Search Results

Search found 88909 results on 3557 pages for 'inline code'.

Page 14/3557 | < Previous Page | 10 11 12 13 14 15 16 17 18 19 20 21  | Next Page >

  • Inline function v. Macro in C -- What's the Overhead (Memory/Speed)?

    - by Jason R. Mick
    I searched Stack Overflow for the pros/cons of function-like macros v. inline functions. I found the following discussion: Pros and Cons of Different macro function / inline methods in C ...but it didn't answer my primary burning question. Namely, what is the overhead in c of using a macro function (with variables, possibly other function calls) v. an inline function, in terms of memory usage and execution speed? Are there any compiler-dependent differences in overhead? I have both icc and gcc at my disposal. My code snippet I'm modularizing is: double AttractiveTerm = pow(SigmaSquared/RadialDistanceSquared,3); double RepulsiveTerm = AttractiveTerm * AttractiveTerm; EnergyContribution += 4 * Epsilon * (RepulsiveTerm - AttractiveTerm); My reason for turning it into an inline function/macro is so I can drop it into a c file and then conditionally compile other similar, but slightly different functions/macros. e.g.: double AttractiveTerm = pow(SigmaSquared/RadialDistanceSquared,3); double RepulsiveTerm = pow(SigmaSquared/RadialDistanceSquared,9); EnergyContribution += 4 * Epsilon * (RepulsiveTerm - AttractiveTerm); (note the difference in the second line...) This function is a central one to my code and gets called thousands of times per step in my program and my program performs millions of steps. Thus I want to have the LEAST overhead possible, hence why I'm wasting time worrying about the overhead of inlining v. transforming the code into a macro. Based on the prior discussion I already realize other pros/cons (type independence and resulting errors from that) of macros... but what I want to know most, and don't currently know is the PERFORMANCE. I know some of you C veterans will have some great insight for me!!

    Read the article

  • When to do code reviews when doing continuous integration?

    - by SpecialEd
    We are trying to switch to a continuous integration environment but are not sure when to do code reviews. From what I've read of continuous integration, we should be attempting to check in code as often as multiple times a day. I assume, this even means for features that are not yet complete. So the question is, when do we do the code reviews? We can't do it before we check in the code, because that would slow down the process where we will not be able to do daily checkins, let alone multiple checkins per day. Also, if the code we are checking in merely compiles but is not feature complete, doing a code review then is pointless, as most code reviews are best done as the feature is finalized. Does this mean we should do code reviews when a feature is completed, but that unreviewed code will get into the repository?

    Read the article

  • Richmond Code Camp 2010.1

    - by andyleonard
    I can't believe it - Richmond Code Camp 2010.1 is less than two weeks away! Once again, the leadership team has outdone themselves. We have a bunch of great speakers, 9 tracks, 45 sessions - there's something for everyone. If you're going to be in the area and are interested, register today. :{> Share this post: email it! | bookmark it! | digg it! | reddit! | kick it! | live it!...(read more)

    Read the article

  • Clean Code says to avoid protected variables

    - by Matsemann
    I have a question to a statement in Clean Code. I don't fully understand the reasoning to why we should avoid protected variables. It's from the chapter about Formatting, section about Vertical Distance: Concepts that are closely related should be kept vertically close to each other. Clearly this rule doesn't work for concepts that belong in separate files. But then closely related concepts should not be separated into different files unless you have a very good reason. Indeed, this is one of the reasons that protected variables should be avoided.

    Read the article

  • Extended Events Code Generator v1.001 - A Quick Fix

    - by Adam Machanic
    If you're one of the estimated 3-5 people who've downloaded and are using my XE Code Generator , please note that version 1.000 has a small bug: text data (such as query text) larger than 8000 bytes is truncated. I've fixed this issue and am pleased to present version 1.001, attached to this post. Enjoy, and stay tuned for slightly more interesting enhancements! Share this post: email it! | bookmark it! | digg it! | reddit! | kick it! | live it!...(read more)

    Read the article

  • SQL Query Builder/Designer and code Formating

    - by DavRob60
    I write SQL query every now and then, I could easily write them freehand, but sometimes I do create SQL queries using SQL Query Designers for various reason. (I wont start to enumerate them here and/or argue about their usefulness, so let's just say they are sometime useful.) Anyway, I currently use 2 Query Designers : SQL server management studio's Query Designer. Visual Studio 2010's Query Builder (must often within the Table adapter Query Configuration Wizard.) There's something I hate about those two (I don't know about the others), it's the way they throw away my Code formatting of SQL queries after an edit. Is there any way to configure something to automatically reformat the SQL output or is there any external tool/plug-in that I could use to do that job?

    Read the article

  • Presenting at Roanoke Code Camp Saturday!

    - by andyleonard
    Introduction I am honored to once again be selected to present at Roanoke Code Camp ! An Introductory Topic One of my presentations is titled "I See a Control Flow Tab. Now What?" It's a Level 100 talk for those wishing to learn how to build their very first SSIS package. This highly-interactive, demo-intense presentation is for beginners and developers just getting started with SSIS. Attend and learn how to build SSIS packages from the ground up . Designing an SSIS Framework I'm also presenting...(read more)

    Read the article

  • Coding standards in programming?

    - by vicky
    I am an WordPress Plugin Developer. I am not sure how to follow the coding standard while creating a plugin of wordpress. I check with some of the plugins like woocommerce and All in one SEO Plugin in that they are maintaining the proper coding standard. Basically I am Using the NetBeans IDE. Is it possible to make the proper space and coding standards in that IDE. I am Wondering to View his code is very neat and clean. How can i do this or how they are maintaining this. Anyone suggest me to make the wordpress plugin with well coding standards. Thanks, vicky

    Read the article

  • Who owns code if its written by one person with another person directing [on hold]

    - by user136226
    I have an Issue that I need some info on. Basically what Im looking to find out is if I create software,and someone else gives me direction on what they want the software to look like,e.g. an image here,this font of text and it must behave in a certain way. Also some of the code was not developed on my computer and there is no official agreement in place. Not looking to screw anyone over here but need to protect myself if things go sour. Do I own the software or is it jointly owned? Thanks

    Read the article

  • Do you think code is self documenting?

    - by Desolate Planet
    This is a question that was put to me many years ago as a gradute in a job interview and it's kind of picked at my brain now and again and I've never really found a good answer that satisfies me. The interviewer in question was looking for a black and white answer, there was no middle ground. I never got the chance to ask about the rationale behind the question, but I'm curious why that question would be put to a developer and what you would learn from a yes or no answer? From my own point of view, I can read Java, Python, Delphi etc, but if my manager comes up to me and asks me how far along in a project I am and I say "The code is 80% complete" (and before you start shooting me down, I've heard this uttered in a couple of offices by developers), how exactly is that self documenting? Apologies if this question seems strange, but I'd rather ask and get some opinions on it to gain a better understanding of why it would be put to someone in an interview.

    Read the article

  • Single or multiple return statements in a function [on hold]

    - by Juan Carlos Coto
    When writing a function that can have several different return values, particularly when different branches of code return different values, what is the cleanest or sanest way of returning? Please note the following are really contrived examples meant only to illustrate different styles. Example 1: Single return def my_function(): if some_condition: return_value = 1 elif another_condition: return_value = 2 else: return_value = 3 return return_value Example 2: Multiple returns def my_function(): if some_condition: return 1 elif another_condition: return 2 else: return 3 The second example seems simpler and is perhaps more readable. The first one, however, might describe the overall logic a bit better (the conditions affect the assignment of the value, not whether it's returned or not). Is the second way preferable to the first? Why?

    Read the article

  • Shouldn't we count characters of code and comments instead of lines of code and comments? [closed]

    - by Gabriel
    Counting lines of code and comments is sometimes bogus, since most of what we write may be written in one or more lines, depending column count limitations, screen size, style and so forth. Since the commonly used languages (say C, C++, C# and Java) are free-form, wouldn't it be more clever to count characters instead? Edit: I'm not considering LOC-oriented programming where coders try to artificially match requirements by adding irrelevant comments or using multiple lines where less would be enough (or the opposite). I'm interested in better metrics that would be independent of coding style, to be used by honest programmers.

    Read the article

  • Do you think code is self documenting?

    - by Desolate Planet
    This is a question that was put to me many years ago as a gradute in a job interview and it's nagged at my brain now and again and I've never really found a good answer that satisfied me. The interviewer in question was looking for a black and white answer, there was no middle ground. I never got the chance to ask about the rationale behind the question, but I'm curious why that question would be put to a developer and what you would learn from a yes or no answer? From my own point of view, I can read Java, Python, Delphi etc, but if my manager comes up to me and asks me how far along in a project I am and I say "The code is 80% complete" (and before you start shooting me down, I've heard this uttered in a couple of offices by developers), how exactly is that self documenting? Apologies if this question seems strange, but I'd rather ask and get some opinions on it to gain a better understanding of why it would be put to someone in an interview.

    Read the article

  • Better control on code updates

    - by yes123
    I will briefly explain my situation. I have a website in PHP, this website is powered by a custom framework + some "plug-in" made ad hoc for it. I am the only developer of this. Until now I just test locally any changes than I upload the php files via FTP. I don't feel confortable anymore with this. The code base has grown quite a lot and I need some sort of system that helps to keep track of changes (line by line) and can restore to an old version easly if something goes wrong. Are there any good solution for this? Note: I never used something like version control or subversion because I think they are too much for this situation (I am the only developer and I just need basic feature) Note2: Something with a nice web interface would be perfect, I can pay for a good service too As now I found: http://beanstalkapp.com/ http://github.com/ http://www.codespaces.com/ http://codesion.com/ https://bitbucket.org/

    Read the article

  • Peer code review for full application

    - by bswinnerton
    My sincerest apologies if this is the wrong place to post something like this, but this seemed like the best fit. I was wondering if there are any websites or resources for a full site peer code review. I'm new to Ruby specifically and want to make sure that my logic is following the overall best standards. I've pieced together multiple different tutorials and I feel like while my understanding is getting better, it'd be great if the overall structure of such an application could be critiqued, and for someone that doesn't really know another Ruby developer - I find that I've run into a roadblock and don't want to develop bad habits now.

    Read the article

  • How to prevent code from leaking outside work?

    - by AeroCross
    I'm working on an institution that has a really strong sense of "possession" - each line of software we write should be only ours. Ironically, I'm the only programmer (ATM), but we're planning in hiring others. Since my bosses wouldn't count the new programmers as people they can trust, they have an issue with the copies of the source code. We use Git, so they would have a entire copy of each of the projects they work on, when they clone the repository. We can restrict access to them to a single key with Gitolite and bind that to their PC's, but they can copy those keys to another computer and they would have the repository access in another PC. Also (and the most obvious method) they could just upload the files somewhere else, add another remote, or just copy the files to an USB drive. Is there any (perhaps clever) way to prevent events like these?

    Read the article

  • Pluggable Rules for Entity Framework Code First

    - by Ricardo Peres
    Suppose you want a system that lets you plug custom validation rules on your Entity Framework context. The rules would control whether an entity can be saved, updated or deleted, and would be implemented in plain .NET. Yes, I know I already talked about plugable validation in Entity Framework Code First, but this is a different approach. An example API is in order, first, a ruleset, which will hold the collection of rules: 1: public interface IRuleset : IDisposable 2: { 3: void AddRule<T>(IRule<T> rule); 4: IEnumerable<IRule<T>> GetRules<T>(); 5: } Next, a rule: 1: public interface IRule<T> 2: { 3: Boolean CanSave(T entity, DbContext ctx); 4: Boolean CanUpdate(T entity, DbContext ctx); 5: Boolean CanDelete(T entity, DbContext ctx); 6: String Name 7: { 8: get; 9: } 10: } Let’s analyze what we have, starting with the ruleset: Only has methods for adding a rule, specific to an entity type, and to list all rules of this entity type; By implementing IDisposable, we allow it to be cancelled, by disposing of it when we no longer want its rules to be applied. A rule, on the other hand: Has discrete methods for checking if a given entity can be saved, updated or deleted, which receive as parameters the entity itself and a pointer to the DbContext to which the ruleset was applied; Has a name property for helping us identifying what failed. A ruleset really doesn’t need a public implementation, all we need is its interface. The private (internal) implementation might look like this: 1: sealed class Ruleset : IRuleset 2: { 3: private readonly IDictionary<Type, HashSet<Object>> rules = new Dictionary<Type, HashSet<Object>>(); 4: private ObjectContext octx = null; 5:  6: internal Ruleset(ObjectContext octx) 7: { 8: this.octx = octx; 9: } 10:  11: public void AddRule<T>(IRule<T> rule) 12: { 13: if (this.rules.ContainsKey(typeof(T)) == false) 14: { 15: this.rules[typeof(T)] = new HashSet<Object>(); 16: } 17:  18: this.rules[typeof(T)].Add(rule); 19: } 20:  21: public IEnumerable<IRule<T>> GetRules<T>() 22: { 23: if (this.rules.ContainsKey(typeof(T)) == true) 24: { 25: foreach (IRule<T> rule in this.rules[typeof(T)]) 26: { 27: yield return (rule); 28: } 29: } 30: } 31:  32: public void Dispose() 33: { 34: this.octx.SavingChanges -= RulesExtensions.OnSaving; 35: RulesExtensions.rulesets.Remove(this.octx); 36: this.octx = null; 37:  38: this.rules.Clear(); 39: } 40: } Basically, this implementation: Stores the ObjectContext of the DbContext to which it was created for, this is so that later we can remove the association; Has a collection - a set, actually, which does not allow duplication - of rules indexed by the real Type of an entity (because of proxying, an entity may be of a type that inherits from the class that we declared); Has generic methods for adding and enumerating rules of a given type; Has a Dispose method for cancelling the enforcement of the rules. A (really dumb) rule applied to Product might look like this: 1: class ProductRule : IRule<Product> 2: { 3: #region IRule<Product> Members 4:  5: public String Name 6: { 7: get 8: { 9: return ("Rule 1"); 10: } 11: } 12:  13: public Boolean CanSave(Product entity, DbContext ctx) 14: { 15: return (entity.Price > 10000); 16: } 17:  18: public Boolean CanUpdate(Product entity, DbContext ctx) 19: { 20: return (true); 21: } 22:  23: public Boolean CanDelete(Product entity, DbContext ctx) 24: { 25: return (true); 26: } 27:  28: #endregion 29: } The DbContext is there because we may need to check something else in the database before deciding whether to allow an operation or not. And here’s how to apply this mechanism to any DbContext, without requiring the usage of a subclass, by means of an extension method: 1: public static class RulesExtensions 2: { 3: private static readonly MethodInfo getRulesMethod = typeof(IRuleset).GetMethod("GetRules"); 4: internal static readonly IDictionary<ObjectContext, Tuple<IRuleset, DbContext>> rulesets = new Dictionary<ObjectContext, Tuple<IRuleset, DbContext>>(); 5:  6: private static Type GetRealType(Object entity) 7: { 8: return (entity.GetType().Assembly.IsDynamic == true ? entity.GetType().BaseType : entity.GetType()); 9: } 10:  11: internal static void OnSaving(Object sender, EventArgs e) 12: { 13: ObjectContext octx = sender as ObjectContext; 14: IRuleset ruleset = rulesets[octx].Item1; 15: DbContext ctx = rulesets[octx].Item2; 16:  17: foreach (ObjectStateEntry entry in octx.ObjectStateManager.GetObjectStateEntries(EntityState.Added)) 18: { 19: Object entity = entry.Entity; 20: Type realType = GetRealType(entity); 21:  22: foreach (dynamic rule in (getRulesMethod.MakeGenericMethod(realType).Invoke(ruleset, null) as IEnumerable)) 23: { 24: if (rule.CanSave(entity, ctx) == false) 25: { 26: throw (new Exception(String.Format("Cannot save entity {0} due to rule {1}", entity, rule.Name))); 27: } 28: } 29: } 30:  31: foreach (ObjectStateEntry entry in octx.ObjectStateManager.GetObjectStateEntries(EntityState.Deleted)) 32: { 33: Object entity = entry.Entity; 34: Type realType = GetRealType(entity); 35:  36: foreach (dynamic rule in (getRulesMethod.MakeGenericMethod(realType).Invoke(ruleset, null) as IEnumerable)) 37: { 38: if (rule.CanDelete(entity, ctx) == false) 39: { 40: throw (new Exception(String.Format("Cannot delete entity {0} due to rule {1}", entity, rule.Name))); 41: } 42: } 43: } 44:  45: foreach (ObjectStateEntry entry in octx.ObjectStateManager.GetObjectStateEntries(EntityState.Modified)) 46: { 47: Object entity = entry.Entity; 48: Type realType = GetRealType(entity); 49:  50: foreach (dynamic rule in (getRulesMethod.MakeGenericMethod(realType).Invoke(ruleset, null) as IEnumerable)) 51: { 52: if (rule.CanUpdate(entity, ctx) == false) 53: { 54: throw (new Exception(String.Format("Cannot update entity {0} due to rule {1}", entity, rule.Name))); 55: } 56: } 57: } 58: } 59:  60: public static IRuleset CreateRuleset(this DbContext context) 61: { 62: Tuple<IRuleset, DbContext> ruleset = null; 63: ObjectContext octx = (context as IObjectContextAdapter).ObjectContext; 64:  65: if (rulesets.TryGetValue(octx, out ruleset) == false) 66: { 67: ruleset = rulesets[octx] = new Tuple<IRuleset, DbContext>(new Ruleset(octx), context); 68: 69: octx.SavingChanges += OnSaving; 70: } 71:  72: return (ruleset.Item1); 73: } 74: } It relies on the SavingChanges event of the ObjectContext to intercept the saving operations before they are actually issued. Yes, it uses a bit of dynamic magic! Very handy, by the way! So, let’s put it all together: 1: using (MyContext ctx = new MyContext()) 2: { 3: IRuleset rules = ctx.CreateRuleset(); 4: rules.AddRule(new ProductRule()); 5:  6: ctx.Products.Add(new Product() { Name = "xyz", Price = 50000 }); 7:  8: ctx.SaveChanges(); //an exception is fired here 9:  10: //when we no longer need to apply the rules 11: rules.Dispose(); 12: } Feel free to use it and extend it any way you like, and do give me your feedback! As a final note, this can be easily changed to support plain old Entity Framework (not Code First, that is), if that is what you are using.

    Read the article

  • Gathering all data in single iteration vs using functions for readable code

    - by user828584
    Say I have an array of runners with which I need to find the tallest runner, the fastest runner, and the lightest runner. It seems like the most readable solution would be: runners = getRunners(); tallestRunner = getTallestRunner(runners); fastestRunner = getFastestRunner(runners); lightestRunner = getLightestRunner(runners); ..where each function iterates over the runners and keeps track of the largest height, greatest speed, and lowest weight. Iterating over the array three times, however, doesn't seem like a very good idea. It would instead be better to do: int greatestHeght, greatestSpeed, leastWeight; Runner tallestRunner, fastestRunner, lightestRunner; for(runner in runners){ if(runner.height > greatestHeight) { greatestHeight = runner.height; tallestRunner = runner; } if(runner.speed > ... } While this isn't too unreadable, it can get messy when there is more logic for each piece of information being extracted in the iteration. What's the middle ground here? How can I use only a single iteration while still keeping the code divided into logical units?

    Read the article

  • Formatting php, what works more efficiently?

    - by JamesM-SiteGen
    Hello fellow programmers, I was just wondering what makes php work faster, I have a few methods that I always go and do, but that only improves the way I can read it, but how about the interpreter? Should I include the curly braces when there is only one statement to run? if(...){ echo "test"; } # Or.. if(...) echo "test"; === Which should be used? I have also found http://beta.phpformatter.com/ and I find the following settings to be good, but are they? Indentation: Indentation style: {K&R (One true brace style)} Indent with: {Tabs} Starting indentation: [1] Indentation: [1] Common: [x] Remove all comments [x] Remove empty lines [x] Align assignments statements nicely [ ] Put a comment with the condition after if, while, for, foreach, declare and catch statements Improvement: [x] Remove lines with just a semicolon (;) [x] Make normal comments (//) from perl comments (#) [x] Make long opening tag (<?php) from short one (<?) Brackets: [x] Space inside brackets- ( ) [x] Space inside empty brackets- ( ) [x] Space inside block brackets- [ ] [x] Space inside empty block brackets- [ ] Tiny var names: often I go through my code and change $var1 to $a, $var2 to $b and so on. I do include comments at the start of the file to show to me what each letter(s) mean.. Final note: So am I doing the right thing with the curly braces and the settings? Are there any great tips that help it run faster?

    Read the article

  • Duplication of code (backend and javascript - knockout)

    - by Michal B.
    We have a new developer in our team. He seems a smart guy (he just came in so I cannot really judge). He started with implementing some small enhancements in the project (MVC3 web application using javascript - jquery and knockout). Let's say we have two values: A - quite complex calculation C - constant B = A + C On the screen there is value B and user can change it (normal texbox). When B changes, A changes as well because C is constant. So there is linear dependency between A and B. Now, all the calculations are done in the backend, but we need to recalculate A as user changes B (in js, I would use knockout). I thought about storing old A and B and when B changes by 10 then we know that new A will be old A + 10. He says this is dirty, because it's duplication of code (we make use of the fact that they are dependent and according to him that should be only in one place in our app). I understand it's not ideal, but making AJAX request after every key press seems a bit too much. It's a really small thing and I would not post if we haven't had long discussion about it. How do you deal with such problems? Also I can imagine that using knockout implies lots of calculations on the client side, which very often leads to duplication of the same calculations from the backend. Does anyone have links to some articles/thoughts on this topic?

    Read the article

  • Augmenting functionality of subclasses without code duplication in C++

    - by Rob W
    I have to add common functionality to some classes that share the same superclass, preferably without bloating the superclass. The simplified inheritance chain looks like this: Element -> HTMLElement -> HTMLAnchorElement Element -> SVGElement -> SVGAlement The default doSomething() method on Element is no-op by default, but there are some subclasses that need an actual implementation that requires some extra overridden methods and instance members. I cannot put a full implementation of doSomething() in Element because 1) it is only relevant for some of the subclasses, 2) its implementation has a performance impact and 3) it depends on a method that could be overridden by a class in the inheritance chain between the superclass and a subclass, e.g. SVGElement in my example. Especially because of the third point, I wanted to solve the problem using a template class, as follows (it is a kind of decorator for classes): struct Element { virtual void doSomething() {} }; // T should be an instance of Element template<class T> struct AugmentedElement : public T { // doSomething is expensive and uses T virtual void doSomething() override {} // Used by doSomething virtual bool shouldDoSomething() = 0; }; class SVGElement : public Element { /* ... */ }; class SVGAElement : public AugmentedElement<SVGElement> { // some non-trivial check bool shouldDoSomething() { /* ... */ return true; } }; // Similarly for HTMLAElement and others I looked around (in the existing (huge) codebase and on the internet), but didn't find any similar code snippets, let alone an evaluation of the effectiveness and pitfalls of this approach. Is my design the right way to go, or is there a better way to add common functionality to some subclasses of a given superclass?

    Read the article

  • Should accessible members of an internal class be internal too?

    - by Jeff Mercado
    I'm designing a set of APIs for some applications I'm working on. I want to keep the code style consistent in all the classes I write but I've found that there are a few inconsistencies that I'm introducing and I don't know what the best way to resolve them is. My example here is specific to C# but this would apply to any language with similar mechanisms. There are a few classes that I need for implementation purposes that I don't necessarily want to expose in the API so I make them internal whereever needed. Generally what I would do is design the class as I normally would (e.g., make members public/protected/private where necessary) and change the visibility level of the class itself to internal. So I might have a few classes that look like this: internal interface IMyItem { ItemSet AddTo(ItemSet set); } internal class _SmallItem : IMyItem { private readonly /* parameters */; public _SmallItem(/* small item parameters */) { /* ... */ } public ItemSet AddTo(ItemSet set) { /* ... */ } } internal abstract class _CompositeItem: IMyItem { private readonly /* parameters */; public _CompositeItem(/* composite item parameters */) { /* ... */ } public abstract object UsefulInformation { get; } protected void HelperMethod(/* parameters */) { /* ... */ } } internal class _BigItem : _CompositeItem { private readonly /* parameters */; public _BigItem(/* big item parameters */) { /* ... */ } public override object UsefulInformation { get { /* ... */ } } public ItemSet AddTo(ItemSet set) { /* ... */ } } In another generated class (part of a parser/scanner), there is a structure that contains fields for all possible values it can represent. The class generated is internal too but I have control over the visibility of the members and decided to make them internal as well. internal partial struct ValueType { internal string String; internal ItemSet ItemSet; internal IMyItem MyItem; } internal class TokenValue { internal static int EQ(ItemSetScanner scanner) { /* ... */ } internal static int NAME(ItemSetScanner scanner, string value) { /* ... */ } internal static int VALUE(ItemSetScanner scanner, string value) { /* ... */ } //... } To me, this feels odd because the first set of classes, I didn't necessarily have to make some members public, they very well could have been made internal. internal members of an internal type can only be accessed internally anyway so why make them public? I just don't like the idea that the way I write my classes has to change drastically (i.e., change all uses of public to internal) just because the class is internal. Any thoughts on what I should do here? It makes sense to me that I might want to make some members of a class declared public, internal. But it's less clear to me when the class is declared internal.

    Read the article

  • What is this code?

    - by Aerovistae
    This is from the Evolution of a Programmer "joke", at the "Master Programmer" level. It seems to be C++, but I don't know what all this bloated extra stuff is, nor did any Google searches turn up anything except the joke I took it from. Can anyone tell me more about what I'm reading here? [ uuid(2573F8F4-CFEE-101A-9A9F-00AA00342820) ] library LHello { // bring in the master library importlib("actimp.tlb"); importlib("actexp.tlb"); // bring in my interfaces #include "pshlo.idl" [ uuid(2573F8F5-CFEE-101A-9A9F-00AA00342820) ] cotype THello { interface IHello; interface IPersistFile; }; }; [ exe, uuid(2573F890-CFEE-101A-9A9F-00AA00342820) ] module CHelloLib { // some code related header files importheader(<windows.h>); importheader(<ole2.h>); importheader(<except.hxx>); importheader("pshlo.h"); importheader("shlo.hxx"); importheader("mycls.hxx"); // needed typelibs importlib("actimp.tlb"); importlib("actexp.tlb"); importlib("thlo.tlb"); [ uuid(2573F891-CFEE-101A-9A9F-00AA00342820), aggregatable ] coclass CHello { cotype THello; }; }; #include "ipfix.hxx" extern HANDLE hEvent; class CHello : public CHelloBase { public: IPFIX(CLSID_CHello); CHello(IUnknown *pUnk); ~CHello(); HRESULT __stdcall PrintSz(LPWSTR pwszString); private: static int cObjRef; }; #include <windows.h> #include <ole2.h> #include <stdio.h> #include <stdlib.h> #include "thlo.h" #include "pshlo.h" #include "shlo.hxx" #include "mycls.hxx" int CHello:cObjRef = 0; CHello::CHello(IUnknown *pUnk) : CHelloBase(pUnk) { cObjRef++; return; } HRESULT __stdcall CHello::PrintSz(LPWSTR pwszString) { printf("%ws\n", pwszString); return(ResultFromScode(S_OK)); } CHello::~CHello(void) { // when the object count goes to zero, stop the server cObjRef--; if( cObjRef == 0 ) PulseEvent(hEvent); return; } #include <windows.h> #include <ole2.h> #include "pshlo.h" #include "shlo.hxx" #include "mycls.hxx" HANDLE hEvent; int _cdecl main( int argc, char * argv[] ) { ULONG ulRef; DWORD dwRegistration; CHelloCF *pCF = new CHelloCF(); hEvent = CreateEvent(NULL, FALSE, FALSE, NULL); // Initialize the OLE libraries CoInitiali, NULL); // Initialize the OLE libraries CoInitializeEx(NULL, COINIT_MULTITHREADED); CoRegisterClassObject(CLSID_CHello, pCF, CLSCTX_LOCAL_SERVER, REGCLS_MULTIPLEUSE, &dwRegistration); // wait on an event to stop WaitForSingleObject(hEvent, INFINITE); // revoke and release the class object CoRevokeClassObject(dwRegistration); ulRef = pCF->Release(); // Tell OLE we are going away. CoUninitialize(); return(0); } extern CLSID CLSID_CHello; extern UUID LIBID_CHelloLib; CLSID CLSID_CHello = { /* 2573F891-CFEE-101A-9A9F-00AA00342820 */ 0x2573F891, 0xCFEE, 0x101A, { 0x9A, 0x9F, 0x00, 0xAA, 0x00, 0x34, 0x28, 0x20 } }; UUID LIBID_CHelloLib = { /* 2573F890-CFEE-101A-9A9F-00AA00342820 */ 0x2573F890, 0xCFEE, 0x101A, { 0x9A, 0x9F, 0x00, 0xAA, 0x00, 0x34, 0x28, 0x20 } }; #include <windows.h> #include <ole2.h> #include <stdlib.h> #include <string.h> #include <stdio.h> #include "pshlo.h" #include "shlo.hxx" #include "clsid.h" int _cdecl main( int argc, char * argv[] ) { HRESULT hRslt; IHello *pHello; ULONG ulCnt; IMoniker * pmk; WCHAR wcsT[_MAX_PATH]; WCHAR wcsPath[2 * _MAX_PATH]; // get object path wcsPath[0] = '\0'; wcsT[0] = '\0'; if( argc > 1) { mbstowcs(wcsPath, argv[1], strlen(argv[1]) + 1); wcsupr(wcsPath); } else { fprintf(stderr, "Object path must be specified\n"); return(1); } // get print string if(argc > 2) mbstowcs(wcsT, argv[2], strlen(argv[2]) + 1); else wcscpy(wcsT, L"Hello World"); printf("Linking to object %ws\n", wcsPath); printf("Text String %ws\n", wcsT); // Initialize the OLE libraries hRslt = CoInitializeEx(NULL, COINIT_MULTITHREADED); if(SUCCEEDED(hRslt)) { hRslt = CreateFileMoniker(wcsPath, &pmk); if(SUCCEEDED(hRslt)) hRslt = BindMoniker(pmk, 0, IID_IHello, (void **)&pHello); if(SUCCEEDED(hRslt)) { // print a string out pHello->PrintSz(wcsT); Sleep(2000); ulCnt = pHello->Release(); } else printf("Failure to connect, status: %lx", hRslt); // Tell OLE we are going away. CoUninitialize(); } return(0); }

    Read the article

  • Django Inline formset for editing multiple related records at once - the right way to go?

    - by Bert
    Hi, When using inline formsets, how does one do paging? I'm using django 1.1. The situation I'm in, is that the user needs to be able to edit the related objects quickly and easily (which is why I think I should be using an inline formset). However, there can be a more than a hundred objects to edit, which makes a pretty large formset, so paging would make sense. Is there a better way to be doing this? Thanks Bert

    Read the article

< Previous Page | 10 11 12 13 14 15 16 17 18 19 20 21  | Next Page >