polterguy / Lizzie
Labels
Lizzie a scripting language for .Net
Lizzie is a dynamic scripting language for .Net based upon a design pattern called "Symbolic Delegates". This allows you to execute dynamically created scripts, that does neither compile nor are interpreted, but instead "compiles" directly down to managed CLR delegates. Below is an example of using Lizzie from C#.
using System;
using lizzie;
class MainClass
{
public static void Main(string[] args)
{
// Some inline Lizzie code
var code = @"
// Multiples 10 by 2 and adds to 57
+(57, *(10, 2))
";
// Creating a Lizzie lambda object from the above code, and evaluating it
var lambda = LambdaCompiler.Compile(code);
var result = lambda();
// Writing the result of the above evaluation to the console
Console.WriteLine("Result was: " + result);
// Waiting for user input
Console.Read();
}
}
Lizzie is highly influenced and inspired from Lisp, but without the unintuitive "Polish notation". In such a way, it arguably is dynamic Lisp for the CLR. Its dynamic nature allows you to execute snippets of Lizzie code, inline in your C# code, by loading your code from files, or by for instance fetching the code from some database of some sort, or even transmit code over the network to have a server endpoint (securely) evaluate your code.
You can easily create your own "keywords" in Lizzie, which allows you to create your own DSL or "Domain Specific programming Languages". Lizzie hence easily lends itself to richer rule based engines, and similar domain specific problems, where your code needs to be more dynamic in nature than that which the CLR allows you to through C#, VB.NET or F#.
What is a Symbolic Delegate?
A Symbolic Delegate is a CLR delegate that is dynamically looked up during runtime from a dictionary of delegates with the same signature. This allows you to dynamically wire together delegates to an "execution tree" during runtime, based upon whatever delegate happens to be the value for your "symbol". Lizzie is literally a dictionary of delegates, where the key to lookup your delegates are of type string. This allows you to easily extend Lizzie by simply creating a new delegate, and associating it with a "symbol", to such have access to execute CLR methods from your Lizzie script code.
MSDN Article about symbolic delegates written by yours truly
Binding Lizzie to your own types
If you want to, you can "bind" your Lizzie code to a CLR type. This allows you to extend your Lizzie code with your own C# "keywords", to create your own "DSL". Below is an example.
using System;
using lizzie;
class MainClass
{
[Bind(Name = "write")]
object Write(Binder<MainClass> ctx, Arguments arguments)
{
Console.WriteLine(arguments.Get<string>(0));
return null;
}
public static void Main(string[] args)
{
// Some inline Lizzie code
var code = @"
write('Hello World!!')
";
// Creating a lambda function from our code, and evaluating it
var function = LambdaCompiler.Compile(new MainClass(), code);
function();
}
}
How small is Lizzie?
The entire reference documentation for Lizze is roughly 12 pages if you choose to print it. This is the entire reference documentation for the language. This allows you to learn the entire programming language literally in 20 minutes. The "compiler" for the language is less than 500 lines of code, and all "keywords" are less than 1,000 lines of code in total. The project as a whole has roughly 2,200 lines of code, but 50% of these are comments. When built, the DLL is roughly 45KB on disc. There are 7 public classes in the project, one attribute, and one interface. There are less than 30 methods in total, and you don't have to use more than a handful of these to start adding dynamic scripting abilities to your CLR code. This arguably makes Lizzie the smallest (useful) programming language on the planet, if you ignore languages such as "brainfuck", arguably created more or less as a joke.
The Lizzie tokenizer also contains only 7 different tokens. There are no operators in the language, no keywords, and only one type of statement. In fact all "statements" are "functions", that all have the same signature. Compare this to the 500+ keywords, and 50+ operators of C#, and the 1,000+ pages of reference documentation for C#, and hopefully you understand the advantage.
How fast is Lizzie
When profiling a language such as Lizzie, there are two important things to measure.
- Compilation speed
- Execution speed
Compilation is blistering fast, at least if you consider the fact that the compiler is written in C#. Below is an example that compiles 10,000 snippets of Lizzie code. On my machine this code is finished after ~2 seconds.
using System;
using System.Diagnostics;
using lizzie;
class MainClass
{
public static void Main(string[] args)
{
// Some inline Lizzie code
var code = @"
+(5, 2, 50)
-(100, 30, 3)
*(5, 3, 2)
/(100, 4)
%(18, 4)
";
// Compiling the above code 10,000 times
Console.WriteLine("Compiling some Lizzie code 10,000 times, please wait ...");
Stopwatch sw = Stopwatch.StartNew();
for (var idx = 0; idx < 10000; idx++) {
var lambda = LambdaCompiler.Compile(code);
}
sw.Stop();
Console.WriteLine($"We compiled the above code 10,000x in {sw.ElapsedMilliseconds} milliseconds");
// Waiting for user input
Console.Read();
}
}
On my computer, which is a MacBook Air from 2016, the above code compiles 10,000 times in roughly 2,100 milliseconds. Since Lizzie is a dynamic scripting language, intended to frequently retrieve snippets of dynamic code, and compile these, before it executes the result - The compilation speed is hence arguably equally important as its execution speed. On my computer, compiling Lizzie itself, and its unit tests once requires 4.62 seconds! Compiling the above Lizzie code 10,000 times took me only 2 seconds.
Execution speed
If we slightly modify our above code, to execute the code 10,000 times, instead of compiling it 10,000 times, such that it resembles the following ...
using System;
using System.Diagnostics;
using lizzie;
class MainClass
{
public static void Main(string[] args)
{
// Some inline Lizzie code.
var code = @"
+(5, 2, 50)
-(100, 30, 3)
*(5, 3, 2)
/(100, 4)
%(18, 4)
";
// Executing the above code 10,000 times!
Console.WriteLine("Executing some Lizzie code 10,000 times, please wait ...");
var lambda = LambdaCompiler.Compile(code);
Stopwatch sw = Stopwatch.StartNew();
for (var idx = 0; idx < 10000; idx++) {
lambda();
}
sw.Stop();
Console.WriteLine($"We executed the above code 10,000x in {sw.ElapsedMilliseconds} milliseconds!");
// Waiting for user input.
Console.Read();
}
}
The results on my computer says 722 milliseconds.
Lizzie is not as fast as C#, since each function invocation also requires a lookup
into a Dictionary. Each function invocation also implies evaluating a
delegate, which has an additional overhead of 20% compared to invoking a virtual
method. So you can't expect a Lizzie lambda object to evaluate nearly as fast as the
equivalent C# method. However, compared to the execution speed of an interpreter
written on the CLR, and/or a "true compiler" written on the CLR, Lizzie will
purely mathematically outperform both of these for all practical concerns,
assuming you have an interest in executing dynamic code. Since most practical
snippets of code does complex tasks, such as accessing the file system, and
reading/writing to databases, fetching data over sockets, etc - The execution
speed overhead of your Lizzie code for most practical concerns will be irrelevant.
DISCLAIMER - I would not encourage you to use Lizzie for extremely CPU resource demanding tasks, such as polygon rendering, algorithmic intensive math operations, complex parsing, etc. Because after all, it will never execute as fast as the equivalent C# code, due to its dynamic nature.
A 5 minutes introductory video to Lizzie
Reference documentation
Installation
PM > Install-Package lizzie
Or visit the download page to get its source code