001/*
002 * Licensed to the Apache Software Foundation (ASF) under one
003 * or more contributor license agreements.  See the NOTICE file
004 * distributed with this work for additional information
005 * regarding copyright ownership.  The ASF licenses this file
006 * to you under the Apache License, Version 2.0 (the
007 * "License"); you may not use this file except in compliance
008 * with the License.  You may obtain a copy of the License at
009 *
010 *   https://www.apache.org/licenses/LICENSE-2.0
011 *
012 * Unless required by applicable law or agreed to in writing,
013 * software distributed under the License is distributed on an
014 * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
015 * KIND, either express or implied.  See the License for the
016 * specific language governing permissions and limitations
017 * under the License.
018 */
019package org.apache.bcel.util;
020
021import java.util.ArrayList;
022import java.util.Arrays;
023import java.util.HashMap;
024import java.util.Iterator;
025import java.util.List;
026import java.util.Map;
027import java.util.regex.Matcher;
028import java.util.regex.Pattern;
029
030import org.apache.bcel.Const;
031import org.apache.bcel.generic.ClassGenException;
032import org.apache.bcel.generic.InstructionHandle;
033import org.apache.bcel.generic.InstructionList;
034import org.apache.commons.lang3.StringUtils;
035
036/**
037 * InstructionFinder is a tool to search for given instructions patterns, i.e., match sequences of instructions in an
038 * instruction list via regular expressions. This can be used, for example, in order to implement a peep hole optimizer that
039 * looks for code patterns and replaces them with faster equivalents.
040 *
041 * <p>
042 * This class internally uses the java.util.regex package to search for regular expressions.
043 *
044 * A typical application would look like this:
045 *
046 * <pre>
047 *
048 *
049 *   InstructionFinder f   = new InstructionFinder(il);
050 *   String            pat = &quot;IfInstruction ICONST_0 GOTO ICONST_1 NOP (IFEQ|IFNE)&quot;;
051 *
052 *   for (Iterator i = f.search(pat, constraint); i.hasNext(); ) {
053 *   InstructionHandle[] match = (InstructionHandle[])i.next();
054 *   ...
055 *   il.delete(match[1], match[5]);
056 *   ...
057 *   }
058 *
059 *
060 * </pre>
061 *
062 * @see org.apache.bcel.generic.Instruction
063 * @see InstructionList
064 */
065public class InstructionFinder {
066
067    /**
068     * Code patterns found may be checked using an additional user-defined constraint object whether they really match the
069     * needed criterion. I.e., check constraints that cannot expressed with regular expressions.
070     */
071    public interface CodeConstraint {
072
073        /**
074         * @param match array of instructions matching the requested pattern
075         * @return true if the matched area is really useful
076         */
077        boolean checkCode(InstructionHandle[] match);
078    }
079
080    private static final int OFFSET = 32767; // char + OFFSET is outside of LATIN-1
081    private static final int NO_OPCODES = 256; // Potential number, some are not used
082    private static final Map<String, String> map = new HashMap<>();
083
084    // Initialize pattern map
085    static {
086        map.put("arithmeticinstruction",
087            "(irem|lrem|iand|ior|ineg|isub|lneg|fneg|fmul|ldiv|fadd|lxor|frem|idiv|land|ixor|ishr|fsub|lshl|fdiv|iadd|lor|dmul|lsub|ishl|imul|lmul|lushr|dneg|iushr|lshr|ddiv|drem|dadd|ladd|dsub)");
088        map.put("invokeinstruction", "(invokevirtual|invokeinterface|invokestatic|invokespecial|invokedynamic)");
089        map.put("arrayinstruction",
090            "(baload|aastore|saload|caload|fastore|lastore|iaload|castore|iastore|aaload|bastore|sastore|faload|laload|daload|dastore)");
091        map.put("gotoinstruction", "(goto|goto_w)");
092        map.put("conversioninstruction", "(d2l|l2d|i2s|d2i|l2i|i2b|l2f|d2f|f2i|i2d|i2l|f2d|i2c|f2l|i2f)");
093        map.put("localvariableinstruction", "(fstore|iinc|lload|dstore|dload|iload|aload|astore|istore|fload|lstore)");
094        map.put("loadinstruction", "(fload|dload|lload|iload|aload)");
095        map.put("fieldinstruction", "(getfield|putstatic|getstatic|putfield)");
096        map.put("cpinstruction",
097            "(ldc2_w|invokeinterface|invokedynamic|multianewarray|putstatic|instanceof|getstatic|checkcast|getfield|invokespecial|ldc_w|invokestatic|invokevirtual|putfield|ldc|new|anewarray)");
098        map.put("stackinstruction", "(dup2|swap|dup2_x2|pop|pop2|dup|dup2_x1|dup_x2|dup_x1)");
099        map.put("branchinstruction",
100            "(ifle|if_acmpne|if_icmpeq|if_acmpeq|ifnonnull|goto_w|iflt|ifnull|if_icmpne|tableswitch|if_icmple|ifeq|if_icmplt|jsr_w|if_icmpgt|ifgt|jsr|goto|ifne|ifge|lookupswitch|if_icmpge)");
101        map.put("returninstruction", "(lreturn|ireturn|freturn|dreturn|areturn|return)");
102        map.put("storeinstruction", "(istore|fstore|dstore|astore|lstore)");
103        map.put("select", "(tableswitch|lookupswitch)");
104        map.put("ifinstruction",
105            "(ifeq|ifgt|if_icmpne|if_icmpeq|ifge|ifnull|ifne|if_icmple|if_icmpge|if_acmpeq|if_icmplt|if_acmpne|ifnonnull|iflt|if_icmpgt|ifle)");
106        map.put("jsrinstruction", "(jsr|jsr_w)");
107        map.put("variablelengthinstruction", "(tableswitch|jsr|goto|lookupswitch)");
108        map.put("unconditionalbranch", "(goto|jsr|jsr_w|athrow|goto_w)");
109        map.put("constantpushinstruction", "(dconst|bipush|sipush|fconst|iconst|lconst)");
110        map.put("typedinstruction",
111            "(imul|lsub|aload|fload|lor|new|aaload|fcmpg|iand|iaload|lrem|idiv|d2l|isub|dcmpg|dastore|ret|f2d|f2i|drem|iinc|i2c|checkcast|frem|lreturn|astore|lushr|daload|dneg|fastore|istore|lshl|ldiv|lstore|areturn|ishr|ldc_w|invokeinterface|invokedynamic|aastore|lxor|ishl|l2d|i2f|return|faload|sipush|iushr|caload|instanceof|invokespecial|putfield|fmul|ireturn|laload|d2f|lneg|ixor|i2l|fdiv|lastore|multianewarray|i2b|getstatic|i2d|putstatic|fcmpl|saload|ladd|irem|dload|jsr_w|dconst|dcmpl|fsub|freturn|ldc|aconst_null|castore|lmul|ldc2_w|dadd|iconst|f2l|ddiv|dstore|land|jsr|anewarray|dmul|bipush|dsub|sastore|d2i|i2s|lshr|iadd|l2i|lload|bastore|fstore|fneg|iload|fadd|baload|fconst|ior|ineg|dreturn|l2f|lconst|getfield|invokevirtual|invokestatic|iastore)");
112        map.put("popinstruction", "(fstore|dstore|pop|pop2|astore|putstatic|istore|lstore)");
113        map.put("allocationinstruction", "(multianewarray|new|anewarray|newarray)");
114        map.put("indexedinstruction",
115            "(lload|lstore|fload|ldc2_w|invokeinterface|invokedynamic|multianewarray|astore|dload|putstatic|instanceof|getstatic|checkcast|getfield|invokespecial|dstore|istore|iinc|ldc_w|ret|fstore|invokestatic|iload|putfield|invokevirtual|ldc|new|aload|anewarray)");
116        map.put("pushinstruction", "(dup|lload|dup2|bipush|fload|ldc2_w|sipush|lconst|fconst|dload|getstatic|ldc_w|aconst_null|dconst|iload|ldc|iconst|aload)");
117        map.put("stackproducer",
118            "(imul|lsub|aload|fload|lor|new|aaload|fcmpg|iand|iaload|lrem|idiv|d2l|isub|dcmpg|dup|f2d|f2i|drem|i2c|checkcast|frem|lushr|daload|dneg|lshl|ldiv|ishr|ldc_w|invokeinterface|invokedynamic|lxor|ishl|l2d|i2f|faload|sipush|iushr|caload|instanceof|invokespecial|fmul|laload|d2f|lneg|ixor|i2l|fdiv|getstatic|i2b|swap|i2d|dup2|fcmpl|saload|ladd|irem|dload|jsr_w|dconst|dcmpl|fsub|ldc|arraylength|aconst_null|tableswitch|lmul|ldc2_w|iconst|dadd|f2l|ddiv|land|jsr|anewarray|dmul|bipush|dsub|d2i|newarray|i2s|lshr|iadd|lload|l2i|fneg|iload|fadd|baload|fconst|lookupswitch|ior|ineg|lconst|l2f|getfield|invokevirtual|invokestatic)");
119        map.put("stackconsumer",
120            "(imul|lsub|lor|iflt|fcmpg|if_icmpgt|iand|ifeq|if_icmplt|lrem|ifnonnull|idiv|d2l|isub|dcmpg|dastore|if_icmpeq|f2d|f2i|drem|i2c|checkcast|frem|lreturn|astore|lushr|pop2|monitorexit|dneg|fastore|istore|lshl|ldiv|lstore|areturn|if_icmpge|ishr|monitorenter|invokeinterface|invokedynamic|aastore|lxor|ishl|l2d|i2f|return|iushr|instanceof|invokespecial|fmul|ireturn|d2f|lneg|ixor|pop|i2l|ifnull|fdiv|lastore|i2b|if_acmpeq|ifge|swap|i2d|putstatic|fcmpl|ladd|irem|dcmpl|fsub|freturn|ifgt|castore|lmul|dadd|f2l|ddiv|dstore|land|if_icmpne|if_acmpne|dmul|dsub|sastore|ifle|d2i|i2s|lshr|iadd|l2i|bastore|fstore|fneg|fadd|ior|ineg|ifne|dreturn|l2f|if_icmple|getfield|invokevirtual|invokestatic|iastore)");
121        map.put("exceptionthrower",
122            "(irem|lrem|laload|putstatic|baload|dastore|areturn|getstatic|ldiv|anewarray|iastore|castore|idiv|saload|lastore|fastore|putfield|lreturn|caload|getfield|return|aastore|freturn|newarray|instanceof|multianewarray|athrow|faload|iaload|aaload|dreturn|monitorenter|checkcast|bastore|arraylength|new|invokevirtual|sastore|ldc_w|ireturn|invokespecial|monitorexit|invokeinterface|invokedynamic|ldc|invokestatic|daload)");
123        map.put("loadclass",
124            "(multianewarray|invokeinterface|invokedynamic|instanceof|invokespecial|putfield|checkcast|putstatic|invokevirtual|new|getstatic|invokestatic|getfield|anewarray)");
125        map.put("instructiontargeter",
126            "(ifle|if_acmpne|if_icmpeq|if_acmpeq|ifnonnull|goto_w|iflt|ifnull|if_icmpne|tableswitch|if_icmple|ifeq|if_icmplt|jsr_w|if_icmpgt|ifgt|jsr|goto|ifne|ifge|lookupswitch|if_icmpge)");
127        // Some aliases
128        map.put("if_icmp", "(if_icmpne|if_icmpeq|if_icmple|if_icmpge|if_icmplt|if_icmpgt)");
129        map.put("if_acmp", "(if_acmpeq|if_acmpne)");
130        map.put("if", "(ifeq|ifne|iflt|ifge|ifgt|ifle)");
131        // Precompile some aliases first
132        map.put("iconst", precompile(Const.ICONST_0, Const.ICONST_5, Const.ICONST_M1));
133        map.put("lconst", new String(new char[] {'(', makeChar(Const.LCONST_0), '|', makeChar(Const.LCONST_1), ')'}));
134        map.put("dconst", new String(new char[] {'(', makeChar(Const.DCONST_0), '|', makeChar(Const.DCONST_1), ')'}));
135        map.put("fconst", new String(new char[] {'(', makeChar(Const.FCONST_0), '|', makeChar(Const.FCONST_1), '|', makeChar(Const.FCONST_2), ')'}));
136        map.put("lload", precompile(Const.LLOAD_0, Const.LLOAD_3, Const.LLOAD));
137        map.put("iload", precompile(Const.ILOAD_0, Const.ILOAD_3, Const.ILOAD));
138        map.put("dload", precompile(Const.DLOAD_0, Const.DLOAD_3, Const.DLOAD));
139        map.put("fload", precompile(Const.FLOAD_0, Const.FLOAD_3, Const.FLOAD));
140        map.put("aload", precompile(Const.ALOAD_0, Const.ALOAD_3, Const.ALOAD));
141        map.put("lstore", precompile(Const.LSTORE_0, Const.LSTORE_3, Const.LSTORE));
142        map.put("istore", precompile(Const.ISTORE_0, Const.ISTORE_3, Const.ISTORE));
143        map.put("dstore", precompile(Const.DSTORE_0, Const.DSTORE_3, Const.DSTORE));
144        map.put("fstore", precompile(Const.FSTORE_0, Const.FSTORE_3, Const.FSTORE));
145        map.put("astore", precompile(Const.ASTORE_0, Const.ASTORE_3, Const.ASTORE));
146        // Compile strings
147        map.forEach((key, value) -> {
148            final char ch = value.charAt(1); // Omit already precompiled patterns
149            if (ch < OFFSET) {
150                map.put(key, compilePattern(value)); // precompile all patterns
151            }
152        });
153        // Add instruction alias to match anything
154        final StringBuilder buf = new StringBuilder("(");
155        for (short i = 0; i < NO_OPCODES; i++) {
156            if (Const.getNoOfOperands(i) != Const.UNDEFINED) { // Not an invalid opcode
157                buf.append(makeChar(i));
158                if (i < NO_OPCODES - 1) {
159                    buf.append('|');
160                }
161            }
162        }
163        buf.append(')');
164        map.put("instruction", buf.toString());
165    }
166
167    /**
168     * Replace symbolic names of instructions with the appropriate character and remove all white space from string. Meta
169     * characters such as +, * are ignored.
170     *
171     * @param pattern The pattern to compile
172     * @return translated regular expression string
173     */
174    private static String compilePattern(final String pattern) {
175        // Bug: BCEL-77 - Instructions are assumed to be english, to avoid odd Locale issues
176        final String lower = StringUtils.toRootLowerCase(pattern);
177        final StringBuilder buf = new StringBuilder();
178        final int size = pattern.length();
179        for (int i = 0; i < size; i++) {
180            char ch = lower.charAt(i);
181            if (Character.isLetterOrDigit(ch)) {
182                final StringBuilder name = new StringBuilder();
183                while ((Character.isLetterOrDigit(ch) || ch == '_') && i < size) {
184                    name.append(ch);
185                    if (++i >= size) {
186                        break;
187                    }
188                    ch = lower.charAt(i);
189                }
190                i--;
191                buf.append(mapName(name.toString()));
192            } else if (!Character.isWhitespace(ch)) {
193                buf.append(ch);
194            }
195        }
196        return buf.toString();
197    }
198
199    /**
200     * Convert opcode number to char.
201     */
202    private static char makeChar(final short opcode) {
203        return (char) (opcode + OFFSET);
204    }
205
206    /**
207     * Map symbolic instruction names like "getfield" to a single character.
208     *
209     * @param pattern instruction pattern in lower case
210     * @return encoded string for a pattern such as "BranchInstruction".
211     */
212    private static String mapName(final String pattern) {
213        final String result = map.get(pattern);
214        if (result != null) {
215            return result;
216        }
217        for (short i = 0; i < NO_OPCODES; i++) {
218            if (pattern.equals(Const.getOpcodeName(i))) {
219                return String.valueOf(makeChar(i));
220            }
221        }
222        throw new IllegalArgumentException("Instruction unknown: " + pattern);
223    }
224
225    private static String precompile(final short from, final short to, final short extra) {
226        final StringBuilder buf = new StringBuilder("(");
227        for (short i = from; i <= to; i++) {
228            buf.append(makeChar(i));
229            buf.append('|');
230        }
231        buf.append(makeChar(extra));
232        buf.append(")");
233        return buf.toString();
234    }
235
236    private final InstructionList il;
237
238    private String ilString; // instruction list as string
239
240    private InstructionHandle[] handles; // map instruction
241
242    // list to array
243    /**
244     * @param il instruction list to search for given patterns
245     */
246    public InstructionFinder(final InstructionList il) {
247        this.il = il;
248        reread();
249    }
250
251    /**
252     * @return the inquired instruction list
253     */
254    public final InstructionList getInstructionList() {
255        return il;
256    }
257
258    /**
259     * @return the matched piece of code as an array of instruction (handles)
260     */
261    private InstructionHandle[] getMatch(final int matchedFrom, final int matchLength) {
262        return Arrays.copyOfRange(handles, matchedFrom, matchedFrom + matchLength);
263    }
264
265    /**
266     * Reread the instruction list, for example, after you've altered the list upon a match.
267     */
268    public final void reread() {
269        final int size = il.getLength();
270        final char[] buf = new char[size]; // Create a string with length equal to il length
271        handles = il.getInstructionHandles();
272        // Map opcodes to characters
273        for (int i = 0; i < size; i++) {
274            buf[i] = makeChar(handles[i].getInstruction().getOpcode());
275        }
276        ilString = new String(buf);
277    }
278
279    /**
280     * Start search beginning from the start of the given instruction list.
281     *
282     * @param pattern the instruction pattern to search for, where case is ignored
283     * @return iterator of matches where e.nextElement() returns an array of instruction handles describing the matched area
284     */
285    public final Iterator<InstructionHandle[]> search(final String pattern) {
286        return search(pattern, il.getStart(), null);
287    }
288
289    /**
290     * Start search beginning from the start of the given instruction list. Check found matches with the constraint object.
291     *
292     * @param pattern the instruction pattern to search for, case is ignored
293     * @param constraint constraints to be checked on matching code
294     * @return instruction handle or 'null' if the match failed
295     */
296    public final Iterator<InstructionHandle[]> search(final String pattern, final CodeConstraint constraint) {
297        return search(pattern, il.getStart(), constraint);
298    }
299
300    /**
301     * Start search beginning from 'from'.
302     *
303     * @param pattern the instruction pattern to search for, where case is ignored
304     * @param from where to start the search in the instruction list
305     * @return iterator of matches where e.nextElement() returns an array of instruction handles describing the matched area
306     */
307    public final Iterator<InstructionHandle[]> search(final String pattern, final InstructionHandle from) {
308        return search(pattern, from, null);
309    }
310
311    /**
312     * Search for the given pattern in the instruction list. You can search for any valid opcode via its symbolic name, for example
313     * "istore". You can also use a super class or an interface name to match a whole set of instructions, for example
314     * "BranchInstruction" or "LoadInstruction". "istore" is also an alias for all "istore_x" instructions. Additional
315     * aliases are "if" for "ifxx", "if_icmp" for "if_icmpxx", "if_acmp" for "if_acmpxx".
316     *
317     * Consecutive instruction names must be separated by white space which will be removed during the compilation of the
318     * pattern.
319     *
320     * For the rest the usual pattern matching rules for regular expressions apply.
321     * <P>
322     * Example pattern:
323     *
324     * <pre>
325     * search(&quot;BranchInstruction NOP ((IfInstruction|GOTO)+ ISTORE Instruction)*&quot;);
326     * </pre>
327     *
328     * <p>
329     * If you alter the instruction list upon a match such that other matching areas are affected, you should call reread()
330     * to update the finder and call search() again, because the matches are cached.
331     *
332     * @param pattern the instruction pattern to search for, where case is ignored
333     * @param from where to start the search in the instruction list
334     * @param constraint optional CodeConstraint to check the found code pattern for user-defined constraints
335     * @return iterator of matches where e.nextElement() returns an array of instruction handles describing the matched area
336     */
337    public final Iterator<InstructionHandle[]> search(final String pattern, final InstructionHandle from, final CodeConstraint constraint) {
338        final String search = compilePattern(pattern);
339        int start = -1;
340        for (int i = 0; i < handles.length; i++) {
341            if (handles[i] == from) {
342                start = i; // Where to start search from (index)
343                break;
344            }
345        }
346        if (start == -1) {
347            throw new ClassGenException("Instruction handle " + from + " not found in instruction list.");
348        }
349        final Pattern regex = Pattern.compile(search);
350        final List<InstructionHandle[]> matches = new ArrayList<>();
351        final Matcher matcher = regex.matcher(ilString);
352        while (start < ilString.length() && matcher.find(start)) {
353            final int startExpr = matcher.start();
354            final int endExpr = matcher.end();
355            final int lenExpr = endExpr - startExpr;
356            final InstructionHandle[] match = getMatch(startExpr, lenExpr);
357            if (constraint == null || constraint.checkCode(match)) {
358                matches.add(match);
359            }
360            start = endExpr;
361        }
362        return matches.iterator();
363    }
364
365    /*
366     * Internal debugging routines.
367     */
368//    private static final String pattern2string( String pattern ) {
369//        return pattern2string(pattern, true);
370//    }
371
372//    private static final String pattern2string( String pattern, boolean make_string ) {
373//        StringBuilder buf = new StringBuilder();
374//        for (int i = 0; i < pattern.length(); i++) {
375//            char ch = pattern.charAt(i);
376//            if (ch >= OFFSET) {
377//                if (make_string) {
378//                    buf.append(Constants.getOpcodeName(ch - OFFSET));
379//                } else {
380//                    buf.append((ch - OFFSET));
381//                }
382//            } else {
383//                buf.append(ch);
384//            }
385//        }
386//        return buf.toString();
387//    }
388}