|
I don't care for patterns in my code and I definitely don't care for vague terms like "clean" code.
I only care for code that's short, safe to fail, easy to test, and understandable by a college freshman in less than 2 weeks.
Nevertheless, please write about the patterns you think are worthwhile; preferably with a not-too-generic example. You might teach us something worthwhile.
Also, if anyone knows of a must-read regarding the use of factory-pattern, please share. I really hate it and I'm looking for material to nuance my feelings towards it.
|
|
|
|
|
killbot5000 wrote: understandable by a college freshman in less than 2 weeks. If you can't explain your pattern in 5 minutes time, then you probably don't understand it. That is assuming that your freshman knows how to code
killbot5000 wrote: Also, if anyone knows of a must-read regarding the use of factory-pattern, please share. That's the pattern I'm opening the article with. May take another day or two.
killbot5000 wrote: I really hate it Yah, then you had the wrong explanation. Most important question to answer is always "what's in it for me?", and a factory has an easy to explain benefit
Bastard Programmer from Hell
If you can't read my code, try converting it here[^]
|
|
|
|
|
If you can't explain your pattern in 5 minutes time, then you probably don't understand it. True enough. I was only trying to explain our somewhat informal guideline at work; a junior should be able to master any one of our projects in a maximum of 2 weeks time, without any help from a senior. In general, we don't trust juniors to correctly identify or implement various patterns, so we don't approach our code as such.
But, of course, we do want juniors to be able to work for clients almost instantly. *grin*
It's mostly an effort of gathering/vetting the right tools; ones that are easy to use, easy to test, and with a low code footprint. Around 90% of our R&D is spend on simplifying the way we do things.
But, to come back to factory pattern:
I hate it because it often seems the right choice, but, for me, it has never panned out. Every time I try it out, I end up over-designing my data models and adding an abstraction layer I don't really need. In the end, I usually remove it and end up with (almost embarrassingly..) simple classes with distinct data.
It can be infuriating.
I try to come up with a clever looking data structure, hope factory pattern fits the bill, and *boom* suddenly I see there's no need for abstraction whatsoever, vastly reducing my test code and simplifying my DB design.
|
|
|
|
|
killbot5000 wrote: Every time I try it out, I end up over-designing my data models and adding an abstraction layer I don't really need. Might be easier to not try it until you need one. To find a scenario where the use of it would be beneficial, you'd need some decent example. The abstract examples didn't help me much.
killbot5000 wrote: *boom* suddenly I see there's no need for abstraction whatsoever, vastly reducing my test code and simplifying my DB design. KISS is always the best choice. Someone else will inderstand it better when reading, updates are easier and less complicated, and less code in general means fewer bugs.
I'll give you a ping when I'm done and posted it
Bastard Programmer from Hell
If you can't read my code, try converting it here[^]
|
|
|
|
|
As promised[^], with a not-too-generic example
Bastard Programmer from Hell
If you can't read my code, try converting it here[^]
|
|
|
|
|
I was really wondering why you did not want to use a pattern so badly you would strike it from your code.
I bought the GoF book, and was excited to be able to communicate better designs quicker. But, like you, I found zealots who used patterns like an indentation style. The could not code without them, and would convolute a simple program with them.
I moved away from them, but use patterns (factory, most often) where they are clearly a solid approach, but NEVER worrying about doing it exactly as prescribed.
But your comment felt like "I refuse to indent my code because there are indentation extremists"...
|
|
|
|
|
The only pattern worth using is GIWAGO - 'Get It Working And Get Out!'
veni bibi saltavi
|
|
|
|
|
It is the Tinder pattern.
(A collection of singletons)
|
|
|
|
|
Swipe left! Swipe left!
Bad command or file name. Bad, bad command! Sit! Stay! Staaaay...
AntiTwitter: @DalekDave is now a follower!
|
|
|
|
|
Boom! You sir, have won the Internet today.
Jeremy Falcon
|
|
|
|
|
Duncan Edwards Jones wrote: the Tinder pattern That's a very tricky one since you have to implement your own MultiTassking, and MultiBedding.
«While I complain of being able to see only a shadow of the past, I may be insensitive to reality as it is now, since I'm not at a stage of development where I'm capable of seeing it.» Claude Levi-Strauss (Tristes Tropiques, 1955)
|
|
|
|
|
Yes, it's called the "static cling" pattern as nothing put into the dictionary ever gets garbage collected.
|
|
|
|
|
Global variables of any flavor have that problem, that's why I would only do something like that for things that remain mostly constant throughout the program's lifetime.
Remember the Win32 hInstance parameter? It was passed to WinMain() as a parameter and remained unchanged from then on and you needed it for many Win32 API calls. The four bytes for a global variable were not much of a problem (yes, it was memory that was not cleaned up until the program ended) and it sure beats passing around that value to every window and dialog.
When memory hogging is under control, state is not problematic (or even constant), and when I need it all over the place, I will not go through the trouble of routing that object or variable to the remotest parts of the application.
I have lived with several Zen masters - all of them were cats.
|
|
|
|
|
Marc Clifton wrote: nothing put into the dictionary ever gets garbage collected. Is there no way to handle that >?
«While I complain of being able to see only a shadow of the past, I may be insensitive to reality as it is now, since I'm not at a stage of development where I'm capable of seeing it.» Claude Levi-Strauss (Tristes Tropiques, 1955)
|
|
|
|
|
OriginalGriff wrote: So I have a private constructor, a static Dictionary containing all created instances, and a static method which fetches the instance:
I only see a downside
you create an "instance" and fill it with data // no problems
...
later fetch your "instanse" and do the work // ...wow, that went pretty quick
all you've done is taken away the ability of the code editor/compiler to check the item exists and built in another opportunity for a hard to find run-time problem. (if you missed it check the spelling, because the compiler wont do that for you.)
and the upside isn't there, a separately declared list left empty takes few resoures.
Installing Signature...
Do not switch off your computer.
|
|
|
|
|
OriginalGriff wrote: I hope not, I might have to stop using it ... Bullshit!
Design patterns were never intended to be the last word and the answer for everything. There is no law against coming up with your very own way of doing something, no matter what some gurus want to have you believe.
The SOLID faction would condemn you for breaching the single responsibility principle and using something akin to global variables and singletons. So what?
My personal sin is a message broker that I use in almost any UI. When compared to awkward wasteful automatic routing in a large UI tree or fragile code bloating manual routing, having a message broker as a singleton is a blessing. Some object registers with the broker to be notified when a specific message is sent, another object sends the message at some time and the broker looks up all subscribers and calls their event handlers. This is reasonably fast (no searching the entire UI tree) and absolutely unproblematic. The only weak spot is that an object better unregister its subscriptions when it is being destroyed, otherwise the broker may get bogged down servicing dead subscriptions.
Yes, global variables or singletons may be problematic, but sometimes you can have all of the benefits without any complications. I would immediately redesign if any complications materialized, but not simply for the purity of some design philosophy's sake. Why throw away something that has proven itself to be useful often enough?
But that's just me. I have always been conservative with redesigning and been building libraries, even when that meant to store machine code routines on cassette tapes and manually relocating them into a new program.
I have lived with several Zen masters - all of them were cats.
|
|
|
|
|
A behavioral pattern?
Someone's therapist knows all about you!
|
|
|
|
|
Interesting, Griff, I've played with some similar patterns. It probably was not your intent to show all of your code, but your mention of "heavy resources" made me wonder if you implement IDisposable, or the newer CLassName~ finalizer thingee.
«While I complain of being able to see only a shadow of the past, I may be insensitive to reality as it is now, since I'm not at a stage of development where I'm capable of seeing it.» Claude Levi-Strauss (Tristes Tropiques, 1955)
|
|
|
|
|
IDisposable - and needless to say there are ways to remove items from the all collection to prevent things gumming up the garbage collector.
Bad command or file name. Bad, bad command! Sit! Stay! Staaaay...
AntiTwitter: @DalekDave is now a follower!
|
|
|
|
|
Seems like an Appalachian version of DI, just missing a banjo duel
But fortunately we have the nanny-state politicians who can step in to protect us poor stupid consumers, most of whom would not know a JVM from a frozen chicken. Bruce Pierson Because programming is an art, not a science. Marc Clifton I gave up when I couldn't spell "egg". Justine Allen
|
|
|
|
|
Hmm. I don't see anything wrong with it, other than the name "all " for the container, which isn't terribly descriptive.
I've used similar "patterns", where constructors and destructors maintain global constructs as a side effect. The global constructs usually simplify finding one of the instances in some way, or in accessing the entire collection of 'live' instances. In some cases I also use them for orderly shutdowns, to insure that all instances get destroyed properly.
For what it's worth, I've never read the GoF Patterns book.
Software Zen: delete this;
|
|
|
|
|
Shoulda used a concurrent dictionary...
"(I) am amazed to see myself here rather than there ... now rather than then".
― Blaise Pascal
|
|
|
|
|
OriginalGriff wrote: And it works really well. But ... is that a pattern?
Yes. Multiton. Been around a long time.
Multiton pattern - Wikipedia[^]
The general idea is a Singleton per 'context' where the definition of the context is up the implementation.
|
|
|
|
|
A cunning way of asking a programming question in the Lounge!
The all field should be readonly , since you never replace it.
You should probably specify an explicit StringComparer for the dictionary, to make it more obvious that the key is case-sensitive.
I'd be inclined to move the Add to the static method, and leave the constructor alone. I'd also replace the ContainsKey / indexer pair with a single TryGetValue call:
public static MyClass Get(string name, List<string> data)
{
if (!all.TryGetValue(name, out MyClass instance)
{
instance = new MyClass(name, data);
all.Add(name, instance);
}
return instance;
}
And as Gerry said, if there's any possibility of the method being called by multiple threads, use a ConcurrentDictionary[^] instead:
private static readonly ConcurrentDictionary<string, MyClass> all = new ConcurrentDictionary<string, MyClass>(StringComparer.Ordinal);
public static MyClass Get(string name, List<string> data)
{
return all.GetOrAdd(name, key => new MyClass(key, data));
}
Let's hope your class doesn't contain any unmanaged or disposable resources, since you'd have no way of knowing when to clean them up.
"These people looked deep within my soul and assigned me a number based on the order in which I joined."
- Homer
|
|
|
|
|
What are the virtues of a virtual machine?
... such stuff as dreams are made on
|
|
|
|
|