How do I go about creating my own API/Framework/Libraries - winapi

Some libraries are very obvious to me. I mean I can create the math, string etc libraries on my own. I don't understand how people create API's like OpenGL, OpenCV, DirectX, and MFC. I don't understand how to write them on my own like I write libraries to compute math and string functions. Are there any resources on the net that would teach me how I can go about doing this.

Those kind of libraries are driven by two things.
1) is the domain that they're working in (notably GPU architecture, and their capabilities and limitations), and
2) the model of those capabilities as view by the API designer.
Simply, someone with some (ideally) reasonable understanding of the problem domain said "I think if you want to work with a GPU, I'd like the GPU to look like this", and came up with a model to present to the API consumer. Then the framework was written to convert that model view that the API designer contrived with the actual workings of the underlying mechanism (in this case a GPU).
Consider something like an Object Relational Mapper tool. Here, they're trying present an OO view mated to an underlying Relational representation.
The designers likely took some idea of what they wanted, tried them out to see how realistic it was do, and then started filling in the gaps and polishing the edges.
The way for you to start is to simply pick a domain which you have knowledge, but don't like how it works, and think "gosh, it would be better like this" and then start solving THAT problem. If things go well, you'll have some momentum and the process will likely get a bit more organic from there. But, ideally not too organic.
The hard part is putting your new API to work and USING it as someone who doesn't know the API, or necessarily the domain, would use it. Using it also gives you the opportunity to encounter the "gosh, it would be better like this" phase again.
Rinse and repeat until you're happy with the outcome.
Few paper designs survive contact with the actual development of the system. Some do better than others, but it's hard to know how an API is going to feel as a developer until you start using it.
So try a few apis on your own, study those you like, study those you hate and think how to make them better, and work on some proof of concept implementations to see how it goes.

Related

User Interface Design Package

I am working on designing a user interface and want to convey my design to my peers. Typically, I'd use a mock-up of the UI and UML, but given the complexity, size, and multiple asynchronous interactions I'm not sure that this makes the most sense. It doesn't seem to allow to efficiently describe the process.
Does anyone have experience in designing large User Interface's? How would a company that designs UI's go about modeling their process/ design? This seems to be a question for the 'front-end' engineers.
well let's go for an answer even if your question is very large. I'll give you my experience and hope it will help you to decide.
To my mind, using UML in a project is very useful if you use it right, it's often clear, quite easy to understand. On the other side, when working on huge project, it's sometimes costly because it takes time to create your diagrams and nowadays, you have to be as fast as possible !
These last months I have worked on a project with a GUI and to well define it, I have first draw some screen on paper and quickly ask for some drawbacks near the coffee machine.
I have created some activity diagrams and then, created a mock-up.
How to choose now ?
First of all, if your GUI should have a beautiful design, I don't think you can avoid a mock-up, but I'm don't know if it's your goal.
My main tip would be, try to simplify your complexity in your program. If you have multiple asynchronous that could arise everytime/everywhere, you'll encounter a lot a problem.
In my project, let's say my tablet could be preempted by an external computer whenever/wherever. I just have created a class beforehand to manage this asynchronous call but my sub-activity-diagrams stayed intact.
Draw specific diagram for specific functionalities and leave useless diagrams which are not relevant/too simple.
Hard to explain you more than this without giving you any example. Hope it will help a bit.

First time porting a library from one language to another

I am porting a library from C++ to Java. This is my first time and I am not sure
about what "porting" really means? Specifically, what if the author named a variable
as 'A' and I think that a better name would be 'B'. Same for methods, classes and namespaces.
Also, what if I think something can be done better? Does porting mean that I should try
to keep as much of the original code spirit as possible, but still allow myself freedom
to improve stuff?
Thanks
It doesn't necessarily have to be a one-to-one translation (and in many cases, it can't be done). Porting is just rewriting a piece of software in a different language/environment/etc. Sometimes porting will require you to tweak things and implement them in different ways altogether, so I think the last sentence of your post pretty much captures the gist of things.
I view it as comparable to translating a book from English to another language. There will be instances where judgment calls need to be made in terms of how to express the intent/function of the source material.
When porting from System A to System B, the world is your oyster. You can pretty much change anything if you believe it's an improvement. The only caveats to that are when dealing with interfaces. Say, you are porting an API, for example, it wouldn't be a good idea to name externally-available methods, as that would break something down the road. Tracing naming issues across multiple classes is a major pain.
As someone who's done a fair bit of porting from language to language, I would recommend sticking to implementation details first and foremost. A good engineering principle is to change one thing at any time. That way, when things don't run as expected, you'll know that it's your implementation that is to blame, and not some silly naming issue. And when you do come to renaming, I suppose it goes without saying, be very careful and backup often. This is one case where software versioning may save you hours of time.
When "porting" a library from one platform to another, you are porting functionality. You are not porting style of code. It isn't like in literature, where one must maintain the style of the piece, keeping in mind metaphors and iambic pentameter or what have you.

Knowledge of windows internals?

I wondered if any of you have knowledge of the internal workings of windows (kernel, interrupts, etc) and if you've found that you've become a better developer as a result?
Do you find that the more knowledge the better is a good motto to have as a developer?
I find myself studying a lot of things, thinking with more understanding, I'll be a better developer. Of course practice and experience also comes into play.
This is a no brainier - absolutely (assuming you're a developer primarily on the Windows platform, of course). A working knowledge of how the car engine works will make a lot of common programming tasks (debugging, performance work, etc) a lot easier.
Windows Internals is the standard reference.
I believe it is valuable to understand how things work underneath. CLR/.NET to C++, native to ASM, ASM to CPU architecture, building registers and ops from logical gates, logical gates from MOSFETs, transistors from quantum physics and the latter from respective mathematical apparatus (group theory, etc).
Understanding low level makes you not only think different but also feel different - like you are in control of things, standing on the shoulders of giants.
More knowledge is always better, and having knowledge at many levels is a lot more valuable than just knowing whatever layer of abstraction you are working at.
A good rule of thumb is that you should have a good knowledge of the layer below the layer where you are working. So, for example, if you write a lot of .NET code, you should know how the CLR works. If you write a lot of web apps, you should understand HTTP. If you writing code that uses HTTP directly, then you should understand TCP/IP. If you are implementing a TCP/IP stack, then you need to understand how Ethernet works.
Knowledge of Windows internals is really helpful if you are writing native Win32 code, or if OS performance issues are critical to what you are doing. At higher levels of abstraction, it may be less helpful, but it never hurts.
I dont think that one requires special or secret knowledge of internals such as those that may be extended to members of the windows team or those with source access but I absolutely contend that understanding internals helps you become a better developer.
Take threading for instance, if you are going to build an application that uses threading in even a moderate way - understanding how windows works, how the threading works, how memory processes work are all keys to being able to do a good job with that code.
I agree to a point with your edict but I would not agree that experience/practice/knowledge are mutually exclusive. That net-net of experience is that you have knowledge gained from that experience. There is also a wisdom component to experience and practice but those are usually intangible situational elements that you apply in the future to avoid mistakes. Bottom line knowledge is a precipitate of experience.
Think of it this way, how many people do you know with 30+ years of experience in IT, think of them and take the top two. Now go into that memory bank and think of the people you know in the industry who are super smart, who know so much about so many things and pick the top two of those. You now have your final 4 - if you had to pick one to start a project with who would it be? Invariably we pick the super smart guy.
Yes, understanding Windows internals helped me to become a better programmer. It also taught be a lot of bad practices, bad ideas, and poor design concepts.
I highly suggest studying OS X or Linux internals as an alternative. It'll take less time, make more sense, and be much more productive.
Read code. Read lots of code. Read lots of good code. jQuery, Django, AIR framework source, Linux kernel, compilers.
Try to learn programming languages that introduce you to new approaches, like Lisp, Ruby, Python, or Javascript. OOP is good, but .net and Java seem to take the brainwash approach on it, and elevate it to some kind of religious level, instead of it just being a good tool in your toolbox.
If you don't understand the code you are reading, it likely means you are on the right track, and learning new techniques.
I'd suggest getting a mac simply because you'll find yourself wanting to make your UIs simpler and easier. It's really important to have a good environment if you want to become a great programmer. Surround yourself with engineers better than yourself (if you can), work with frameworks and languages that take the 'engineer' approach vs. the 'experimenter' approach, and... use a operating system that contains code better than yours.
I'd also reccomend the book "Coders at Work".
It depends. Many programmers who understand the internals of a system begin writing optimised code to exploit that knowledge. This has three very serious side-effects:
1.) It's harder for others without that knowledge to extend or support the code.
2.) System internals may change without notice, whereas interfaces are usually versioned and changes discussed publicly.
3.) Interfaces are generally consistent across platform revisions and hardware, internals do not have this consistency.
In short, There's a lot of broken, unsupportable code out there that's borked because it relies on an internal process that the vendor changed without notice.
Father of language C said that "you don't need to learn all features of language to write great codes. Better you understand the problem, better you write the code." Having knowledge is always better.

Game framework architecture - view components or MVC?

I'm trying to build a very light re-usable framework for my games, rather than starting from scratch each time I start a game. I have a component driven architecture - e.g. Entity composes a Position component and a Health component and Ai component etc.
My big question is whether my model composes view components to allow for more than one view of the model, or whether to use a truer MVC where the model does not know about its views, and they are managed externally.
I have tried both methods but if anyone knows the pros and cons of each approach and which is the industry standard, it would be great to know.
depends on your audience, game devs, myself included aren't very used to the MVC model, although most know it, it's not as easy to keep it clean cut, because of development casualties (not any serious technical reasons). So from experience, I've seen dozens of game frameworks start as MVC, but only a pair were able to maintain it until the end. My theory is MVC adds too much complexity and little benefits for small throwaway games (with normally a few devs), and it's to hard to keep really cleanly separate most game objects into these layers for large/complex games. And since games have a release date, they many times sacrifice code clarity and reusability for performance and quick adhoc solutions (that will get rewritten if necessarry in the sequel (if there is one)).
However, with the caveat above, it's better to aim high, because if you succeed it's better :) and if you fail, well to bad. So you should probably try the MVC, but don't worry if it fails, profesional game devs have all failed at the task many times :)
I’d certainly vote for the model to know nothing about its views. Loose coupling is good: Simpler model code, easier testing, more choices.
I know this question might be outdated, but I need to reply on it.
Actually, I started programming a game in Lua (with LÖVE) and I started programming a MVC - Framework for it.
At first, to use MVC really depends on what you want.
I know my problems with game programming, when the program becomes bigger, and mostly the structure becoms too complex to maintain.
Next thing is, I know that I will change all the graphics when I find an artist who is willing to work for it. But until then, I'm gonna use my own dummy graphics.
I want the artist to feel free to do what ever he wants, without beeing dependend on any resolution or color restriction.
That means, I might have to change the whole (!) presentation code. Maybe even the way objects interact (collision detection, f.e.).
The game logic is captured in the models, so I can concentrate on that. And I think game logic is the most important part of making a game. Isn't it?
Hope you see my point.
But, if you have everything together: all the graphics, sounds, the whole thing; then you can code straight forward.
My MVC is a configuration-over-convention-ass, that slows down prototyping a bit.
BUT(!) iterations of development can be made much more easily. Testing, especially Unit-Tests are done much more faster.
I would say MVC turns you development-speed-curve (which is normally an anti-exponential curve) into an exponential curve. Slow at the beginning, but more and more fast at the end.
MVC works really well for games, at least for my games which are designed for cross-platform.
It really depends on how you implement it in order to get the benefit.

Feature bloat - how much is too much?

I'm a computer science student designing a project and I've started wondering what are good examples or software, or even hardware that are toeing the line between being feature rich with good usable features for regular users and being too intimidating for new users. Also could anyone recommend any good tips/books for designing good quality applications that are feature rich but not "bloated"?
"Make everything as simple as possible, but not simpler." - Albert Einstein
"Perfection is reached not when there is nothing left to add, but when there is nothing left to take away." - Antoine de Saint-Exupéry
I am not trying to be flippant but these quotes really are the best advice. Simplicity of design should be your goal. Not that achieving simplicity is easy! On the contrary, it is quite difficult but it is possible.
Try thinking about things a bit differently. Rather than
How many things can I add before this becomes bloated?
try
What are the fewest number of features and elements I can include while still providing a superior experience for my users?
Here's a good set of slides from a presentation on the topic: Rescue Princess 2.0.
The first order of business should just be keeping the application easy to use. Beyond that, all I can say is, beware of writing features for an imaginary user: make sure someone actually needs it before you start coding.
As a direct answer to your question: pretty much any Microsoft product. I'm showing my bias here, but Microsoft has a strong tendency to keep their codebase, and add features on top of features until the original functionality of the app is nearly lost beneath mounds of accreted crud.
Look at MS Word, for example; while you can still just open it up and start typing, god forbid if you want to renumber a section of your document while leaving the rest alone. Heaven forbid if you want to generate a Table of Contents that includes references to an Appendix. This sort of stuff is something that is de rigeur for Word Processors, and Word supports it, it just supports it in a way that you cannot get it done without a manual, several cups of coffee, and bandages to stop the bleeding from banging your head on the desk.
Microsoft isn't alone in doing this; this thing tends to happen all the time, with all sorts of products; but they are among the worst offenders, I've found.
1: What do your users need, and want, and
2: Which features will you have time to implement?
Your question is pretty general. Which features constitute bloat? That kind of depends on whether you're writing an antivirus scanner, an OS or a word processor.
There is no clear barrier between "good" and "too much".
However, it depends on what you want to do.
If you're developing a SDK, I recommend splitting your implementation in several small libraries(rather than just one big SDL library, there is the SDL core, SDL_Mixer, SDL_Image, etc.)
If you're developing an application, keep a module-based system and a plug-in mechanism.
That way, new features can be added more easily and bloat can be more easily detected.
You may get to a point where you'll add new features some will consider "great" and others "bloat". Otherwise, your application may reach a point that some will call it "feature-poor" and others will call it "just enough".
This isn't an exact quote, but the idea was something like this:
A piece of software is perfect not when there is nothing more to add, but when there is nothing more to remove.
In essence, the simpler and more to-the-point is a software, the better.
To get examples of good software design, take a look at programs that are popular today. Google applications would be a nice place to look. Skype perhaps. Heh, even StackOverflow. :)
If you want intimidating, go to the world of CAD. Check out for example Blender. That's a freeware 3D designer software. Good tool I'm told, but the UI has so many buttons/panels/menus/etc. that it makes baby bunnies cry. Unfortunately I cannot say if this would be a good example of a "bad" UI. 3D designing is a very complex process and all those tools are probably in the right place. But it's definately intimidating. :)
A bad UI design can often be found with propieritary software that comes with propieritary hardware. Unfortunately I cannot give you any examples from the top of my head.
I always tend to design my projects in a way that they're just skeletons which are as extensible as possible. Limiting factors are performance, complexity or Thirdparty-limitations.
This way you could add additional features after finishing the basic structure. A user could also add his needed features.
This probably does not work very good for GUI-applications which should have a good usability without much configuration, but I'm sticking good with this approach for those libs I develop. (They're used by other coders who like to have a highly modifable piece of software)
It's not very hard to develop an application/lib which is bloated with features. But it is to develop an app which could be easily extended by other developers/users to match their own needs.
Develop a wide-ranging plug-in system so you add and take out stuff at any time. Problem solved. If only that was as easy as writing spaghetti code. ;)

Resources