CSC/ECE 517 Summer 2008/wiki1 7 ev
Introduction
Through the use of the eval() function in Ruby you can dynamically interpret code at run-time. This capability gives the programmer the power to design applications that can rewrite themselves by generating new code and executing it.
Reasons eval() Can Be Useful
This feature could be useful in dynamically adapting a program to different versions of interfaces, in generating tables or arrays of functions to be called in a state machine, or in creating a new class from existing classes. Other reasons to use eval include allowing a user to submit arbitrary command sequences for evaluation, saving the developer from creating a complex expression parser (for more information see Macros and Evaluation).
eval() Support in Ruby and Java
eval() capability is fully supported at the language level in Ruby, thus it is available without adding to or extending Ruby in any way.
The base Java language does have the capability, with libraries such as Simple Expression Evaluator for Java, and JFormula 3.6 - Math expressions API, to evaluate mathematical expressions, but not arbitrary code. This means that even with these libraries in Java, a programmer can evaluate some math at run time, but can't define new methods and then call them. Because these features do not permit the evaluation of arbitrary code, it can not define new classes and methods, making it less powerful than Ruby's eval().
Although Java doesn't have a built in eval() type of function, the BeanShell addon is available which provides full eval() functionality, equivalent to Ruby's eval().
It is more common to see eval() type functions in interpreted languages then compiled languages. In compiled languages you will need some sort of mechanism for interpreting code at run-time which usually isn't built in.
Advantages and Disadvantages of Using Eval()
As with many powerful features, the ability of programs to use eval() to dynamically extend themselves at run time, by programatically generating and executing new code, comes with benefits and potential disadvantages. These advantages and disadavantages are similar for Java and Ruby, and apply to any language which offers this dynamic eval() capability.
Advantages
eval() provides some powerful advantages to the programmer, including the ability to write programs that can compose (and execute) arbitrary new code at run time.
Dynamic Metaprogramming Capability
The capability to generate new code at run time provides the application developer the option of having the program dyamically extend itself, leading to a dynamic Metaprogramming capability.
Flexibilility
Ruby's eval() capability permits Ruby developers to create flexible applications able to rewrite and extend themselves on the fly. Even a statically typed, compiled language such as Java can take advantage of this flexibility through the BeanShell addon.
Ease of Use
eval() makes it easy to extend the functionality of programs at run time, to adapt to new requirements or changes in the underlying data model. Because the interpreter or virtual machine instance retains state across invocations during the instance lifetime, the current process, and all new code passed to eval(), are available to the application, easing the burden on the programmer.
Disadvantages
The power and flexibility comes with the price of some performance and security.
Performance
eval() must parse the code and compile it at run time, and in view of this, the performance implications of interpreted languages apply. The speed penalty may not be worth the flexibility of generating new code at run time in some cases. In addition, adding new code will increase the memory required by the application, and the increased load on the processor may have to be considered for non-trivial run time evaluation.
Security
If eval() is used to generate and invoke code created with arbitrary data received from another, less trusted application, extreme care should be used. Allowing an untrusted application to modify a running program is an enormous security risk. It is difficult to imagine circumstances in which eval() should be used in safety critical applications.
Ruby Example of Eval()
The eval function can be used for dynamically calling functions. For example in Ruby:
# Create an array of functions
functions = ['function1','function2','function3']
# Define functions
def function1()
puts "inside function1"
end
def function2()
puts "inside function2"
end
def function3()
puts "inside function3"
end
# Function that will dynamically call other methods
def CallFunction(functionname)
eval functionname
end
#Results
irb(main):039:0> CallFunction functions[0] # Call Function1
inside function1
=> nil
irb(main):040:0> CallFunction functions[1] # Call Function2
inside function2
=> nil
irb(main):041:0> CallFunction functions[2] # Call Function3
inside function3
=> nil
The method names can be stored in an array and called dynamically at run-time. Taking this one step further we can store the function definition in an array as well:
#Create an array for the function name and function definition function1 = ['function4'] # We could have more then one function2 =['def function4 () puts "inside function4" end'] # We could have more then one # Dynamically create/define function eval function2[0] # Dynamically call the the name of the function irb(main):060:0> eval function1[0] inside function4 => nil
So, in this example function4() is created dynamically at run-time. This allows us to store parts of the code as data and to create functions when needed. An application written using eval to dynamically generate functions would allow new functions to be added easily. For example, you could create new definitions of functions and insert them into a database table.
Java Example
The base Java language does not have a direct equivalent to Ruby's eval() facility, because Java is a compiled, statically typed language. However, Java environments can use a facility similar to Ruby's eval() through the BeanShell addon, available from bsh-2.0b4.
Because a Java program can create instances of the BeanShell interpreter, and submit arbitrary Java code to the interpreter for evaluation at run time, Java developers can take advantage of BeanShell's eval facility to create programs that create and execute new programs on the fly.
A BeanShell interpreter object retains state across invocations, so that previously evaluated classes, objects, methods, and variables are available to classes, objects, methods, and variables that are evaluated later in the lifetime of each BeanShell instance. For details, see the BeanShell Manual, Embedded Mode
BeanShell is distributed as an executable jar, which when run, provides a BeanShell workspace in which Java code can be developed in a command line environment. To view the output from the code below in the BeanShell Workspace, the user must invoke "File->Capture System in/out/err" from the "Bsh Workspace" menu, before running the code.
import java.lang.StringBuffer; // needed for mutable strings
// Create an array for function names
StringBuffer[] functions = new StringBuffer[4];
// Define functions
void function1() {System.out.println("inside function1");}
void function2() {System.out.println("inside function2");}
void function3() {System.out.println("inside function3");}
// Factory Pattern
void CallFunction (functionname) {eval(functionname + "();");}
// Initialize function names
functions[0] = new StringBuffer("function1");
functions[1] = new StringBuffer("function2");
functions[2] = new StringBuffer("function3");
// Results
bsh % CallFunction (functions[0]);
inside function1
bsh % CallFunction (functions[1]);
inside function2
bsh % CallFunction (functions[2]);
inside function3
bsh %
As in the Ruby example, the function names in the array can be stored in an array, or someplace else like a database or xml file.
In addition, similar to the Ruby example, we can store the function definition in an array as well:
// Create an array for the function name and function definition
StringBuffer[] function1 = new StringBuffer[10]; // We could have more than one
StringBuffer[] function2 = new StringBuffer[10]; // We could have more than one
// Dynamically create/define function
function1[0] = new StringBuffer("function4");
// Dynamically call the the name of the function
bsh % function2[0] = new StringBuffer("void function4() {System.out.println(\"inside
function 4\");}");
bsh % eval(function2[0].toString());
bsh % CallFunction (function1[0]);
inside function 4
bsh %
As in the Ruby example, in this Java example using BeanShell eval(), function4() is created dynamically at run-time. This allows us to store parts of the code as data and to create functions when needed. An application written using eval to dynamically generate functions would allow new functions to be added easily. For example, you could create new definitions of functions and insert them into a database table.
Links
Ruby Links
Java Links
Alternative Java Expression Evaluators
Janini Java Expression Evaluator
JEPLite: java expression parser enlited
Message Board Threads on eval() in Java
Java Programming - evaluate expression
eval() in Javascript
Is there an equivalent to eval() in javascript?
Invoke dynamic languages dynamically, Part 1: Introducing the Java scripting API