我的想法：
- 在编译脚本时，您必须使用自己的类加载器，以避免从脚本中访问任何其他类。不确定在 GAE 中是否可行。
- 您必须使用 Java 的 SecurityManager 功能来避免脚本能够访问文件系统、网络等。不确定在 GAE 中是否可行。
只看上面的两个项目，对我来说它看起来非常复杂和脆弱。如果你找不到现有的沙盒功能作为现有项目，你应该远离它。
设计一种允许您认为合法的表达式的领域特定语言要安全得多，并且查看上述项目，无论如何您都必须非常努力地思考您想要允许的内容。从那里到设计语言并不是一个很大的步骤。
注意不要使用 groovy 闭包（内部 DSL）来实现 DSL，因为那只是 groovy，而且你也可以破解。您需要定义一种外部语言并对其进行解析。我推荐使用解析器组合器 jparsec 来定义语法。在这种情况下不需要编译器编译器。
http://jparsec.codehaus.org/
仅供参考，这是我用 jparsec 编写的一个小解析器（常规代码）：
//import some static methods, this will allow more concise code
import static org.codehaus.jparsec.Parsers.*
import static org.codehaus.jparsec.Terminals.*
import static org.codehaus.jparsec.Scanners.*
import org.codehaus.jparsec.functors.Map as FMap
import org.codehaus.jparsec.functors.Map4 as FMap4
import org.codehaus.jparsec.functors.Map3 as FMap3
import org.codehaus.jparsec.functors.Map2 as FMap2
/**
* Uses jparsec combinator parser library to construct an external DSL parser for the following grammar:
* <pre>
* pipeline := routingStep*
* routingStep := IDENTIFIER '(' parameters? ')'
* parameters := parameter (',' parameter)*
* parameter := (IDENTIFIER | QUOTED_STRING) ':' QUOTED_STRING
* </pre>
*/
class PipelineParser {
//=======================================================
//Pass 1: Define which terminals are part of the grammar
//=======================================================
//operators
private static def OPERATORS = operators(',', '(', ')', ':')
private static def LPAREN = OPERATORS.token('(')
private static def RPAREN = OPERATORS.token(')')
private static def COLON = OPERATORS.token(':')
private static def COMMA = OPERATORS.token(',')
//identifiers tokenizer
private static def IDENTIFIER = Identifier.TOKENIZER
//single quoted strings tokenizer
private static def SINGLE_QUOTED_STRING = StringLiteral.SINGLE_QUOTE_TOKENIZER
//=======================================================
//Pass 2: Define the syntax of the grammar
//=======================================================
//PRODUCTION RULE: parameter := (IDENTIFIER | QUOTED_STRING) ':' QUOTED_STRING
@SuppressWarnings("GroovyAssignabilityCheck")
private static def parameter = sequence(or(Identifier.PARSER,StringLiteral.PARSER), COLON, StringLiteral.PARSER, new FMap3() {
def map(paramName, colon, paramValue) {
new Parameter(name: paramName, value: paramValue)
}
})
//PRODUCTION RULE: parameters := parameter (',' parameter)*
@SuppressWarnings("GroovyAssignabilityCheck")
private static def parameters = sequence(parameter, sequence(COMMA, parameter).many(), new FMap2() {
def map(parameter1, otherParameters) {
if (otherParameters != null) {
[parameter1, otherParameters].flatten()
} else {
[parameter1]
}
}
})
//PRODUCTION RULE: routingStep := IDENTIFIER '(' parameters? ')'
@SuppressWarnings("GroovyAssignabilityCheck")
private static def routingStep = sequence(Identifier.PARSER, LPAREN, parameters.optional(), RPAREN, new FMap4() {
def map(routingStepName, lParen, parameters, rParen) {
new RoutingStep(
name: routingStepName,
parameters: parameters ?: []
)
}
})
//PRODUCTION RULE: pipeline := routingStep*
@SuppressWarnings("GroovyAssignabilityCheck")
private static def pipeline = routingStep.many().map(new FMap() {
def map(from) {
new Pipeline(
routingSteps: from
)
}
})
//Combine the above tokenizers to create the tokenizer that will parse the stream and spit out the tokens of the grammar
private static def tokenizer = or(OPERATORS.tokenizer(), SINGLE_QUOTED_STRING, IDENTIFIER)
//This parser will be used to define which input sequences need to be ignored
private static def ignored = or(JAVA_LINE_COMMENT, JAVA_BLOCK_COMMENT, WHITESPACES)
/**
* Parser that is used to parse extender pipelines.
* <pre>
* def parser=PipelineParser.parser
* Pipeline pipeline=parser.parse(pipelineStr)
* </pre>
* Returns an instance of {@link Pipeline} containing the AST representation of the parsed string.
*/
//Create a syntactic pipeline parser that will use the given tokenizer to parse the input into tokens, and will ignore sequences that are matched by the given parser.
static def parser = pipeline.from(tokenizer, ignored.skipMany())
}