Aviator源码：Aviator表达式引擎执行过程源码分析
目录
1.if执行脚本示例
2.源码分析
2.1 compile执行过程
2.1.1 CodeGenerator
2.1.2 ExpressionParser
2.1.3 if脚本ASM反编译结果
2.2 compiledExpression.execute执行过程
3.总结概述
由于Aviator支持的脚本语法较多，下面通过项目中使用较多的if语句来对aviator执行过程进行一次源码剖析，以便更清楚的了解aviator具体的解析执行过程，并加深技术储备，方便后续aviator源码相关问题的快速深入分析，同时希望达到“窥一斑而知全貌”的效果；
1.if执行脚本示例
Map env = new HashMap<>();
env.put("a", 2);
env.put("b", 3);
env.put("c", 4);
Object result = AviatorEvaluator.execute("if(a > 1) { return b+2; } else { return c; }", env); 通过执行上述的if脚本，着重分析aviator框架内部的解析逻辑，比如通过asm字节码技术动态生成class、LambdaFunctionBootstrap构造和设计理念以及LambdaFunction函数的构造和设计理念等；
2.源码分析
跟踪AviatorEvaluator.execute方法，内部实现如下：
/**
* Execute a text expression with environment
*
* @param cacheKey unique key for caching
* @param expression text expression
* @param env Binding variable environment
* @param cached Whether to cache the compiled result,make true to cache it.
*/
public Object execute(final String cacheKey, final String expression,
final Map env, final boolean cached) {
Expression compiledExpression = compile(cacheKey, expression, cached);
if (compiledExpression != null) {
return compiledExpression.execute(env);
} else {
throw new ExpressionNotFoundException("Null compiled expression for " + expression);
}
} 其中主要包含2个部分：
1）compile方法通过asm技术生成ClassExpression的子类：这里将整个表达式脚本编译为ClassExpression子类；
2）执行compiledExpression.execute方法：将外部传入的变量集env实参带入方法中进行执行，返回脚本执行结果；
下面对这2个部分分别进行分析：
2.1 compile执行过程
compile方法体中使用了LRU或HashMap对编译后的字节码进行缓存，编译的字节码是通过方法innerCompile来获取的：
private Expression compile(final String cacheKey, final String expression,
final String sourceFile, final boolean cached) {
if (expression == null || expression.trim().length() == 0) {
throw new CompileExpressionErrorException("Blank expression");
}
if (cacheKey == null || cacheKey.trim().length() == 0) {
throw new CompileExpressionErrorException("Blank cacheKey");
}
if (cached) {
FutureTask existedTask = null;
if (this.expressionLRUCache != null) {
boolean runTask = false;
synchronized (this.expressionLRUCache) {
existedTask = this.expressionLRUCache.get(cacheKey);
if (existedTask == null) {
existedTask = newCompileTask(expression, sourceFile, cached);
runTask = true;
this.expressionLRUCache.put(cacheKey, existedTask);
}
}
if (runTask) {
existedTask.run();
}
} else {
FutureTask task = this.expressionCache.get(cacheKey);
if (task != null) {
return getCompiledExpression(expression, task);
}
task = newCompileTask(expression, sourceFile, cached);
existedTask = this.expressionCache.putIfAbsent(cacheKey, task);
if (existedTask == null) {
existedTask = task;
existedTask.run();
}
}
return getCompiledExpression(cacheKey, existedTask);
} else {
return innerCompile(expression, sourceFile, cached);
}
} innerCompile方法实现如下：
private Expression innerCompile(final String expression, final String sourceFile,
final boolean cached) {
ExpressionLexer lexer = new ExpressionLexer(this, expression);
CodeGenerator codeGenerator = newCodeGenerator(sourceFile, cached);
ExpressionParser parser = new ExpressionParser(this, lexer, codeGenerator);
Expression exp = parser.parse();
if (getOptionValue(Options.TRACE_EVAL).bool) {
((BaseExpression) exp).setExpression(expression);
}
return exp;
}这里说明几个主要用到的类：
ExpressionLexer：表达式词法分析器，用来对aviator脚本进行词法解析，如将脚本解析为变量、数字、字符串、注释等，并构造Token流进行后续处理；
CodeGenerator：字节码生成器，用于动态生成自定义的字节码；
ExpressionParser：表达式解析器，用于将脚本编译为表达式对象（ClassExpression），支持对多种aviator支持的脚本进行解析，具体解析过程是将ExpressionLexer构造的Token流通过字节码生成器编译为表达式对象ClassExpression
2.1.1 CodeGenerator
这里用到的CodeGenerator主要包括：
CodeGenerator：定义了字节码生成的顶层接口
ASMCodeGenerator：默认的ASM字节码生成器，对CodeGenerator声明的操作进行ASM具体实现
OptimizeCodeGenerator：为加快执行效率而生的字节码生成器（默认的字节码生成器），可以执行一些前置的计算逻辑，然后再委托ASMCodeGenerator进行字节码生成
2.1.2 ExpressionParser
ExpressionParser完成对多种aviator支持的脚本的解析工作，比如if，for，let等脚本特性，解析逻辑是依赖parse方法完成的，如下：
public Expression parse(final boolean reportErrorIfNotEOF) {
StatementType statementType = statements();
if (this.lookhead != null && reportErrorIfNotEOF) {
if (statementType == StatementType.Ternary) {
reportSyntaxError("unexpect token '" + currentTokenLexeme()
+ "', maybe forget to insert ';' to complete last expression ");
} else {
reportSyntaxError("unexpect token '" + currentTokenLexeme() + "'");
}
}
return getCodeGeneratorWithTimes().getResult(true);
}这里主要分为2个部分：
1）调用statements完成脚本解析工作，产出Token流（以默认的OptimizeCodeGenerator进行分析）
2）调用OptimizeCodeGenerator的getResult方法完成ASM字节码生成，并根据字节码构造完成ClassExpression实例对象
2.1.1.1 statements方法
statements中主要的解析过程如下：
private StatementType statements() {
if (this.lookhead == null) {
return StatementType.Empty;
}
StatementType stmtType = statement();
ensureDepthState();
while (expectChar(';') || stmtType == StatementType.Other || stmtType == StatementType.Return) {
ensureNoStatementAfterReturn(stmtType);
if (this.lookhead != null && this.lookhead != Variable.END && !expectChar('}')) {
getCodeGeneratorWithTimes().onTernaryEnd(this.lookhead);
}
if (expectChar(';')) {
move(true);
}
if (this.lookhead == null) {
break;
}
StatementType nextStmtType = statement();
if (nextStmtType == StatementType.Empty) {
break;
}
stmtType = nextStmtType;
ensureDepthState();
}
ensureNoStatementAfterReturn(stmtType);
// If the last statement is ternary,it must be ended with END TOKEN such as null token, '}',
// 'end' keyword, or ';'
// Otherwise report syntax error.
if (stmtType == StatementType.Ternary) {
if (lookhead != null && !expectChar(';') && !expectChar('}') && lookhead != Variable.END) {
this.back();
reportSyntaxError("unexpect token '" + currentTokenLexeme()
+ "', maybe forget to insert ';' to complete last expression ");
}
}
return stmtType;
}具体的解析过程是在statement方法中，如下：
private StatementType statement() {
if (this.lookhead == Variable.IF) {
ensureFeatureEnabled(Feature.If);
if (ifStatement(false, false)) {
return StatementType.Return;
} else {
return StatementType.Other;
}
} else if (this.lookhead == Variable.FOR) {
ensureFeatureEnabled(Feature.ForLoop);
forStatement();
return StatementType.Other;
} else if (this.lookhead == Variable.RETURN) {
ensureFeatureEnabled(Feature.Return);
returnStatement();
return StatementType.Return;
} else if (this.lookhead == Variable.BREAK) {
breakStatement();
return StatementType.Return;
} else if (this.lookhead == Variable.CONTINUE) {
continueStatement();
return StatementType.Return;
} else if (this.lookhead == Variable.LET) {
ensureFeatureEnabled(Feature.Let);
letStatement();
return StatementType.Other;
} else if (this.lookhead == Variable.WHILE) {
ensureFeatureEnabled(Feature.WhileLoop);
whileStatement();
return StatementType.Other;
} else if (this.lookhead == Variable.FN) {
ensureFeatureEnabled(Feature.Fn);
fnStatement();
return StatementType.Other;
} else if (this.lookhead == Variable.TRY) {
ensureFeatureEnabled(Feature.ExceptionHandle);
tryStatement();
return StatementType.Other;
} else if (this.lookhead == Variable.THROW) {
ensureFeatureEnabled(Feature.ExceptionHandle);
throwStatement();
return StatementType.Other;
} else if (expectChar('{')) {
ensureFeatureEnabled(Feature.LexicalScope);
if (scopeStatement()) {
return StatementType.Return;
} else {
return StatementType.Other;
}
} else if (this.lookhead == Variable.USE) {
ensureFeatureEnabled(Feature.Use);
useStatement();
return StatementType.Other;
} else {
if (ternary()) {
return StatementType.Ternary;
} else {
return StatementType.Empty;
}
}
}这里可以看到：aviator支持的多种脚本特性，比如if、for、let、return等语法都是在这里解析完成的；
以if语法举例说明，继续跟踪ifStatement方法，如下：
/**
*
* if(test) {
* ...if-body...
* }else {
* ...else-body...
* }
* ...statements...
*
*
* ===>
*
*
* __if_callcc(test ? (lambda() -> ...if-body... end)() : (lambda() -> ...else-body... end)(),
* lambda()- >
* ...statements...
* end);
*
*/
private boolean ifStatement(final boolean isWhile, final boolean isElsif) {
if (!isWhile) {
move(true);
}
boolean ifBodyHasReturn = false;
boolean elseBodyHasReturn = false;
boolean newLexicalScope = this.scope.newLexicalScope;
this.scope.newLexicalScope = true;
// prepare to call __if_callcc(result, statements)
getCodeGeneratorWithTimes().onMethodName(Constants.IfReturnFn);
{
if (!ternary()) {
reportSyntaxError("missing test statement for if");
}
getCodeGeneratorWithTimes().onTernaryBoolean(this.lookhead);
if (expectChar('{')) {
move(true);
this.scope.enterBrace();
getCodeGeneratorWithTimes().onLambdaDefineStart(
getPrevToken().withMeta(Constants.SCOPE_META, this.scope.newLexicalScope));
getCodeGeneratorWithTimes().onLambdaBodyStart(this.lookhead);
ifBodyHasReturn = statements() == StatementType.Return;
getCodeGeneratorWithTimes().onLambdaBodyEnd(this.lookhead);
getCodeGeneratorWithTimes().onMethodName(anonymousMethodName());
getCodeGeneratorWithTimes().onMethodInvoke(this.lookhead);
getCodeGeneratorWithTimes().onTernaryLeft(this.lookhead);
} else {
reportSyntaxError("expect '{' for " + getLoopKeyword(isWhile) + " statement");
}
if (!expectChar('}')) {
reportSyntaxError("missing '}' to close " + getLoopKeyword(isWhile) + " body");
}
this.scope.leaveBrace();
move(true);
elseBodyHasReturn = elseStatement(isWhile, ifBodyHasReturn);
getCodeGeneratorWithTimes().onMethodParameter(this.lookhead);
}
{
//
if (isWhile || isElsif) {
// Load ReducerEmptyVal directly.
getCodeGenerator().onConstant(Constants.ReducerEmptyVal);
} else {
if (expectChar(';')) {
// the statement is ended.
getCodeGenerator().onConstant(Constants.ReducerEmptyVal);
} else {
// create a lambda function wraps statements after if statement (statements)
getCodeGeneratorWithTimes().onLambdaDefineStart(
getPrevToken().withMeta(Constants.SCOPE_META, this.scope.newLexicalScope) //
.withMeta(Constants.INHERIT_ENV_META, true));
getCodeGeneratorWithTimes().onLambdaBodyStart(this.lookhead);
if (statements() == StatementType.Empty) {
getCodeGenerator().onConstant(Constants.ReducerEmptyVal);
} else {
if (ifBodyHasReturn && elseBodyHasReturn && !isElsif) {
reportSyntaxError("unreachable code");
}
}
getCodeGeneratorWithTimes().onLambdaBodyEnd(this.lookhead);
}
}
getCodeGenerator().onMethodParameter(this.lookhead);
// call __if_callcc(result, statements)
getCodeGenerator().onMethodInvoke(this.lookhead);
this.scope.newLexicalScope = newLexicalScope;
}
return ifBodyHasReturn && elseBodyHasReturn;
}
if语法的解析过程主要分为以下几个部分：
1）通过ternary方法完成对if条件语句的解析（解析结果放入OptimizeCodeGenerator的Token流中，后续统一生成字节码）
2）if的方法体（ifBody）抽象为一个lambda表达式，并通过委托给lambdaGenerator进行解析，后面着重分析该解析过程；
3）对elseBody进行解析：这里实际的解析过程和ifBody的解析过程类似，也是委托给新构建的lambdaGenerator进行解析；
private boolean elseStatement(final boolean isWhile, final boolean ifBodyHasReturn) {
if (isWhile) {
// Call __reducer_break(nil)
final CodeGenerator cg = getCodeGeneratorWithTimes();
cg.onMethodName(Constants.ReducerBreakFn);
cg.onConstant(Variable.NIL);
cg.onMethodParameter(this.lookhead);
cg.onMethodInvoke(this.lookhead);
cg.onTernaryRight(this.lookhead);
return false;
}
if (expectChar(';')) {
return withoutElse();
}
boolean hasReturn = false;
boolean hasElsif = this.lookhead == Variable.ELSIF;
boolean hasElse = this.lookhead == Variable.ELSE;
if (this.lookhead != null && (hasElse || hasElsif || ifBodyHasReturn)) {
if (hasElse) {
move(true);
if (expectChar('{')) {
this.scope.enterBrace();
move(true);
hasReturn = elseBody(false);
if (expectChar('}')) {
this.scope.leaveBrace();
move(true);
} else {
reportSyntaxError("missing '}' to close 'else' body");
}
} else {
reportSyntaxError("expect '{' for else statement");
}
} else if (hasElsif) {
hasReturn = ifStatement(false, true);
getCodeGenerator().onTernaryRight(this.lookhead);
} else if (ifBodyHasReturn) {
hasReturn = elseBody(true);
} else {
return withoutElse();
}
return hasReturn;
} else {
// Missing else statement, always nil.
return withoutElse();
}
}4）这里对if语句后面的整个脚本抽象为一个lambda表达式，也是通过委托给lambdaGenerator进行解析；
如上的解析过程也可以借助方法注释进行较好的理解：
2.1.1.1.1 lambdaGenerator解析过程分析
对2.1.1.1 节的注释2的代码进一步分析：
1）lambda脚本解析前置环节
onLambdaDefineStart方法构造lambdaGenerator；
onLambdaBodyStart方法将ExpressionParser的codeGenerator替换为新构造的lambdaGenerator，同时维护好lambdaGenerator的父parentCodeGenerator=OptimizeCodeGenerator，lamdba解析完成后再替换回parentCodeGenerator，也即OptimizeCodeGenerator，也及整个脚本的ASM操作是通过OptimizeCodeGenerator和lambdaGenerator（如果包含lambda语法）交替完成的
2）lambda脚本解析完成后
lambda脚本解析完成后，这里会调用getLmabdaBootstrap构造LambdaFunctionBootstrap实例对象，并将LambdaFunctionBootstrap缓存到OptimizeCodeGenerator的
Map
进一步展开getLmabdaBootstrap函数：
可以看到这里也是通过getResult方法通过ASM字节码技术动态构造Expression子类实例（具体分析过程见2.1.1.2），构造的结果放到了LambdaFunctionBootstrap实例对象中，也即lambda表达式的解析结果即为LambdaFunctionBootstrap；
2.1.1.2 OptimizeCodeGenerator.getResult()过程分析
aviator脚本解析完成之后，解析结果Token流会存放到OptimizeCodeGenerator的成员变量List
getResult方法具体实现如下：
@Override
public Expression getResult(final boolean unboxObject) {
// execute literal expression
while (execute() > 0) {
;
}
Map();
Map();
Set> constants = new HashSet<>();
for (Token token : this.tokenList) {
if (ExpressionParser.isConstant(token, this.instance)) {
constants.add(token);
}
switch (token.getType()) {
case Variable:
if (SymbolTable.isReservedKeyword((Variable) token)) {
continue;
}
String varName = token.getLexeme();
VariableMeta meta = variables.get(varName);
if (meta == null) {
meta = new VariableMeta((CompileTypes) token.getMeta(Constants.TYPE_META), varName,
token.getMeta(Constants.INIT_META, false), token.getStartIndex());
variables.put(varName, meta);
} else {
meta.add(token);
}
break;
case Delegate:
DelegateToken delegateToken = (DelegateToken) token;
if (delegateToken.getDelegateTokenType() == DelegateTokenType.Method_Name) {
Token realToken = delegateToken.getToken();
if (realToken == null) {
continue;
}
if (realToken.getType() == TokenType.Variable) {
String methodName = token.getLexeme();
if (!methods.containsKey(methodName)) {
methods.put(methodName, 1);
} else {
methods.put(methodName, methods.get(methodName) + 1);
}
}
} else if (delegateToken.getDelegateTokenType() == DelegateTokenType.Array) {
Token realToken = delegateToken.getToken();
if (realToken.getType() == TokenType.Variable) {
varName = token.getLexeme();
VariableMeta varMeta = variables.get(varName);
if (varMeta == null) {
varMeta =
new VariableMeta((CompileTypes) realToken.getMeta(Constants.TYPE_META), varName,
realToken.getMeta(Constants.INIT_META, false), realToken.getStartIndex());
variables.put(varName, varMeta);
} else {
varMeta.add(realToken);
}
}
}
break;
}
}
Expression exp = null;
// Last token is a literal token,then return a LiteralExpression
if (this.tokenList.size() <= 1) {
if (this.tokenList.isEmpty()) {
exp = new LiteralExpression(this.instance, null, new ArrayList<>(variables.values()));
} else {
final Token lastToken = this.tokenList.get(0);
if (ExpressionParser.isLiteralToken(lastToken, this.instance)) {
exp = new LiteralExpression(this.instance,
getAviatorObjectFromToken(lastToken).getValue(getCompileEnv()),
new ArrayList<>(variables.values()));
}
}
}
if (exp == null) {
// call asm to generate byte codes
callASM(variables, methods, constants);
// get result from asm
exp = this.codeGen.getResult(unboxObject);
}
if (exp instanceof BaseExpression) {
((BaseExpression) exp).setCompileEnv(getCompileEnv());
((BaseExpression) exp).setSourceFile(this.sourceFile);
}
return exp;
} 这里主要包含以下几部分：
1）可以前置执行的逻辑提前执行，比如文本表达式（1+2）等，先行计算出执行结果，优化执行效率；
2）初始化常量集、变量集、aviator函数实例集合，为后续ASM生成类成员变量和类构造函数使用；
3）调用callASM方法生成字节码，根据不同的token类型进行不同的asm操作；
private void callASM(final Map> constants) {
this.codeGen.initConstants(constants);
this.codeGen.initVariables(variables);
this.codeGen.initMethods(methods);
this.codeGen.setLambdaBootstraps(this.lambdaBootstraps);
this.codeGen.start();
for (int i = 0; i < this.tokenList.size(); i++) {
Token token = this.tokenList.get(i);
switch (token.getType()) {
case Operator:
OperatorToken op = (OperatorToken) token;
switch (op.getOperatorType()) {
case ADD:
this.codeGen.onAdd(token);
break;
case SUB:
this.codeGen.onSub(token);
break;
case MULT:
this.codeGen.onMult(token);
break;
case Exponent:
this.codeGen.onExponent(token);
break;
case DIV:
this.codeGen.onDiv(token);
break;
case MOD:
this.codeGen.onMod(token);
break;
case EQ:
this.codeGen.onEq(token);
break;
case NEQ:
this.codeGen.onNeq(token);
break;
case LT:
this.codeGen.onLt(token);
break;
case LE:
this.codeGen.onLe(token);
break;
case GT:
this.codeGen.onGt(token);
break;
case GE:
this.codeGen.onGe(token);
break;
case NOT:
this.codeGen.onNot(token);
break;
case NEG:
this.codeGen.onNeg(token);
break;
case AND:
this.codeGen.onAndRight(token);
break;
case OR:
this.codeGen.onJoinRight(token);
break;
case FUNC:
this.codeGen.onMethodInvoke(token);
break;
case INDEX:
this.codeGen.onArrayIndexEnd(token);
break;
case MATCH:
this.codeGen.onMatch(token);
break;
case TERNARY:
this.codeGen.onTernaryRight(token);
break;
case BIT_AND:
this.codeGen.onBitAnd(token);
break;
case BIT_OR:
this.codeGen.onBitOr(token);
break;
case BIT_XOR:
this.codeGen.onBitXor(token);
break;
case BIT_NOT:
this.codeGen.onBitNot(token);
break;
case SHIFT_LEFT:
this.codeGen.onShiftLeft(token);
break;
case SHIFT_RIGHT:
this.codeGen.onShiftRight(token);
break;
case DEFINE:
this.codeGen.onAssignment(token.withMeta(Constants.DEFINE_META, true));
break;
case ASSIGNMENT:
this.codeGen.onAssignment(token);
break;
case U_SHIFT_RIGHT:
this.codeGen.onUnsignedShiftRight(token);
break;
}
break;
case Delegate:
DelegateToken delegateToken = (DelegateToken) token;
final Token realToken = delegateToken.getToken();
switch (delegateToken.getDelegateTokenType()) {
case And_Left:
this.codeGen.onAndLeft(realToken);
break;
case Join_Left:
this.codeGen.onJoinLeft(realToken);
break;
case Array:
this.codeGen.onArray(realToken);
break;
case Index_Start:
this.codeGen.onArrayIndexStart(realToken);
break;
case Ternary_Boolean:
this.codeGen.onTernaryBoolean(realToken);
break;
case Ternary_Left:
this.codeGen.onTernaryLeft(realToken);
break;
case Method_Name:
this.codeGen.onMethodName(realToken);
break;
case Method_Param:
this.codeGen.onMethodParameter(realToken);
break;
case Lambda_New:
this.codeGen.genNewLambdaCode(delegateToken.getLambdaFunctionBootstrap());
break;
case Ternay_End:
this.codeGen.onTernaryEnd(realToken);
break;
}
break;
default:
this.codeGen.onConstant(token);
break;
}
}
} 4）调用getResult方法根据生成的字节码构造Expression子类实例（ClassExpression）
@Override
public Expression getResult(final boolean unboxObject) {
end(unboxObject);
byte[] bytes = this.classWriter.toByteArray();
try {
Class defineClass =
ClassDefiner.defineClass(this.className, Expression.class, bytes, this.classLoader);
Constructor constructor =
defineClass.getConstructor(AviatorEvaluatorInstance.class, List.class, SymbolTable.class);
BaseExpression exp = (BaseExpression) constructor.newInstance(this.instance,
new ArrayList(this.variables.values()), this.symbolTable);
exp.setLambdaBootstraps(this.lambdaBootstraps);
exp.setFuncsArgs(this.funcsArgs);
exp.setSourceFile(this.sourceFile);
return exp;
} catch (ExpressionRuntimeException e) {
throw e;
} catch (Throwable e) {
if (e.getCause() instanceof ExpressionRuntimeException) {
throw (ExpressionRuntimeException) e.getCause();
}
throw new CompileExpressionErrorException("define class error", e);
}
} 2.1.3 if脚本ASM反编译结果
if脚本的asm的解析过程已分析完成，结合上面的if具体脚本示例和源码中asm字节码生成过程，可以得到上述的if脚本示例动态生成的类如下：
1)ifBody生成的lassLambda_1684208818128_57）
public super class Script_11313134242424_59 extends ClassExpression {
private final AviatorObject f0=null;
private final AviatorJavaType f1=null;
private final AviatorFunction f2=null;
public void (final AviatorEvaluatorInstance instance, final List vars,
final SymbolTable symbolTable){
super(instance,vars,symbolTable);
f0=AviatorLong.valueOf(2);
f1=new AviatorJavaType("b", symbolTable);
f2=instance.getFunction("__reducer_return",symbolTable);
}
public final Object execute0(Env env){
return RuntimeUtils.assertNotNull(f2.call(env,f1.add(f0, env))).deref(env);
}
} 2）elseBody生成的classambda_1684208818128_58）
public super class Script_11313134242424_60 extends ClassExpression {
private final AviatorJavaType f0=null;
private final AviatorFunction f1=null;
public void (final AviatorEvaluatorInstance instance, final List vars,
final SymbolTable symbolTable){
super(instance,vars,symbolTable);
f0=new AviatorJavaType("c", symbolTable);
f1=instance.getFunction("__reducer_return",symbolTable);
}
public final Object execute0(Env env){
return RuntimeUtils.assertNotNull(f1.call(env,f0)).deref(env);
}
}
3）if语句后面的语句生成的classambda_1684208818128_59）
public super class Script_11313134242424_61 extends ClassExpression {
private final AviatorJavaType f0=null;
public void (final AviatorEvaluatorInstance instance, final List vars,
final SymbolTable symbolTable){
super(instance,vars,symbolTable);
f0=new AviatorJavaType("_reducer_empty", symbolTable);
}
public final Object execute0(Env env){
return f0.deref(env); // return null;
}
} 4）整个if脚本生成的class
public super class Script_11313134242424_58 extends ClassExpression {
private final AviatorObject f0=null;
private final AviatorJavaType f1=null;
private final AviatorFunction f2=null;
public void (final AviatorEvaluatorInstance instance, final List vars,
final SymbolTable symbolTable){
super(instance,vars,symbolTable);
f0=AviatorLong.valueOf(1);
f1=new AviatorJavaType("a", symbolTable);
f2=AviatorEvaluatorInstance.getFunction("__if_callcc",symbolTable);
}
public final Object execute0(Env env){
RuntimeUtils.assertNotNull(f2.call(env,
if(f1.compare(f0, env).booleanValue(env)){
RuntimeUtils.assertNotNull(RuntimeUtils.getFunction(this.newLambda(env, "Lambda_1684208818128_57"), env).call(env));
}else{
RuntimeUtils.assertNotNull(RuntimeUtils.getFunction(this.newLambda(env, "Lambda_1684208818128_58"), env).call(env));
},
this.newLambda(env, "Lambda_1684208818128_59")
)).getValue(env);
}
} 2.2 compiledExpression.execute执行过程
上述根据if脚本通过asm字节码最终生成ClassExpression类后，下面即传出变量集env实参进行执行，execute方法如下：
@Override
public Object execute(Map map) {
if (map == null) {
map = Collections.emptyMap();
}
Env env = genTopEnv(map);
EnvProcessor envProcessor = this.instance.getEnvProcessor();
if (envProcessor != null) {
envProcessor.beforeExecute(env, this);
}
try {
return executeDirectly(env);
} finally {
if (envProcessor != null) {
envProcessor.afterExecute(env, this);
}
}
} 这里包含了EnvProcessor前置拦截和后置拦截器，下面主要分析下executeDirectly方法具体执行过程：
@Override
public Object executeDirectly(final Map env) {
try {
Object result = execute0((Env) env);
if (RuntimeUtils.isTracedEval(env)) {
RuntimeUtils.printlnTrace(env, "Result : " + result);
}
return result;
} catch (ExpressionRuntimeException e) {
throw e;
} catch (Throwable t) {
throw Reflector.sneakyThrow(t);
}
} ClassExpression的方法executeDirectly中又调用了execute0进行执行，产出结果，这里的execute0即为上面asm字节码生成部分通过asm生成的成员方法，
针对上述的if脚本示例，生成的ClassExpression子类和实现的execute0方法如下：
public super class Script_11313134242424_58 extends ClassExpression {
private final AviatorObject f0=null;
private final AviatorJavaType f1=null;
private final AviatorFunction f2=null;
public void (final AviatorEvaluatorInstance instance, final List vars,
final SymbolTable symbolTable){
super(instance,vars,symbolTable);
f0=AviatorLong.valueOf(1);
f1=new AviatorJavaType("a", symbolTable);
f2=AviatorEvaluatorInstance.getFunction("__if_callcc",symbolTable);
}
public final Object execute0(Env env){
RuntimeUtils.assertNotNull(f2.call(env,
if(f1.compare(f0, env).booleanValue(env)){
RuntimeUtils.assertNotNull(RuntimeUtils.getFunction(this.newLambda(env, "Lambda_1684208818128_57"), env).call(env));
}else{
RuntimeUtils.assertNotNull(RuntimeUtils.getFunction(this.newLambda(env, "Lambda_1684208818128_58"), env).call(env));
},
this.newLambda(env, "Lambda_1684208818128_59")
)).getValue(env);
}
} 这样，execute0通过传入env实参，执行的方法体即完整实现了if示例脚本的内容，最终产出if脚本计算结果；
其中在执行到某个具体if分支时，会调用newLambda函数：
public LambdaFunction newLambda(final Env env, final String name) {
LambdaFunctionBootstrap bootstrap = this.lambdaBootstraps.get(name);
if (bootstrap == null) {
throw new ExpressionNotFoundException("Lambda " + name + " not found");
}
return bootstrap.newInstance(env);
}newLambda函数中会调用缓存的lambdaBootstraps，获取对应的LambdaFunctionBootstrap，然后通过newInstance方法创建对应的LambdaFunction，如下：
/**
* Create a lambda function.
*
* @param env
* @return
*/
public LambdaFunction newInstance(final Env env) {
Reference ref = null;
if (this.inheritEnv && (ref = this.fnLocal.get()) != null) {
LambdaFunction fn = ref.get();
if (fn != null) {
fn.setContext(env);
return fn;
} else {
this.fnLocal.remove();
}
}
LambdaFunction fn = new LambdaFunction(this.name, this.params, this.expression, env);
fn.setInheritEnv(this.inheritEnv);
if (this.inheritEnv) {
this.fnLocal.set(new SoftReference<>(fn));
}
return fn;
} 后面继续调用LambdaFunction的call函数：
@Override
public AviatorObject call(final Map env) {
try {
if (this.isVariadic && !this.installed) {
return variadicCall(env, true);
}
return AviatorRuntimeJavaType.valueOf(this.expression.executeDirectly(newEnv(env)));
} finally {
if (this.inheritEnv) {
this.context = null;
}
}
} call方法体中，又进一步调用了lambda脚本通过asm生成的expression，进而执行了对应分支的逻辑，至此，最终产出计算结果；
3.总结概述
1）LambdaFunctionBootstrap代表的含义
LambdaFunctionBootstrap是对if语法中ifBody，elseBody，elsifBody以及if语句之后的语句构造的模型（也包括其他的while、for、lambda语法等），是对lambda类型脚本通过asm字节码技术编译后的抽象，包含了lambda类型脚本编译后的Expression
2）LRU缓存中缓存的内容
LRU缓存的key为执行脚本本身，String类型
LRU缓存的value为Expression对象，对于ClassExpression子类对象，内部包含了解析的多个LambdaFunctionBootstrap实例，对于上述if脚本示例，即包含3个LambdaFunctionBootstrap实例
3）LambdaFunction代表的含义
在if脚本示例编译后的ClassExpression子类实例中，方法execute0中调用了newLambda方法，在传入参数env后的执行过程中，通过这里的newLambda会创建LambdaFunction对象，LambdaFunction是对lambda类型脚本在执行过程中的抽象，包含了lambda类型脚本编译后的Expression；
在LambdaFunctionBootstrap中实现了对于LambdaFunction（非线程安全）的ThreadLocal线程本地缓存，提交执行效率；
附，LambdaFunction函数作为线程本地缓存，aviator低版本（version<5.3.3）存在内存泄漏问题