The bytecode name must not contain dangerous characters.
jaroslav@1646: * In particular, it must not be qualified or segmented by colon {@code ':'}.
jaroslav@1646: * @param s the bytecode name
jaroslav@1646: * @return the source name, which may possibly have unsafe characters
jaroslav@1646: * @throws IllegalArgumentException if the bytecode name is not {@link #isSafeBytecodeName safe}
jaroslav@1646: * @see #isSafeBytecodeName(java.lang.String)
jaroslav@1646: */
jaroslav@1646: public static String toSourceName(String s) {
jaroslav@1646: checkSafeBytecodeName(s);
jaroslav@1646: String sn = s;
jaroslav@1646: if (looksMangled(s)) {
jaroslav@1646: sn = demangle(s);
jaroslav@1646: assert(s.equals(mangle(sn))) : s+" => "+sn+" => "+mangle(sn);
jaroslav@1646: }
jaroslav@1646: return sn;
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: /**
jaroslav@1646: * Given a bytecode name from a classfile, separate it into
jaroslav@1646: * components delimited by dangerous characters.
jaroslav@1646: * Each resulting array element will be either a dangerous character,
jaroslav@1646: * or else a safe bytecode name.
jaroslav@1646: * (The safe name might possibly be mangled to hide further dangerous characters.)
jaroslav@1646: * For example, the qualified class name {@code java/lang/String}
jaroslav@1646: * will be parsed into the array {@code {"java", '/', "lang", '/', "String"}}.
jaroslav@1646: * The name {@code <init>} will be parsed into { '<', "init", '>'}}
jaroslav@1646: * The name {@code foo/bar$:baz} will be parsed into
jaroslav@1646: * {@code {"foo", '/', "bar", '$', ':', "baz"}}.
jaroslav@1646: * The name {@code ::\=:foo:\=bar\!baz} will be parsed into
jaroslav@1646: * {@code {':', ':', "", ':', "foo", ':', "bar:baz"}}.
jaroslav@1646: */
jaroslav@1646: public static Object[] parseBytecodeName(String s) {
jaroslav@1646: int slen = s.length();
jaroslav@1646: Object[] res = null;
jaroslav@1646: for (int pass = 0; pass <= 1; pass++) {
jaroslav@1646: int fillp = 0;
jaroslav@1646: int lasti = 0;
jaroslav@1646: for (int i = 0; i <= slen; i++) {
jaroslav@1646: int whichDC = -1;
jaroslav@1646: if (i < slen) {
jaroslav@1646: whichDC = DANGEROUS_CHARS.indexOf(s.charAt(i));
jaroslav@1646: if (whichDC < DANGEROUS_CHAR_FIRST_INDEX) continue;
jaroslav@1646: }
jaroslav@1646: // got to end of string or next dangerous char
jaroslav@1646: if (lasti < i) {
jaroslav@1646: // normal component
jaroslav@1646: if (pass != 0)
jaroslav@1646: res[fillp] = toSourceName(s.substring(lasti, i));
jaroslav@1646: fillp++;
jaroslav@1646: lasti = i+1;
jaroslav@1646: }
jaroslav@1646: if (whichDC >= DANGEROUS_CHAR_FIRST_INDEX) {
jaroslav@1646: if (pass != 0)
jaroslav@1646: res[fillp] = DANGEROUS_CHARS_CA[whichDC];
jaroslav@1646: fillp++;
jaroslav@1646: lasti = i+1;
jaroslav@1646: }
jaroslav@1646: }
jaroslav@1646: if (pass != 0) break;
jaroslav@1646: // between passes, build the result array
jaroslav@1646: res = new Object[fillp];
jaroslav@1646: if (fillp <= 1 && lasti == 0) {
jaroslav@1646: if (fillp != 0) res[0] = toSourceName(s);
jaroslav@1646: break;
jaroslav@1646: }
jaroslav@1646: }
jaroslav@1646: return res;
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: /**
jaroslav@1646: * Given a series of components, create a bytecode name for a classfile.
jaroslav@1646: * This is the inverse of {@link #parseBytecodeName(java.lang.String)}.
jaroslav@1646: * Each component must either be an interned one-character string of
jaroslav@1646: * a dangerous character, or else a safe bytecode name.
jaroslav@1646: * @param components a series of name components
jaroslav@1646: * @return the concatenation of all components
jaroslav@1646: * @throws IllegalArgumentException if any component contains an unsafe
jaroslav@1646: * character, and is not an interned one-character string
jaroslav@1646: * @throws NullPointerException if any component is null
jaroslav@1646: */
jaroslav@1646: public static String unparseBytecodeName(Object[] components) {
jaroslav@1646: Object[] components0 = components;
jaroslav@1646: for (int i = 0; i < components.length; i++) {
jaroslav@1646: Object c = components[i];
jaroslav@1646: if (c instanceof String) {
jaroslav@1646: String mc = toBytecodeName((String) c);
jaroslav@1646: if (i == 0 && components.length == 1)
jaroslav@1646: return mc; // usual case
jaroslav@1646: if ((Object)mc != c) {
jaroslav@1646: if (components == components0)
jaroslav@1646: components = components.clone();
jaroslav@1646: components[i] = c = mc;
jaroslav@1646: }
jaroslav@1646: }
jaroslav@1646: }
jaroslav@1646: return appendAll(components);
jaroslav@1646: }
jaroslav@1646: private static String appendAll(Object[] components) {
jaroslav@1646: if (components.length <= 1) {
jaroslav@1646: if (components.length == 1) {
jaroslav@1646: return String.valueOf(components[0]);
jaroslav@1646: }
jaroslav@1646: return "";
jaroslav@1646: }
jaroslav@1646: int slen = 0;
jaroslav@1646: for (Object c : components) {
jaroslav@1646: if (c instanceof String)
jaroslav@1646: slen += String.valueOf(c).length();
jaroslav@1646: else
jaroslav@1646: slen += 1;
jaroslav@1646: }
jaroslav@1646: StringBuilder sb = new StringBuilder(slen);
jaroslav@1646: for (Object c : components) {
jaroslav@1646: sb.append(c);
jaroslav@1646: }
jaroslav@1646: return sb.toString();
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: /**
jaroslav@1646: * Given a bytecode name, produce the corresponding display name.
jaroslav@1646: * This is the source name, plus quotes if needed.
jaroslav@1646: * If the bytecode name contains dangerous characters,
jaroslav@1646: * assume that they are being used as punctuation,
jaroslav@1646: * and pass them through unchanged.
jaroslav@1646: * Non-empty runs of non-dangerous characters are demangled
jaroslav@1646: * if necessary, and the resulting names are quoted if
jaroslav@1646: * they are not already valid Java identifiers, or if
jaroslav@1646: * they contain a dangerous character (i.e., dollar sign "$").
jaroslav@1646: * Single quotes are used when quoting.
jaroslav@1646: * Within quoted names, embedded single quotes and backslashes
jaroslav@1646: * are further escaped by prepended backslashes.
jaroslav@1646: *
jaroslav@1646: * @param s the original bytecode name (which may be qualified)
jaroslav@1646: * @return a human-readable presentation
jaroslav@1646: */
jaroslav@1646: public static String toDisplayName(String s) {
jaroslav@1646: Object[] components = parseBytecodeName(s);
jaroslav@1646: for (int i = 0; i < components.length; i++) {
jaroslav@1646: if (!(components[i] instanceof String))
jaroslav@1646: continue;
jaroslav@1646: String sn = (String) components[i];
jaroslav@1646: // note that the name is already demangled!
jaroslav@1646: //sn = toSourceName(sn);
jaroslav@1646: if (!isJavaIdent(sn) || sn.indexOf('$') >=0 ) {
jaroslav@1646: components[i] = quoteDisplay(sn);
jaroslav@1646: }
jaroslav@1646: }
jaroslav@1646: return appendAll(components);
jaroslav@1646: }
jaroslav@1646: private static boolean isJavaIdent(String s) {
jaroslav@1646: int slen = s.length();
jaroslav@1646: if (slen == 0) return false;
jaroslav@1646: if (!Character.isJavaIdentifierStart(s.charAt(0)))
jaroslav@1646: return false;
jaroslav@1646: for (int i = 1; i < slen; i++) {
jaroslav@1646: if (!Character.isJavaIdentifierPart(s.charAt(i)))
jaroslav@1646: return false;
jaroslav@1646: }
jaroslav@1646: return true;
jaroslav@1646: }
jaroslav@1646: private static String quoteDisplay(String s) {
jaroslav@1646: // TO DO: Replace wierd characters in s by C-style escapes.
jaroslav@1646: return "'"+s.replaceAll("['\\\\]", "\\\\$0")+"'";
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: private static void checkSafeBytecodeName(String s)
jaroslav@1646: throws IllegalArgumentException {
jaroslav@1646: if (!isSafeBytecodeName(s)) {
jaroslav@1646: throw new IllegalArgumentException(s);
jaroslav@1646: }
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: /**
jaroslav@1646: * Report whether a simple name is safe as a bytecode name.
jaroslav@1646: * Such names are acceptable in class files as class, method, and field names.
jaroslav@1646: * Additionally, they are free of "dangerous" characters, even if those
jaroslav@1646: * characters are legal in some (or all) names in class files.
jaroslav@1646: * @param s the proposed bytecode name
jaroslav@1646: * @return true if the name is non-empty and all of its characters are safe
jaroslav@1646: */
jaroslav@1646: public static boolean isSafeBytecodeName(String s) {
jaroslav@1646: if (s.length() == 0) return false;
jaroslav@1646: // check occurrences of each DANGEROUS char
jaroslav@1646: for (char xc : DANGEROUS_CHARS_A) {
jaroslav@1646: if (xc == ESCAPE_C) continue; // not really that dangerous
jaroslav@1646: if (s.indexOf(xc) >= 0) return false;
jaroslav@1646: }
jaroslav@1646: return true;
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: /**
jaroslav@1646: * Report whether a character is safe in a bytecode name.
jaroslav@1646: * This is true of any unicode character except the following
jaroslav@1646: * dangerous characters: {@code ".;:$[]<>/"}.
jaroslav@1646: * @param s the proposed character
jaroslav@1646: * @return true if the character is safe to use in classfiles
jaroslav@1646: */
jaroslav@1646: public static boolean isSafeBytecodeChar(char c) {
jaroslav@1646: return DANGEROUS_CHARS.indexOf(c) < DANGEROUS_CHAR_FIRST_INDEX;
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: private static boolean looksMangled(String s) {
jaroslav@1646: return s.charAt(0) == ESCAPE_C;
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: private static String mangle(String s) {
jaroslav@1646: if (s.length() == 0)
jaroslav@1646: return NULL_ESCAPE;
jaroslav@1646:
jaroslav@1646: // build this lazily, when we first need an escape:
jaroslav@1646: StringBuilder sb = null;
jaroslav@1646:
jaroslav@1646: for (int i = 0, slen = s.length(); i < slen; i++) {
jaroslav@1646: char c = s.charAt(i);
jaroslav@1646:
jaroslav@1646: boolean needEscape = false;
jaroslav@1646: if (c == ESCAPE_C) {
jaroslav@1646: if (i+1 < slen) {
jaroslav@1646: char c1 = s.charAt(i+1);
jaroslav@1646: if ((i == 0 && c1 == NULL_ESCAPE_C)
jaroslav@1646: || c1 != originalOfReplacement(c1)) {
jaroslav@1646: // an accidental escape
jaroslav@1646: needEscape = true;
jaroslav@1646: }
jaroslav@1646: }
jaroslav@1646: } else {
jaroslav@1646: needEscape = isDangerous(c);
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: if (!needEscape) {
jaroslav@1646: if (sb != null) sb.append(c);
jaroslav@1646: continue;
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: // build sb if this is the first escape
jaroslav@1646: if (sb == null) {
jaroslav@1646: sb = new StringBuilder(s.length()+10);
jaroslav@1646: // mangled names must begin with a backslash:
jaroslav@1646: if (s.charAt(0) != ESCAPE_C && i > 0)
jaroslav@1646: sb.append(NULL_ESCAPE);
jaroslav@1646: // append the string so far, which is unremarkable:
jaroslav@1646: sb.append(s.substring(0, i));
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: // rewrite \ to \-, / to \|, etc.
jaroslav@1646: sb.append(ESCAPE_C);
jaroslav@1646: sb.append(replacementOf(c));
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: if (sb != null) return sb.toString();
jaroslav@1646:
jaroslav@1646: return s;
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: private static String demangle(String s) {
jaroslav@1646: // build this lazily, when we first meet an escape:
jaroslav@1646: StringBuilder sb = null;
jaroslav@1646:
jaroslav@1646: int stringStart = 0;
jaroslav@1646: if (s.startsWith(NULL_ESCAPE))
jaroslav@1646: stringStart = 2;
jaroslav@1646:
jaroslav@1646: for (int i = stringStart, slen = s.length(); i < slen; i++) {
jaroslav@1646: char c = s.charAt(i);
jaroslav@1646:
jaroslav@1646: if (c == ESCAPE_C && i+1 < slen) {
jaroslav@1646: // might be an escape sequence
jaroslav@1646: char rc = s.charAt(i+1);
jaroslav@1646: char oc = originalOfReplacement(rc);
jaroslav@1646: if (oc != rc) {
jaroslav@1646: // build sb if this is the first escape
jaroslav@1646: if (sb == null) {
jaroslav@1646: sb = new StringBuilder(s.length());
jaroslav@1646: // append the string so far, which is unremarkable:
jaroslav@1646: sb.append(s.substring(stringStart, i));
jaroslav@1646: }
jaroslav@1646: ++i; // skip both characters
jaroslav@1646: c = oc;
jaroslav@1646: }
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: if (sb != null)
jaroslav@1646: sb.append(c);
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: if (sb != null) return sb.toString();
jaroslav@1646:
jaroslav@1646: return s.substring(stringStart);
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: static char ESCAPE_C = '\\';
jaroslav@1646: // empty escape sequence to avoid a null name or illegal prefix
jaroslav@1646: static char NULL_ESCAPE_C = '=';
jaroslav@1646: static String NULL_ESCAPE = ESCAPE_C+""+NULL_ESCAPE_C;
jaroslav@1646:
jaroslav@1646: static final String DANGEROUS_CHARS = "\\/.;:$[]<>"; // \\ must be first
jaroslav@1646: static final String REPLACEMENT_CHARS = "-|,?!%{}^_";
jaroslav@1646: static final int DANGEROUS_CHAR_FIRST_INDEX = 1; // index after \\
jaroslav@1646: static char[] DANGEROUS_CHARS_A = DANGEROUS_CHARS.toCharArray();
jaroslav@1646: static char[] REPLACEMENT_CHARS_A = REPLACEMENT_CHARS.toCharArray();
jaroslav@1646: static final Character[] DANGEROUS_CHARS_CA;
jaroslav@1646: static {
jaroslav@1646: Character[] dcca = new Character[DANGEROUS_CHARS.length()];
jaroslav@1646: for (int i = 0; i < dcca.length; i++)
jaroslav@1646: dcca[i] = Character.valueOf(DANGEROUS_CHARS.charAt(i));
jaroslav@1646: DANGEROUS_CHARS_CA = dcca;
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646: static final long[] SPECIAL_BITMAP = new long[2]; // 128 bits
jaroslav@1646: static {
jaroslav@1646: String SPECIAL = DANGEROUS_CHARS + REPLACEMENT_CHARS;
jaroslav@1646: //System.out.println("SPECIAL = "+SPECIAL);
jaroslav@1646: for (char c : SPECIAL.toCharArray()) {
jaroslav@1646: SPECIAL_BITMAP[c >>> 6] |= 1L << c;
jaroslav@1646: }
jaroslav@1646: }
jaroslav@1646: static boolean isSpecial(char c) {
jaroslav@1646: if ((c >>> 6) < SPECIAL_BITMAP.length)
jaroslav@1646: return ((SPECIAL_BITMAP[c >>> 6] >> c) & 1) != 0;
jaroslav@1646: else
jaroslav@1646: return false;
jaroslav@1646: }
jaroslav@1646: static char replacementOf(char c) {
jaroslav@1646: if (!isSpecial(c)) return c;
jaroslav@1646: int i = DANGEROUS_CHARS.indexOf(c);
jaroslav@1646: if (i < 0) return c;
jaroslav@1646: return REPLACEMENT_CHARS.charAt(i);
jaroslav@1646: }
jaroslav@1646: static char originalOfReplacement(char c) {
jaroslav@1646: if (!isSpecial(c)) return c;
jaroslav@1646: int i = REPLACEMENT_CHARS.indexOf(c);
jaroslav@1646: if (i < 0) return c;
jaroslav@1646: return DANGEROUS_CHARS.charAt(i);
jaroslav@1646: }
jaroslav@1646: static boolean isDangerous(char c) {
jaroslav@1646: if (!isSpecial(c)) return false;
jaroslav@1646: return (DANGEROUS_CHARS.indexOf(c) >= DANGEROUS_CHAR_FIRST_INDEX);
jaroslav@1646: }
jaroslav@1646: static int indexOfDangerousChar(String s, int from) {
jaroslav@1646: for (int i = from, slen = s.length(); i < slen; i++) {
jaroslav@1646: if (isDangerous(s.charAt(i)))
jaroslav@1646: return i;
jaroslav@1646: }
jaroslav@1646: return -1;
jaroslav@1646: }
jaroslav@1646: static int lastIndexOfDangerousChar(String s, int from) {
jaroslav@1646: for (int i = Math.min(from, s.length()-1); i >= 0; i--) {
jaroslav@1646: if (isDangerous(s.charAt(i)))
jaroslav@1646: return i;
jaroslav@1646: }
jaroslav@1646: return -1;
jaroslav@1646: }
jaroslav@1646:
jaroslav@1646:
jaroslav@1646: }