Lambdas and Anonymous Delegates pt 2

So I did exactly what I suggested to myself, and sent an email to Eric Lippert regarding the inline lambda and anonymous delegate business I blogged about earlier today. He's quick - here's his reply:

Ah, good question.

You can make this work by casting to a delegate type.

((Func<bool>)()=>M()).Invoke();
((Func<bool>)(delegate (){return M();}).Invoke();

For lambdas, we don't know whether the lambda is to be converted to a delegate or an expression tree. Since an expression tree cannot be invoked, we cannot allow you to invoke it before we know what you intend.

For anonymous methods, yeah, I suppose we could have designed C# 2.0 to allow invoking them directly. But what would the point be? Why say (delegate(){return M();}).Invoke()" when you could just say "M()"? I agree that for reasons of generality and consistency it would be nice to have, but I don't think disallowing this actually disables many scenarios.

He's right - it's not like this is some sort of necessary functionality. On the one hand, it's great that I can get exactly this functionality by casting to a delegate type first, but on the other hand it's extra cruft. All the same, if some inspired or courageous EverDroid user needs this sort of functionality, it's there.

Thanks, Eric! =)

Inline execution of a lambda expression or anonymous delegate

I had an idea I wanted to try to exploit that would make my VB.NET expression a bit more, well, expressive.

The way I built my expression builder, essentially you pass in a valid VB.NET expression and the expression builder sticks it onto the end of a Return statement inside a static method of a new class, which it then compiles with the stock VB.NET code compiler. Then I use reflection to dig out the compiled method and returns a Func<bool> that invokes the method, returning the result.

This works great for its intended purpose, but it unfortunately restricts my expressions to being, well, expressions. Boolean expressions. Which means they can't do a whole lot of processing, or whatever. That's okay - I don't really *need* all that much. But I enjoy a challenge, and I thought to myself 'Self, why don't we try to use an inline lambda!'

Great idea - except you can't INVOKE a lambda inline. Example:

bool x = ( () => true ).Invoke();

Doesn't work. Operator '.' cannot be applied to operand of type 'lambda expression'. 'Fine,' says I, 'I'll use an anonymous delegate!'

bool x = ( delegate { return true; } ).Invoke();

Operator '.' cannot be applied to operand of type 'anonymous delegate'. Argh. So as far as i'm aware, this is impossible, probably by design. I'll have to ask Eric Lippert about it, since he's been doing his C# design rationale posts lately.

Progress thus far

Truth be told, at this point, progress is slow going. I've been doing a lot of work for my friend Brian, and as it pays more immediately than Macrocrafter work I have to focus on that. It's been calming down of late (unfortunately) so that should afford me more time to work on this stuff.

I've been slowly organizing the work I need to do on this system. So far I've divided it up into several main components that I'll need to build before I can start stitching them together.

Also I've decided to skip the intermediate step and just make a damned hunter bot out of this, so I'll also need to integrate targeting and walking into the system.

Anyway - one of those systems includes a VB.NET compiler, as I mentioned before. All I need to do with this is to take my proof of concept code and make a proper system out of it - something reusable. I'm going to add it to Autocrat. I'll also need a timer that can produce timing events in 150ms intervals - that's the interval at which I'm going to have the system pull down updated EQ2 data. Any of the .NET timers should be capable of doing this so I'm not too worried here. I'm also going to need a state machine engine that can be configured via XML - this will be responsible for persisting the user-editable 'scripts' that will make up the brains of the bot itself. I've got a schema all figured up; all that remains is actually putting the work into it - and then I get to build a UI on top of it, because the last thing I want to require is for people (or myself) to have to hand-edit XML to get it to work.

Coupled with the timer is going to be a system for describing stateful property changes. Here's the problem: my bot engine is going to be evaluating event trigger expressions every 150ms. The expressions are boolean, so they'll be testing data based on the content of the current tick's EQ2 properties. If an expression looks similar to: "Return EQ2.GroupMember(2).Health < 50" - then when group member 2's health drops below 50%, it will trigger the event. This is desired - BUT it will ALSO trigger once every 150ms tick until group member 2's health rises to or above 50. This is NOT desired, so I need a way to expose the concept of state changes: for example, I'd like to rewrite the same expression as: "Return EQ2.GroupMember(2).Health.DroppedBelow(50)". That would return True if, and only if, the previous tick's Health was at or above 50 AND the current tick's Health is below 50.

So I'll need to figure out an efficient system of encapsulating old state and exposing it via these state descriptors. I'm sure I'll figure it out.

But you can see why I've been making slow progress here. =) In between kids, contract work, looking for full-time work, EQ2 updates, and other life, I don't have much time for experimentation, which is a shame.

Happy Birthday baby girl =) And some other stuff.

So my daughter just turned 2 years old today. We had a party at CiCi's, and fun, pizza, and cake was had all around. =)

A friend of mine in NYC has a lot of extra work he's unable to do himself, due to having his own full-time job, family, etc., so he's farming it out to me. It's a great situation as I currently have a work deficit, so taking it off his hands helps me out a lot. So far I've updated three websites owned by the same company - took me the better part of a day to get it all completed, but it wasn't terribly hard. Mostly grunt-work, but it pays, and it's not brainless. I am by no means complaining. =)

Finding work in Charleston, SC has been difficult - the two large development companies out here have both rejected me as being "not quite what they want" - I don't have any team-oriented experience, or large enterprise experience, which is all they're looking for out here. My argument is that if everyone wants to hire people with prior team experience, how can anyone who doesn't have it get it? Eventually there'll be nobody left except people *without* team experience, and then what will they do?

Bah. Perhaps I'm bitter - I need a job so I can put food on the table and pay rent and all that rot. It's frustrating. I'm actually hoping I find a computer / network tech job I can do, rather than programming - I think I'll have more fun doing that, since I'll get to work with people, help solve their immediate problems, etc.

Maybe I'll see if I can find investors for that computer recycling business I've been thinking of starting.

Anyway - very little work has occurred on my next EQ2 project, which one of my friends has dubbed EverDroid - a name I think I'll use. I need to build an Eval system into Autocrat, because this is something I'm going to use a lot in projects to come, for the sake of configuration and behavior customization. That won't be hard - I've got a simple console proof of concept working just fine, compiling expressions in VB.NET (which I think non-programmers will find easier to reason about - Microsoft agrees, and has done the same thing for their expression compiler in Windows Workflow 4).

EverHarvest had a good month last month - decent sales, despite the economic issues. Not nearly what it used to be, but then again I never expected it to be. Most people seem to be happy with it, when it works - which it's not at the moment; I've been working on the next update in response to Sony's latest changes, but they're doing their rapid-fire patching again, which always makes me want to wait... I was half-way through the last set of updates when they patched again yesterday, ruining my work. Ah well - such is the price to pay. Overall it's nothing to complain about - I work on it maybe two days total out of every month.

So, that's my update today. I'll get back to the millstone again and crank out the EH update. I really need to write some helper tools for this thing...

So far, so good.

Early proofs of concept indicate this might be a viable option. Here are some of the hurdles ahead of me:

Trigger condition evaluation: I'm planning on using VB.NET as the language for trigger condition expressions. I've built a simple test that takes an arbitrary boolean expression, compiles it via CodeDOM into an in-memory assembly, and returns a delegate to invoke it with all the necessary information that will be required of the bot. This works well so far, but of course it remains to be seen how well it'll work in practice. I may change this to having the expression compiled into a System.Linq.Expression, which I could then compile into a delegate - I think that would be faster than the CodeDOM method I'm using now. Profiling will tell.

Dynamic state machines: Part of the system will be handling different states. I'm using states as a logical grouping of events for a common scenario. For example, a Fighter bot could have many states that it could be in at any one time; an out of combat state, an in-combat state, a dead state, and so on. Each of these states can have their own grouping of events, possibly having actions that force the machine to transition between states. I have some experience with state machines so this shouldn't be too much of a challenge.

Ability mapping and timing: Another vital part of the system is my Ability Mapping concept. Instead of having each user modify their scripts to include their character's own personal spells and combat arts, scripts will instead reference ability mappings. These mappings are then configured by the user to refer to one or more in-game abilities, each of which has its own cast, reuse, and duration timers. When a script action invokes the mapping, the system will use the first ability it's able to, according to its position in the mapping (top to bottom) and the status of its timers. This way, said Fighter bot can have a mapping called 'attack' that its user could have configured with all his character's single-target combat arts, so that every time that 'attack' mapping is invoked, one of those combat arts is used. I'm still conceptualizing this particular feature so any more detail would be speculation / brainstorming, which I will save for another post.

Configuration and UI: Of course, all of this complexity needs to be presented as simply as possible. I do not intend for people to manually edit XML in order to change these scripts, so I'll be building a script editing UI for that purpose. In addition, the UI will need to include a mapping editor, as well as a place for entering script-declared variables. All of these things will need to be data-aware, and some of them dynamically created. I've been debating with myself whether I want to use XAML and WPF for the UI, and while this sounds like it might be a good use of WPF, I still haven't taken the time to crest the learning wall, and with my current job situation I don't know if I want to take that time now.

There's more, but this is getting long already - I'll post again either tomorrow or the next day with more.

Big plans.

So I hinted, last time, about some big plans I have for a new EQ2 project. I think I may have gone off the deep end a bit with this one, but I'm thinking I might have a winner.

People have been begging for a hunter bot for a while now, and while I could make one, I don't think I could do it justice. EQ2 is a very complex beast, with combat being especially complex. Despite what many think, it's *not* as simple as walking around, targeting stuff, and mashing buttons until you're done. When you play EQ2, and you fight things, you create little strategies out of the abilities you're given, and considering that there are 24 classes and umpteen billion ways of customizing (AA, etc), there are far too many of these strategies to bake in to the software.

So, in order to appeal to everyone, I'm going to stop working on the hunter bot, and start working on my new project, a group assistant (or you could call it a two-box helper).

This will, however, be like no ordinary two-box helper. My system will be completely customizable, down to every last detail, thanks to a custom rules engine I plan to build. I'm scrapping my DSL idea for this - it works, but I don't want to trust it for a production build, not yet - and going with an XML rules system with a visual designer on top of it.

This will not be a general purpose scripting application. The rules engine will support the project's only goal - to make two-boxing (or three, four, five, twenty, etc) a hands-off endeavor.

But I have ulterior motives as well... This application will facilitate a rules-based system upon which the final hunter-bot can be built. Once the rules engine is in place, and all the wiring is complete, it'll be go time.

More details to come later. =)

Moving!

Despite a number of promising leads, all of my job resources have run dry. As such, in three days, I am moving back home to Charleston, SC. My spirits are up, actually - I expected to feel thoroughly defeated (which I did), but at the moment I'm happy. I'll get to see my friends and family again. That's the important thing.

Meanwhile, I haven't let the out-of-work situation hold me back from programming. If anything, I'm more prolific on my own projects than I was able to be before, what with the 8-hr a day grind. Most recently, I'm working on a new EQ2-related project. The project itself is, well, top-secret at the moment - but I can discuss one of its major components: a script lexer and parser.

Compiler design and implementation is definitely one of my favorite topics in computing. I may be one of the few people in the world who enjoys hand-writing parsers. I can easily say that the grammar I wrote for this particular lexer/parser is the most complex grammar I've written with the intention of actually parsing.

In the span of about 3 hours tonight, I've hand-written the lexer and recursive descent parser for this language, which builds the inputs into a state machine. I can't give away too many details, but suffice it to say that this system will automate some interesting tasks in EQ2 (much like other MacroCrafter software I've written) and will allow the user to change its behavior on the fly.

To be honest, I don't really expect the user to do much editing - although I intend for the scripts to be user-editable, I will likely be writing all the scripts myself and releasing them with the new software. What I really wanted out of this was an easy way for me to describe the logic that this software will execute.

Sorry I'm being vague. I haven't made any official posts regarding this stuff yet on the MC forums, and I don't want to jump the gun. I've made the mistake in the past regarding talking about projects and such on this and other venues, getting people's hopes up, and then either abandoning or shelving said project. People's feelings get hurt, I get complaints, etc. I shouldn't promise what I might not deliver - and this vagueness is part of that.

As for when I'll be announcing this -- well, it'll come once I flesh the concept out more, build more proof of concept bits, and start pinning stuff together. I think it'll be well-received - possibly more so than a hunter bot would...