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