Search Results

Search found 10773 results on 431 pages for 'concurrency testing'.

Page 11/431 | < Previous Page | 7 8 9 10 11 12 13 14 15 16 17 18  | Next Page >

  • How to handle data concurrency in ASP.NET?

    - by Itsgkiran
    Hi! all I have an application, that is accessing by number of users at the same time. Those users, who are accessing the application getting the same id. Here what i am doing in the code is, when they are creating new user i am getting a max id from DB and increasing the value to 1. So that they are getting same ID. so that i am facing concurrency in this situation. How to solve this problem. I need to display different numbers when the users click on NewUser. I am using SQL server 2008 and .NET 3.5 and C#.NET Thanks in advance.

    Read the article

  • How to handle concurrency control in dynamic data?

    - by Andrew
    I've been quite impressed with dynamic data and how easy and quick it is to get a simple site up and running. I'm planning on using it for a simple internal HR admin site for registering people's skills/degrees/etc. I've been watching the intro videos at www.asp.net/dynamicdata and one thing they never mention is how to handle concurrency control. It seems that DD does not handle it right out of the box (unless there is some setting I haven't seen) as I manually generated a change conflict exception and the app failed without any user friendly message. Anybody know if DD handles it out of the box? Or do you have to somehow build it into the site?

    Read the article

  • Understanding Clojure concurrency example

    - by dusha
    Hello, I just go through various documentation on Clojure concurrency and came accross the example on the website (http://clojure.org/concurrent_programming). (import '(java.util.concurrent Executors)) (defn test-stm [nitems nthreads niters] (let [refs (map ref (replicate nitems 0)) pool (Executors/newFixedThreadPool nthreads) tasks (map (fn [t] (fn [] (dotimes [n niters] (dosync (doseq [r refs] (alter r + 1 t)))))) (range nthreads))] (doseq [future (.invokeAll pool tasks)] (.get future)) (.shutdown pool) (map deref refs))) I understand what it does and how it works, but I don't get why the second anonymous function fn[] is needed? Many thanks, dusha. P.S. Without this second fn [] I get NullPointerException.

    Read the article

  • How fast are my services? Comparing basicHttpBinding and ws2007HttpBinding using the SO-Aware Test Workbench

    - by gsusx
    When working on real world WCF solutions, we become pretty aware of the performance implications of the binding and behavior configuration of WCF services. However, whether it’s a known fact the different binding and behavior configurations have direct reflections on the performance of WCF services, developers often struggle to figure out the real performance behavior of the services. We can attribute this to the lack of tools for correctly testing the performance characteristics of WCF services...(read more)

    Read the article

  • SO-Aware at the Atlanta Connected Systems User Group

    - by gsusx
    Today my colleague Don Demsak will be presenting a session about WCF management, testing and governance using SO-Aware and the SO-Aware Test Workbench at the Connected Systems User Group in Atlanta . Don is a very engaging speaker and has prepared some very cool demos based on lessons of real world WCF solutions. If you are in the ATL area and interested in WCF, AppFabric, BizTalk you should definitely swing by Don’s session . Don’t forget to heckle him a bit (you can blame it for it ;) )...(read more)

    Read the article

  • Does concurrency inherently introduce "randomness" into a game?

    - by Jeff
    When a game is implemented with concurrency (as most games are), does this necessarily, by its very nature, introduce an element of randomness into the game that is outside of the players' control? Note that when I use the word "random", I'm not meaning to launch into a philosophical debate about the deterministic nature of the system. I understand that concurrency is deterministic in the sense that the operating system decides which processes to allow time on the CPU and in what order (or the JVM controls which Thread's turn it is to execute, etc). But my understanding of this is that there is no way to control or predict whether one thread's next command will execute before or after another. The reason I'm asking is because this seems like a fundamental difficulty for game development where a game is supposedly designed around a player's skill. Consider a game like League of Legends. Assume that two players are battling it out. It's a very close contest between the two and it's coming down to the wire -- so much so that whoever gets their last attack off will be the one to kill the other and win the game for their team. If the players are implemented using concurrency and the situation really was like this, is it essentially out of the players' hands at this point? Is the outcome of this match all up to whatever system is arbitrarily deciding which player's thread/process will execute next? If not, what am I misunderstanding about concurrency? If so, is there any way around this problem so that a game of skill can always be a game of skill, especially in those most crucial moments?

    Read the article

  • When you should and should not use the 'new' keyword?

    - by skizeey
    I watched a Google Tech Talk presentation on Unit Testing, given by Misko Hevery, and he said to avoid using the new keyword in business logic code. I wrote a program, and I did end up using the new keyword here and there, but they were mostly for instantiating objects that hold data (ie, they didn't have any functions or methods). I'm wondering, did I do something wrong when I used the new keyword for my program. And where can we break that 'rule'?

    Read the article

  • BasicHTTPServer, SimpleHTTPServer and concurrency

    - by braindump
    I'm writing a small web server for testing purposes using python, BasicHTTPServer and SimpleHTTPServer. It looks like it's processing one request at a time. Is there any way to make it a little faster without messing around too deeply? Basicly my code looks as the following and I'd like to keep it this simple ;) os.chdir(webroot) httpd = BaseHTTPServer.HTTPServer(("", port), SimpleHTTPServer.SimpleHTTPRequestHandler) print("Serving directory %s on port %i" %(webroot, port) ) try: httpd.serve_forever() except KeyboardInterrupt: print("Server stopped.")

    Read the article

  • Java Concurrency : Synchronized(this) => and this.wait() and this.notify()

    - by jens
    Hello Experts, I would appreciate your help in understand a "Concurrency Example" from: http://forums.sun.com/thread.jspa?threadID=735386 Qute Start: public synchronized void enqueue(T obj) { // do addition to internal list and then... this.notify(); } public synchronized T dequeue() { while (this.size()==0) { this.wait(); } return // something from the queue } Quote End: My Question is: Why is this code valid? = When I synchronize a method like "public synchronized" = then I synchronize on the "Instance of the Object == this". However in the example above: Calling "dequeue" I will get the "lock/monitor" on this Now I am in the dequeue method. As the list is zero, the calling thread will be "waited" From my understanding I have now a deadlock situation, as I will have no chance of ever enquing an object (from an nother thread), as the "dequeue" method is not yet finised and the dequeue "method" holds the lock on this: So I will never ever get the possibility to call "enequeue" as I will not get the "this" lock. Backround: I have exactly the same problem: I have some kind of connection pool (List of Connections) and need to block if all connections are checked. What is the correct way to synchronize the List to block, if size exceeds a limit or is zero? Thank you very much Jens

    Read the article

  • Concurrency and Coordination Runtime (CCR) Learning Resources

    - by Harry
    I have recently been learning the in's and out's of the Concurrency and Coordination Runtime (CCR). Finding good learning resources for this relatively new technology has been quite difficult. (A quick google search brings up "Creedence Clearwater Revival" as the top result!) Some of the resources I have found: Free e-book chapter from WROX on the Robotics Developer Studio Good Article/post on InfoQ Robotic's Member blog Very active MSDN CCR Forum - Got plenty of help from here! Great MSDN Magazine by Jeffrey Richter Official CCR User Guide - Didn't find this very helpful Great blogging series on CCR iodyner CCR Related Blog - Update: Moved to here Eight or so Videos on Channel9.msdn.com CCR Patterns page on MS Robotics Studio - I haven't read this yet 4 x CCR Questions on Stackoverflow - Most of the questions have been Mine! LOL CCR and DSS toolkit has now been released to MSDN Members Do you have any good learning resources for the CCR? I really hope that Microsoft will publish more material, so far it has been too Robotics specific. I believe that MS needs to acknowledge that most people are using the CCR in issolation from the DSS and Robotics Studio. Update The Mix 2010 conference had a presentation by Myspace about how they have used the CCR framework in their middle tier. They also open sourced the code base. MySpace DataRelay Mix Video Presentation

    Read the article

  • Java Concurrency in practice sample question

    - by andy boot
    I am reading "Java Concurrency in practice" and looking at the example code on page 51. According to the book this piece of code is at risk of of failure if it has not been published properly. Because I like to code examples and break them to prove how they work. I have tried to make it throw an AssertionError but have failed. (Leading me to my previous question) Can anyone post sample code so that an AssertionError is thrown? Rule: Do not modify the Holder class. public class Holder{ private int n; public Holder(int n){ this.n = n; } public void assertSanity(){ if (n != n) { throw new AssertionError("This statement is false"); } } } I have modified the class to make it more fragile but I still can not get an AssertionError thrown. class Holder2{ private int n; private int n2; public Holder2(int n) throws InterruptedException{ this.n = n; Thread.sleep(200); this.n2 = n; } public void assertSanity(){ if (n != n2) { throw new AssertionError("This statement is false"); } } } Is it possible to make either of the above classes throw an AssertionError? Or do we have to accept that they may occasionally do so and we can't write code to prove it?

    Read the article

  • Java Concurrency in practice sample question

    - by andy boot
    I am reading "Java Concurrency in practice" and looking at the example code on page 51. This states that if a thread has references to a shared object then other threads may be able to access that object before the constructor has finished executing. I have tried to put this into practice and so I wrote this code thinking that if I ran it enough times a RuntimeException("World is f*cked") would occur. But it isn't doing. Is this a case of the Java spec not guaranting something but my particular implementation of java guaranteeing it for me? (java version: 1.5.0 on Ubuntu) Or have I misread something in the book? Code: (I expect an exception but it is never thrown) public class Threads { private Widgit w; public static void main(String[] s) throws Exception { while(true){ Threads t = new Threads(); t.runThreads(); } } private void runThreads() throws Exception{ new Checker().start(); w = new Widgit((int)(Math.random() * 100) + 1); } private class Checker extends Thread{ private static final int LOOP_TIMES = 1000; public void run() { int count = 0; for(int i = 0; i < LOOP_TIMES; i++){ try { w.checkMe(); count++; } catch(NullPointerException npe){ //ignore } } System.out.println("checked: "+count+" times out of "+LOOP_TIMES); } } private static class Widgit{ private int n; private int n2; Widgit(int n) throws InterruptedException{ this.n = n; Thread.sleep(2); this.n2 = n; } void checkMe(){ if (n != n2) { throw new RuntimeException("World is f*cked"); } } } }

    Read the article

  • Concurrency problem with arrays (Java)

    - by Johannes
    For an algorithm I'm working on I tried to develop a blacklisting mechanism that can blacklist arrays in a specific way: If "1, 2, 3" is blacklisted "1, 2, 3, 4, 5" is also considered blacklisted. I'm quite happy with the solution I've come up with so far. But there seem to be some serious problems when I access a blacklist from multiple threads. The method "contains" (see code below) sometimes returns true, even if an array is not blacklisted. This problem does not occur if I only use one thread, so it most likely is a concurrency problem. I've tried adding some synchronization, but it didn't change anything. I also tried some slightly different implementations using java.util.concurrent classes. Any ideas on how to fix this? public class Blacklist { private static final int ARRAY_GROWTH = 10; private final Node root = new Node(); private static class Node{ private volatile Node[] childNodes = new Node[ARRAY_GROWTH]; private volatile boolean blacklisted = false; public void blacklist(){ this.blacklisted = true; this.childNodes = null; } } public void add(final int[] array){ synchronized (root) { Node currentNode = this.root; for(final int edge : array){ if(currentNode.blacklisted) return; else if(currentNode.childNodes.length <= edge) { currentNode.childNodes = Arrays.copyOf(currentNode.childNodes, edge + ARRAY_GROWTH); } if(currentNode.childNodes[edge] == null) { currentNode.childNodes[edge] = new Node(); } currentNode = currentNode.childNodes[edge]; } currentNode.blacklist(); } } public boolean contains(final int[] array){ synchronized (root) { Node currentNode = this.root; for(final int edge : array){ if(currentNode.blacklisted) return true; else if(currentNode.childNodes.length <= edge || currentNode.childNodes[edge] == null) return false; currentNode = currentNode.childNodes[edge]; } return currentNode.blacklisted; } } }

    Read the article

  • Survey: how do you unit test your T-SQL?

    - by Alexander Kuznetsov
    How do you unit test your T-SQL? Which libraries/tools do you use? What percentage of your code is covered by unit tests and how do you measure it? Do you think the time and effort which you invested in your unit testing harness has paid off or not? Share this post: email it! | bookmark it! | digg it! | reddit! | kick it! | live it!...(read more)

    Read the article

  • Junit: splitting integration test and Unit tests.

    - by jeff porter
    Hello all, I've inherited a load of Junit test, but these tests (apart from most not working) are a mixture of actual unit test and integration tests (requiring external systems, db etc). So I'm trying to think of a way to actually separate them out, so that I can run the unit test nice and quickly and the integration tests after that. The options are.. 1: Split them into separate directories. 2: Move to Junit4 and annotate the classes to separate them. 3: Use a file naming convention to tell what a class is , i.e. AdapterATest and AdapterAIntergrationTest. 3 has the issue that Eclipse has the option to "Run all tests in the selected project/package or folder". So it would make it very hard to just run the integration tests. 2: runs the risk that developers might start writing integration tests in unit test classes and it just gets messy. 1: Seems like the neatest solution, but my gut says there must be a better solution out there. So that is my question, how do you lot break apart integration tests and proper unit tests?

    Read the article

  • What are the best resources for learning about concurrency and multi-threaded applications?

    - by Zepee
    I realised I have a massive knowledge gap when it comes to multi-threaded applications and concurrent programming. I've covered some basics in the past, but most of it seems to be gone from my mind, and it is definitely a field that I want, and need, to be more knowledgeable about. What are the best resources for learning about building concurrent applications? I'm a very practical oriented person, so if said book contains concrete examples the better, but I'm open to suggestions. I personally prefer to work in pseudocode or C++, and a slant toward game development would be best, but not required.

    Read the article

  • Is there a real difference between dynamic analysis and testing?

    - by user970696
    Often testing is regarded as a dynamic analysis of a software. Yet while writing my thesis, the reviewer noted to me that dynamic analysis is about analyzing the program behind the scenes - e.g. profiling and that it is not the same as testing because its "analysis" which looks inside and observes. I know that "static analysis" is not testing, should we then separate this "dynamic analysis" also from testing? Some books do refer to dynamic analysis in this sense. I would maybe say that testing is a one mean of dynamic analysis?

    Read the article

  • Concurrency with listeners and upload: is this solution sound?

    - by cbrulak
    I'm working on an Android app. We have listeners for position data and camera capture which is timed on a background thread (not a service). The timer thread dictates when the image is captured and also when the data is cached in a local sqlite db. So, my question is how to properly store the location data as it comes in based on the listeners and pull that data so that the database can be updated as the camera capture is executed. I can't put the location data into the database as it arrives because it is processed more frequently than the camera images and the camera images are what dominates the architecture at the moment. (Location data is supplement). However I need the location data to get a rough location of the where the image is captured. My first thought was to have singleton store the location data. And have a semaphore on the getInstance method so if the database update is happening (after an image is captured) we don't have an error. The location data can wait for the database update or it can be lost for that particular event it doesn't really matter. What are you thoughts? Am I on the right track? (And is this the right sub-site or would this be better on stackoverflow?)

    Read the article

  • Concurrency Utilities for Java EE 6: JSR 236 Rebooting

    - by arungupta
    JSR 166 added support for concurrency utilities in the Java platform. The JSR 236's, a.k.a Concurrency Utilities for Java EE, goal was to extend that support to the Java EE platform by adding asynchronous abilities to different application components. The EG was however stagnant since Dec 2003. Its coming back to life with the co-spec lead Anthony Lai's message to the JSR 236 EG (archived here). The JSR will be operating under JCP 2.8's transparency rules and can be tracked at concurrency-spec.java.net. All the mailing lists are archived here. The final release is expected in Q1 2013 and the APIs will live in the javax.enterprise.concurrent package. Please submit your nomination if you would like to join this EG.

    Read the article

  • Test Doubles : Do they go in "source packages" or "test packages"?

    - by sbrattla
    I've got a couple of data access objects (DefaultPersonServices.class, DefaultAddressServices.class) which is responsible for various CRUD operations in a database. A few different classes use these services, but as the services requires that a connection is established with a database I can't really use them in unit tests as they take too long. Thus, I'd like to create a test doubles for them and simply do FakePersonServices.class and FakeAddressService.class implementations which I can use throughout testing. Now, this is all good (I assume)...but my question relates to where I put the test doubles. Should I keep them along with the default implementations (aka "real" implementations) or should I keep them in a corresponding test package. The default implementations are found in Source Packages : com.company.data.services. Should I keep the test doubles here too, or should the test doubles rather be in Test Packages : com.company.data.services?

    Read the article

  • What Are Some Tips For Writing A Large Number of Unit Tests?

    - by joshin4colours
    I've recently been tasked with testing some COM objects of the desktop app I work on. What this means in practice is writing a large number (100) unit tests to test different but related methods and objects. While the unit tests themselves are fairly straight forward (usually one or two Assert()-type checks per test), I'm struggling to figure out the best way to write these tests in a coherent, organized manner. What I have found is that copy and Paste coding should be avoided. It creates more problems than it's worth, and it's even worse than copy-and-paste code in production code because test code has to be more frequently updated and modified. I'm leaning toward trying an OO-approach using but again, the sheer number makes even this approach daunting from an organizational standpoint due to concern with maintenance. It also doesn't help that the tests are currently written in C++, which adds some complexity with memory management issues. Any thoughts or suggestions?

    Read the article

  • If you should only have one assertion per test; how to test multiple inputs?

    - by speg
    I'm trying to build up some test cases, and have read that you should try and limit the number of assertions per test case. So my question is, what is the best way to go about testing a function w/ multiple inputs. For example, I have a function that parses a string from the user and returns the number of minutes. The string can be in the form "5w6h2d1m", where w, h, d, m correspond to the number of weeks, hours, days, and minutes. If I wanted to follow the '1 assertion per test rule' I'd have to make multiple tests for each variation of input? That seems silly so instead I just have something like: self.assertEqual(parse_date('5m'), 5) self.assertEqual(parse_date('5h'), 300) self.assertEqual(parse_date('5d') ,7200) self.assertEqual(parse_date('1d4h20m'), 1700) In the one test case. Is there a better way?

    Read the article

  • How to write automated tests for SQL queries?

    - by James
    The current system we are adopting at work is to write some extremely complex queries which perform multiple calculations and have multiple joins / sub-queries. I don't think I am experienced enough to say if this is correct or not so I am agreeing and attempting to function with this system as it has clear benefits. The problem we are having at the moment is that the person writing the queries makes a lot of mistakes and assumes everything is correct. We have now assigned a tester to analyse all of the queries but this still proves extremely time consuming and stressful. I would like to know how we could create an automated procedure (without specifically writing it with code if possible as I can work out how to do that the long way) to verify a set of 10+ different inputs, verify the output data and say if the calculations are correct. I know I could write a script using specific data in the database and create a script using c# (the db is SQL Server) and verify all the values coming out but I would like to know what the official "standard" is as my experience is lacking in this area and I would like to improve. I am happy to add more information if required, add a comment if necessary. Thank you. Edit: I am using c#

    Read the article

  • What does well written, readable tests look like?

    - by Industrial
    Doing unit testing for the first time at a large scale, I find myself writing a lot of repetitive unit tests for my business logic. Sure, to create complete test suites I need to test all possibilities but readability feels compromised doing what I do - as shown in the psuedocode below. How would a well written, readable test suit look like? describe "UserEntity" -> it "valid name validates" ... it "invalid name doesnt validate" ... it "valid list of followers validate" ..

    Read the article

  • Self-updating collection concurrency issues

    - by DEHAAS
    I am trying to build a self-updating collection. Each item in the collection has a position (x,y). When the position is changed, an event is fired, and the collection will relocate the item. Internally the collection is using a “jagged dictionary”. The outer dictionary uses the x-coordinate a key, while the nested dictionary uses the y-coordinate a key. The nested dictionary then has a list of items as value. The collection also maintains a dictionary to store the items position as stored in the nested dictionaries – item to stored location lookup. I am having some trouble making the collection thread safe, which I really need. Source code for the collection: public class PositionCollection<TItem, TCoordinate> : ICollection<TItem> where TItem : IPositionable<TCoordinate> where TCoordinate : struct, IConvertible { private readonly object itemsLock = new object(); private readonly Dictionary<TCoordinate, Dictionary<TCoordinate, List<TItem>>> items; private readonly Dictionary<TItem, Vector<TCoordinate>> storedPositionLookup; public PositionCollection() { this.items = new Dictionary<TCoordinate, Dictionary<TCoordinate, List<TItem>>>(); this.storedPositionLookup = new Dictionary<TItem, Vector<TCoordinate>>(); } public void Add(TItem item) { if (item.Position == null) { throw new ArgumentException("Item must have a valid position."); } lock (this.itemsLock) { if (!this.items.ContainsKey(item.Position.X)) { this.items.Add(item.Position.X, new Dictionary<TCoordinate, List<TItem>>()); } Dictionary<TCoordinate, List<TItem>> xRow = this.items[item.Position.X]; if (!xRow.ContainsKey(item.Position.Y)) { xRow.Add(item.Position.Y, new List<TItem>()); } xRow[item.Position.Y].Add(item); if (this.storedPositionLookup.ContainsKey(item)) { this.storedPositionLookup[item] = new Vector<TCoordinate>(item.Position); } else { this.storedPositionLookup.Add(item, new Vector<TCoordinate>(item.Position)); // Store a copy of the original position } item.Position.PropertyChanged += (object sender, PropertyChangedEventArgs eventArgs) => this.UpdatePosition(item, eventArgs.PropertyName); } } private void UpdatePosition(TItem item, string propertyName) { lock (this.itemsLock) { Vector<TCoordinate> storedPosition = this.storedPositionLookup[item]; this.RemoveAt(storedPosition, item); this.storedPositionLookup.Remove(item); } } } I have written a simple unit test to check for concurrency issues: [TestMethod] public void TestThreadedPositionChange() { PositionCollection<Crate, int> collection = new PositionCollection<Crate, int>(); Crate crate = new Crate(new Vector<int>(5, 5)); collection.Add(crate); Parallel.For(0, 100, new Action<int>((i) => crate.Position.X += 1)); Crate same = collection[105, 5].First(); Assert.AreEqual(crate, same); } The actual stored position varies every time I run the test. I appreciate any feedback you may have.

    Read the article

< Previous Page | 7 8 9 10 11 12 13 14 15 16 17 18  | Next Page >