Skip to content

dingodb/dingo-expr

Repository files navigation

Dingo Expression

Build

Dingo Expression is an expression parsing and evaluating library, written in Java and C++.

Getting Started

// The original expression string.
String exprString = "(1 + 2) * (5 - (3 + 4))";
// parse it into an Expr object.
Expr expr = ExprParser.DEFAULT.parse(exprString);
// Compile the Expr to get a compiled Expr object.
Expr expr1 = ExprCompiler.SIMPLE.visit(expr);
// Evaluate it
Object result = expr1.eval();

Expr object can also be constructed by code, not only by parsing expression strings

Expr expr = Exprs.op(Exprs.MUL, Exprs.op(Exprs.ADD, 1, 2), Exprs.op(Exprs.SUB, 5, Exprs.op(Exprs.ADD, 3, 4)));

There are also an "advanced" compiler, which can simplify expressions when compiling

Expr expr1 = ExprCompiler.ADVANCED.visit(expr);

Variables

Variables are allowed in expressions. In order to compile an expression with variables, an implementation of interface CompileContext must be provided, which tells the compiler the type and runtime identification of all the existing variables

Expr expr1 = ExprCompiler.SIMPLE.visit(expr, compileContext);

To evaluate the expression, an implementation of interface EvalContext must be provided to get the real value of variables

Object result = expr1.eval(evalContext, config);

where config is an object implementing ExprConfig which can be ExprConfig.SIMPLE, ExprConfig.ADVANCED or any customized implementations.

Functions

Functions are all pre-defined and there is no syntax to define a function in expressions, but you can implement a customized function and register it to the parsing, compiling and evaluating system.

All this begins with an interface FunFactory

public interface FunFactory {
    // Register a function with no parameters
    void registerNullaryFun(@NonNull String funName, @NonNull NullaryOp op);

    // Register a function with one parameter
    void registerUnaryFun(@NonNull String funName, @NonNull UnaryOp op);

    // Register a function with two parameters
    void registerBinaryFun(@NonNull String funName, @NonNull BinaryOp op);

    // Register a function with three parameters
    void registerTertiaryFun(@NonNull String funName, @NonNull TertiaryOp op);

    // Register a function with varidic parameters
    void registerVariadicFun(@NonNull String funName, @NonNull VariadicOp op);
}

An implementation of FunFactory is then used in a ExprParser

ExprParser parser = new ExprParser(new DefaultFunFactory(ExprConfig.SIMPLE));

where the DefaultFunFactory is a provided implementation of FunFactory which contains the pre-defined functions. Actually there is a ExprParser.DEFAULT defined as above, which can be used. In this case, you can register a function by ExprParser.DEFAULT.getFunFactory().registerUnaryFun(funName, UnaryOp op).

Note, in DefaultFunFactory, the five kinds of functions are stored in separate maps, so different kinds of functions with the same name is allowed.

After register the function, the ExprParser can recognize the new function and convert to the corresponding Op. Then, the only thing left is to implement an Op.

For example, if you want to register an UnaryOp, which has only one parameter, by calling registerUnaryFun, you should extend the class UnaryOp

public class CustomizedOp extends UnaryOp {
    private static final long serialVersionUID = 12345678L;

    // Implement this if `NULL` is returned when the parameter is `NULL`.
    @Override
    protected Object evalNonNullValue(@NonNull Object value, ExprConfig config);

    // Implement this if `NULL` parameter need to be processed.
    @Override
    public Object evalValue(Object value, ExprConfig config);

    // Implement this if the parameter (before evaluating, is an `Expr`) need to be processed.
    @Override
    public Object eval(@NonNull Expr expr, EvalContext context, ExprConfig config);

    @Override
    public boolean isConst(@NonNull UnaryOpExpr expr);

    @Override
    public OpKey keyOf(@NonNull Type type);

    @Override
    public OpKey bestKeyOf(@NonNull Type @NonNull [] types);

    @Override
    public UnaryOp getOp(OpKey key);

Only one of the methods evalNonNullValue, evalValue and eval is required to be implemented, which do the evaluating.

The method isConst is to predicate if the expression is a const value. If true, the Op may be replaced by its evaluated value in compiling time. The default implementation is to see if all the operands of the Expr is true. If the Op ought not to be replaced, this method should be overridden and return false.

The methods keyOf, bestKeyOf and getOp are used to support multiple signatures of the function and are called in compiling time. The compiler call these methods as the following steps

  1. call keyOf to get the OpKey for the given types of parameters
  2. call getOp to get the compiled Op for the given OpKey (indirectly, for the given types of parameters)
  3. if getOp returns null, which means the given types of parameters is not supported, then try to call bestKeyOf to decide if type conversions are available

In the implementation of bestKeyOf, the required types of parameters should be set into the array, then the corresponding OpKey is returned. The compiler will insert casting Op to fulfill the type requirement.

To simplify the implementation of customized function further, the annotation @Operators can be used. For example

@Operators
abstract class AddOp extends BinaryOp {
    private static final long serialVersionUID = 7159909541314089027L;

    static int add(int value0, int value1) {
        return value0 + value1;
    }

    static long add(long value0, long value1) {
        return value0 + value1;
    }

    static float add(float value0, float value1) {
        return value0 + value1;
    }

    static double add(double value0, double value1) {
        return value0 + value1;
    }

    @Override
    public @NonNull String getName() {
        return "ADD";
    }

    @Override
    public @Nullable OpKey keyOf(@NonNull Type type0, @NonNull Type type1) {
        return OpKeys.DEFAULT.keyOf(type0, type1);
    }

    @Override
    public final OpKey bestKeyOf(@NonNull Type @NonNull [] types) {
        return OpKeys.DEFAULT.bestKeyOf(types);
    }
}

where only the evaluating method of different types are required. A final class which extends this class will be generated, which can wrap these methods into Ops.

Dependencies

For Maven POM

<dependencies>
    <!-- Required if you want to do expression parsing -->
    <dependency>
        <groupId>io.dingodb.expr</groupId>
        <artifactId>dingo-expr-parser</artifactId>
        <version>0.7.0</version>
    </dependency>

    <!-- Core module to compile and evaluate an expression -->
    <dependency>
        <groupId>io.dingodb.expr</groupId>
        <artifactId>dingo-expr-runtime</artifactId>
        <version>0.7.0</version>
    </dependency>

    <!-- Required if annotations are used -->
    <dependency>
        <groupId>io.dingodb.expr</groupId>
        <artifactId>dingo-expr-annotations</artifactId>
        <version>0.7.0</version>
    </dependency>
</dependencies>

For Gradle build

dependencies {
    // Required if you want to do expression parsing
    implementation group: 'io.dingodb.expr', name: 'dingo-expr-runtime', version: '0.7.0'
    // Core module to compile and evaluate an expression
    implementation group: 'io.dingodb.expr', name: 'dingo-expr-parser', version: '0.7.0'
    // Required if annotations are used
    implementation group: 'io.dingodb.expr', name: 'dingo-expr-annotations', version: '0.7.0'
}

Used By

About

Expression library used by DingoDB.

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •  

Languages