Search Results

Search found 3773 results on 151 pages for 'paul james'.

Page 17/151 | < Previous Page | 13 14 15 16 17 18 19 20 21 22 23 24  | Next Page >

  • My first SQL Saturday

    - by Paul Nielsen
    I’m leaving soon for an exciting journey with a thrilling destination – my first SQL Saturday. So I decided to do it right and I’m taking the Amtrak Acela Express from Boston to New York. I love New York! If you’re headed to SQL Saturday #39, and you love database design, I invite you to come to my session on Temporal Database Designs – how to design a table so it can be queried as of any pervious point in time. The proof of concept code is posted at http://temporalsql.codeplex.com/ . See you there....(read more)

    Read the article

  • Upgraded Linux, now CMS Made Simple is spewing errors

    - by Paul Tomblin
    I upgraded my host from Debian Lenny to Debian Squeeze, and now my CMS Made Simple site is spewing PHP errors all over the screen. I thought I'd upgrade the CMS because I haven't done so in a while, but Google Chrome tells me that the CMS Made Simple site is infested with malware. What are my options now? Example errors: Deprecated: Assigning the return value of new by reference is deprecated in /www/danmurn/cms/include.php on line 73 Deprecated: Assigning the return value of new by reference is deprecated in /www/danmurn/cms/include.php on line 162 Deprecated: Assigning the return value of new by reference is deprecated in /www/danmurn/cms/include.php on line 240 Warning: session_start() [function.session-start]: Cannot send session cookie - headers already sent by (output started at /www/danmurn/cms/include.php:73) in /www/danmurn/cms/include.php on line 34 Warning: session_start() [function.session-start]: Cannot send session cache limiter - headers already sent (output started at /www/danmurn/cms/include.php:73) in /www/danmurn/cms/include.php on line 34 Deprecated: Function set_magic_quotes_runtime() is deprecated in /www/danmurn/cms/include.php on line 62 Deprecated: Assigning the return value of new by reference is deprecated in /www/danmurn/cms/lib/classes/class.global.inc.php on line 184 Deprecated: Assigning the return value of new by reference is deprecated in /www/danmurn/cms/lib/classes/class.global.inc.php on line 196

    Read the article

  • Hugh Bin-Haad's SQL Rumours

    - by Paul Nielsen
    Insider rumours and gossip from the murky world of the Database Industry, and from the colourful characters that inhabit it http://www.simple-talk.com/sql/editors-corner/insider-insights/ Share this post: email it! | bookmark it! | digg it! | reddit! | kick it! | live it!...(read more)

    Read the article

  • Fresh Voices

    - by Paul Nielsen
    The PASS Summit 2010 Call for Speakers is closed. When the call went out I offered to review abstracts for anyone interested, which gave a peak into the content of the next PASS, albeit a skewed peak, but what I did see was encouraging. In all I reviewed about 50 abstracts (several of these I found on my junk folder, so I apologize to any that I didn’t see.) The abstracts I reviewed had a mix of new topics and core technologies from new speakers, regulars, and a few MVPs. The observation that stood...(read more)

    Read the article

  • C#/.NET Little Wonders: Of LINQ and Lambdas - A Presentation

    - by James Michael Hare
    Once again, in this series of posts I look at the parts of the .NET Framework that may seem trivial, but can help improve your code by making it easier to write and maintain. The index of all my past little wonders posts can be found here. Today I’m giving a brief beginner’s guide to LINQ and Lambdas at the St. Louis .NET User’s Group so I thought I’d post the presentation here as well.  I updated the presentation a bit as well as added some notes on the query syntax.  Enjoy! The C#/.NET Fundaments: Of Lambdas and LINQ Presentation Of Lambdas and LINQ View more presentations from BlackRabbitCoder   Technorati Tags: C#, CSharp, .NET, Little Wonders, LINQ, Lambdas

    Read the article

  • Hugh Bin-Haad's SQL Rumours

    - by Paul Nielsen
    Insider rumours and gossip from the murky world of the Database Industry, and from the colourful characters that inhabit it http://www.simple-talk.com/sql/editors-corner/insider-insights/ Share this post: email it! | bookmark it! | digg it! | reddit! | kick it! | live it!...(read more)

    Read the article

  • Kernel compile error with iw_ndis.c

    - by James
    Hi, I have a hp pavilion dm3t with intel HD graphics running ubuntu 10.10 64 bit. I'm trying to compile and install a patched kernel according to this, https://launchpad.net/~kamalmostafa/+archive/linux-kamal-mjgbacklight So I downloaded the tarball from here (linked to from the page above): http://kernel.ubuntu.com/git?p=kamal/ubuntu-maverick.git;a=shortlog;h=refs/heads/mjg-backlight I untar'd it to a directory, entered the directory and did: make defconfig I'm not sure if that's what I should have done but it was successful, so I did: make which seemed to work fine until it gave these errors: ubuntu/ndiswrapper/iw_ndis.c:1966: error: unknown field ‘num_private’ specified in initializer ubuntu/ndiswrapper/iw_ndis.c:1966: warning: initialization makes pointer from integer without a cast ubuntu/ndiswrapper/iw_ndis.c:1967: error: unknown field ‘num_private_args’ specified in initializer ubuntu/ndiswrapper/iw_ndis.c:1967: warning: excess elements in struct initializer ubuntu/ndiswrapper/iw_ndis.c:1967: warning: (near initialization for ‘ndis_handler_def’) ubuntu/ndiswrapper/iw_ndis.c:1970: error: unknown field ‘private’ specified in initializer ubuntu/ndiswrapper/iw_ndis.c:1970: warning: initialization makes integer from pointer without a cast ubuntu/ndiswrapper/iw_ndis.c:1970: error: initializer element is not computable at load time ubuntu/ndiswrapper/iw_ndis.c:1970: error: (near initialization for ‘ndis_handler_def.num_standard’) ubuntu/ndiswrapper/iw_ndis.c:1971: error: unknown field ‘private_args’ specified in initializer ubuntu/ndiswrapper/iw_ndis.c:1971: warning: initialization from incompatible pointer type make[2]: *** [ubuntu/ndiswrapper/iw_ndis.o] Error 1 make[1]: *** [ubuntu/ndiswrapper] Error 2 make: *** [ubuntu] Error 2 How can I compile and install this kernel successfully? I'm new to this and would appreciate any help.

    Read the article

  • C#/.NET Little Wonders: The EventHandler and EventHandler&lt;TEventArgs&gt; delegates

    - by James Michael Hare
    Once again, in this series of posts I look at the parts of the .NET Framework that may seem trivial, but can help improve your code by making it easier to write and maintain. The index of all my past little wonders posts can be found here. In the last two weeks, we examined the Action family of delegates (and delegates in general), and the Func family of delegates and how they can be used to support generic, reusable algorithms and classes. So this week, we are going to look at a handy pair of delegates that can be used to eliminate the need for defining custom delegates when creating events: the EventHandler and EventHandler<TEventArgs> delegates. Events and delegates Before we begin, let’s quickly consider events in .NET.  According to the MSDN: An event in C# is a way for a class to provide notifications to clients of that class when some interesting thing happens to an object. So, basically, you can create an event in a type so that users of that type can subscribe to notifications of things of interest.  How is this different than some of the delegate programming that we talked about in the last two weeks?  Well, you can think of an event as a special access modifier on a delegate.  Some differences between the two are: Events are a special access case of delegates They behave much like delegates instances inside the type they are declared in, but outside of that type they can only be (un)subscribed to. Events can specify add/remove behavior explicitly If you want to do additional work when someone subscribes or unsubscribes to an event, you can specify the add and remove actions explicitly. Events have access modifiers, but these only specify the access level of those who can (un)subscribe A public event, for example, means anyone can (un)subscribe, but it does not mean that anyone can raise (invoke) the event directly.  Events can only be raised by the type that contains them In contrast, if a delegate is visible, it can be invoked outside of the object (not even in a sub-class!). Events tend to be for notifications only, and should be treated as optional Semantically speaking, events typically don’t perform work on the the class directly, but tend to just notify subscribers when something of note occurs. My basic rule-of-thumb is that if you are just wanting to notify any listeners (who may or may not care) that something has happened, use an event.  However, if you want the caller to provide some function to perform to direct the class about how it should perform work, make it a delegate. Declaring events using custom delegates To declare an event in a type, we simply use the event keyword and specify its delegate type.  For example, let’s say you wanted to create a new TimeOfDayTimer that triggers at a given time of the day (as opposed to on an interval).  We could write something like this: 1: public delegate void TimeOfDayHandler(object source, ElapsedEventArgs e); 2:  3: // A timer that will fire at time of day each day. 4: public class TimeOfDayTimer : IDisposable 5: { 6: // Event that is triggered at time of day. 7: public event TimeOfDayHandler Elapsed; 8:  9: // ... 10: } The first thing to note is that the event is a delegate type, which tells us what types of methods may subscribe to it.  The second thing to note is the signature of the event handler delegate, according to the MSDN: The standard signature of an event handler delegate defines a method that does not return a value, whose first parameter is of type Object and refers to the instance that raises the event, and whose second parameter is derived from type EventArgs and holds the event data. If the event does not generate event data, the second parameter is simply an instance of EventArgs. Otherwise, the second parameter is a custom type derived from EventArgs and supplies any fields or properties needed to hold the event data. So, in a nutshell, the event handler delegates should return void and take two parameters: An object reference to the object that raised the event. An EventArgs (or a subclass of EventArgs) reference to event specific information. Even if your event has no additional information to provide, you are still expected to provide an EventArgs instance.  In this case, feel free to pass the EventArgs.Empty singleton instead of creating new instances of EventArgs (to avoid generating unneeded memory garbage). The EventHandler delegate Because many events have no additional information to pass, and thus do not require custom EventArgs, the signature of the delegates for subscribing to these events is typically: 1: // always takes an object and an EventArgs reference 2: public delegate void EventHandler(object sender, EventArgs e) It would be insane to recreate this delegate for every class that had a basic event with no additional event data, so there already exists a delegate for you called EventHandler that has this very definition!  Feel free to use it to define any events which supply no additional event information: 1: public class Cache 2: { 3: // event that is raised whenever the cache performs a cleanup 4: public event EventHandler OnCleanup; 5:  6: // ... 7: } This will handle any event with the standard EventArgs (no additional information).  But what of events that do need to supply additional information?  Does that mean we’re out of luck for subclasses of EventArgs?  That’s where the generic for of EventHandler comes into play… The generic EventHandler<TEventArgs> delegate Starting with the introduction of generics in .NET 2.0, we have a generic delegate called EventHandler<TEventArgs>.  Its signature is as follows: 1: public delegate void EventHandler<TEventArgs>(object sender, TEventArgs e) 2: where TEventArgs : EventArgs This is similar to EventHandler except it has been made generic to support the more general case.  Thus, it will work for any delegate where the first argument is an object (the sender) and the second argument is a class derived from EventArgs (the event data). For example, let’s say we wanted to create a message receiver, and we wanted it to have a few events such as OnConnected that will tell us when a connection is established (probably with no additional information) and OnMessageReceived that will tell us when a new message arrives (probably with a string for the new message text). So for OnMessageReceived, our MessageReceivedEventArgs might look like this: 1: public sealed class MessageReceivedEventArgs : EventArgs 2: { 3: public string Message { get; set; } 4: } And since OnConnected needs no event argument type defined, our class might look something like this: 1: public class MessageReceiver 2: { 3: // event that is called when the receiver connects with sender 4: public event EventHandler OnConnected; 5:  6: // event that is called when a new message is received. 7: public event EventHandler<MessageReceivedEventArgs> OnMessageReceived; 8:  9: // ... 10: } Notice, nowhere did we have to define a delegate to fit our event definition, the EventHandler and generic EventHandler<TEventArgs> delegates fit almost anything we’d need to do with events. Sidebar: Thread-safety and raising an event When the time comes to raise an event, we should always check to make sure there are subscribers, and then only raise the event if anyone is subscribed.  This is important because if no one is subscribed to the event, then the instance will be null and we will get a NullReferenceException if we attempt to raise the event. 1: // This protects against NullReferenceException... or does it? 2: if (OnMessageReceived != null) 3: { 4: OnMessageReceived(this, new MessageReceivedEventArgs(aMessage)); 5: } The above code seems to handle the null reference if no one is subscribed, but there’s a problem if this is being used in multi-threaded environments.  For example, assume we have thread A which is about to raise the event, and it checks and clears the null check and is about to raise the event.  However, before it can do that thread B unsubscribes to the event, which sets the delegate to null.  Now, when thread A attempts to raise the event, this causes the NullReferenceException that we were hoping to avoid! To counter this, the simplest best-practice method is to copy the event (just a multicast delegate) to a temporary local variable just before we raise it.  Since we are inside the class where this event is being raised, we can copy it to a local variable like this, and it will protect us from multi-threading since multicast delegates are immutable and assignments are atomic: 1: // always make copy of the event multi-cast delegate before checking 2: // for null to avoid race-condition between the null-check and raising it. 3: var handler = OnMessageReceived; 4: 5: if (handler != null) 6: { 7: handler(this, new MessageReceivedEventArgs(aMessage)); 8: } The very slight trade-off is that it’s possible a class may get an event after it unsubscribes in a multi-threaded environment, but this is a small risk and classes should be prepared for this possibility anyway.  For a more detailed discussion on this, check out this excellent Eric Lippert blog post on Events and Races. Summary Generic delegates give us a lot of power to make generic algorithms and classes, and the EventHandler delegate family gives us the flexibility to create events easily, without needing to redefine delegates over and over.  Use them whenever you need to define events with or without specialized EventArgs.   Tweet Technorati Tags: .NET, C#, CSharp, Little Wonders, Generics, Delegates, EventHandler

    Read the article

  • I'm a happy camper

    - by Paul Nielsen
    I’m satisfied with SQL Server 2008. It meets my needs and I can build whatever I want in the database. There’s no feature that it lacks that blocking my development. SQL Server 2008 is what I see when I close my eyes and dream. Maybe I’m just pleased with my current projects, maybe I’m being dense, maybe I lack imagination, tell me if I’m being stupid, but I feel no compelling need for an upgrade. If Microsoft didn’t ship the new version for 3 or 5 more years, I’d be ok with that. I’m sure there...(read more)

    Read the article

  • AutoAudit 1.10c

    - by Paul Nielsen
    AutoAudit is a free SQL Server (2005, 2008) Code-Gen utility that creates Audit Trail Triggers with: · Created, Modified, and RowVersion (incrementing INT) columns to table · Creates View to reconstruct deleted rows · Creates UDF to reconstruct Row History · Schema Audit Trigger to track schema changes · Re-code-gens triggers when Alter Table changes the table Version 1.10c Adds: · Createdby and ModifiedBy columns. Pass the user to the column and AutoAudit records that username instead of the Suser_Sname...(read more)

    Read the article

  • Get the onended event for an AudioBuffer in HTML5/Chrome

    - by Matthew James Davis
    So I am playing audio file in Chrome and I want to detect when playing has ended so I can delete references to it. Here is my code var source = context.createBufferSource(); source.buffer = sound.buffer; source.loop = sound.loop; source.onended = function() { delete playingSounds[soundName]; } source.connect(mainNode); source.start(0, sound.start, sound.length); however, the event handler doesn't fire. Is this not yet supported as described by the W3 specification? Or am I doing something wrong?

    Read the article

  • Survey: Do you write custom SQL CLR procedures/functions/etc

    - by James Luetkehoelter
    I'm quite curious because despite the great capabilities of writing CLR-based stored procedures to off-load those nasty operations TSQL isn't that great at (like iteration, or complex math), I'm continuing to see a wealth of SQL 2008 databases with complex stored procedures and functions which would make great candidates. The in-house skill to create the CLR code exists as well, but there is flat out resistance to use it. In one scenario I was told "Oh, iteration isn't a problem because we've trained...(read more)

    Read the article

  • C#/.NET Little Wonders: A Redux

    - by James Michael Hare
    I gave my Little Wonders presentation to the Topeka Dot Net Users' Group today, so re-posting the links to all the previous posts for them. The Presentation: C#/.NET Little Wonders: A Presentation The Original Trilogy: C#/.NET Five Little Wonders (part 1) C#/.NET Five More Little Wonders (part 2) C#/.NET Five Final Little Wonders (part 3) The Subsequent Sequels: C#/.NET Little Wonders: ToDictionary() and ToList() C#/.NET Little Wonders: DateTime is Packed With Goodies C#/.NET Little Wonders: Fun With Enum Methods C#/.NET Little Wonders: Cross-Calling Constructors C#/.NET Little Wonders: Constraining Generics With Where Clause C#/.NET Little Wonders: Comparer<T>.Default C#/.NET Little Wonders: The Useful (But Overlooked) Sets The Concurrent Wonders: C#/.NET Little Wonders: The Concurrent Collections (1 of 3) - ConcurrentQueue and ConcurrentStack C#/.NET Little Wonders: The Concurrent Collections (2 of 3) - ConcurrentDictionary Tweet   Technorati Tags: .NET,C#,Little Wonders

    Read the article

  • C#/.NET Little Wonders: The Timeout static class

    - by James Michael Hare
    Once again, in this series of posts I look at the parts of the .NET Framework that may seem trivial, but can help improve your code by making it easier to write and maintain. The index of all my past little wonders posts can be found here. When I started the “Little Wonders” series, I really wanted to pay homage to parts of the .NET Framework that are often small but can help in big ways.  The item I have to discuss today really is a very small item in the .NET BCL, but once again I feel it can help make the intention of code much clearer and thus is worthy of note. The Problem - Magic numbers aren’t very readable or maintainable In my first Little Wonders Post (Five Little Wonders That Make Code Better) I mention the TimeSpan factory methods which, I feel, really help the readability of constructed TimeSpan instances. Just to quickly recap that discussion, ask yourself what the TimeSpan specified in each case below is 1: // Five minutes? Five Seconds? 2: var fiveWhat1 = new TimeSpan(0, 0, 5); 3: var fiveWhat2 = new TimeSpan(0, 0, 5, 0); 4: var fiveWhat3 = new TimeSpan(0, 0, 5, 0, 0); You’d think they’d all be the same unit of time, right?  After all, most overloads tend to tack additional arguments on the end.  But this is not the case with TimeSpan, where the constructor forms are:     TimeSpan(int hours, int minutes, int seconds);     TimeSpan(int days, int hours, int minutes, int seconds);     TimeSpan(int days, int hours, int minutes, int seconds, int milliseconds); Notice how in the 4 and 5 parameter version we suddenly have the parameter days slipping in front of hours?  This can make reading constructors like those above much harder.  Fortunately, there are TimeSpan factory methods to help make your intention crystal clear: 1: // Ah! Much clearer! 2: var fiveSeconds = TimeSpan.FromSeconds(5); These are great because they remove all ambiguity from the reader!  So in short, magic numbers in constructors and methods can be ambiguous, and anything we can do to clean up the intention of the developer will make the code much easier to read and maintain. Timeout – Readable identifiers for infinite timeout values In a similar way to TimeSpan, let’s consider specifying timeouts for some of .NET’s (or our own) many methods that allow you to specify timeout periods. For example, in the TPL Task class, there is a family of Wait() methods that can take TimeSpan or int for timeouts.  Typically, if you want to specify an infinite timeout, you’d just call the version that doesn’t take a timeout parameter at all: 1: myTask.Wait(); // infinite wait But there are versions that take the int or TimeSpan for timeout as well: 1: // Wait for 100 ms 2: myTask.Wait(100); 3:  4: // Wait for 5 seconds 5: myTask.Wait(TimeSpan.FromSeconds(5); Now, if we want to specify an infinite timeout to wait on the Task, we could pass –1 (or a TimeSpan set to –1 ms), which what the .NET BCL methods with timeouts use to represent an infinite timeout: 1: // Also infinite timeouts, but harder to read/maintain 2: myTask.Wait(-1); 3: myTask.Wait(TimeSpan.FromMilliseconds(-1)); However, these are not as readable or maintainable.  If you were writing this code, you might make the mistake of thinking 0 or int.MaxValue was an infinite timeout, and you’d be incorrect.  Also, reading the code above it isn’t as clear that –1 is infinite unless you happen to know that is the specified behavior. To make the code like this easier to read and maintain, there is a static class called Timeout in the System.Threading namespace which contains definition for infinite timeouts specified as both int and TimeSpan forms: Timeout.Infinite An integer constant with a value of –1 Timeout.InfiniteTimeSpan A static readonly TimeSpan which represents –1 ms (only available in .NET 4.5+) This makes our calls to Task.Wait() (or any other calls with timeouts) much more clear: 1: // intention to wait indefinitely is quite clear now 2: myTask.Wait(Timeout.Infinite); 3: myTask.Wait(Timeout.InfiniteTimeSpan); But wait, you may say, why would we care at all?  Why not use the version of Wait() that takes no arguments?  Good question!  When you’re directly calling the method with an infinite timeout that’s what you’d most likely do, but what if you are just passing along a timeout specified by a caller from higher up?  Or perhaps storing a timeout value from a configuration file, and want to default it to infinite? For example, perhaps you are designing a communications module and want to be able to shutdown gracefully, but if you can’t gracefully finish in a specified amount of time you want to force the connection closed.  You could create a Shutdown() method in your class, and take a TimeSpan or an int for the amount of time to wait for a clean shutdown – perhaps waiting for client to acknowledge – before terminating the connection.  So, assume we had a pub/sub system with a class to broadcast messages: 1: // Some class to broadcast messages to connected clients 2: public class Broadcaster 3: { 4: // ... 5:  6: // Shutdown connection to clients, wait for ack back from clients 7: // until all acks received or timeout, whichever happens first 8: public void Shutdown(int timeout) 9: { 10: // Kick off a task here to send shutdown request to clients and wait 11: // for the task to finish below for the specified time... 12:  13: if (!shutdownTask.Wait(timeout)) 14: { 15: // If Wait() returns false, we timed out and task 16: // did not join in time. 17: } 18: } 19: } We could even add an overload to allow us to use TimeSpan instead of int, to give our callers the flexibility to specify timeouts either way: 1: // overload to allow them to specify Timeout in TimeSpan, would 2: // just call the int version passing in the TotalMilliseconds... 3: public void Shutdown(TimeSpan timeout) 4: { 5: Shutdown(timeout.TotalMilliseconds); 6: } Notice in case of this class, we don’t assume the caller wants infinite timeouts, we choose to rely on them to tell us how long to wait.  So now, if they choose an infinite timeout, they could use the –1, which is more cryptic, or use Timeout class to make the intention clear: 1: // shutdown the broadcaster, waiting until all clients ack back 2: // without timing out. 3: myBroadcaster.Shutdown(Timeout.Infinite); We could even add a default argument using the int parameter version so that specifying no arguments to Shutdown() assumes an infinite timeout: 1: // Modified original Shutdown() method to add a default of 2: // Timeout.Infinite, works because Timeout.Infinite is a compile 3: // time constant. 4: public void Shutdown(int timeout = Timeout.Infinite) 5: { 6: // same code as before 7: } Note that you can’t default the ShutDown(TimeSpan) overload with Timeout.InfiniteTimeSpan since it is not a compile-time constant.  The only acceptable default for a TimeSpan parameter would be default(TimeSpan) which is zero milliseconds, which specified no wait, not infinite wait. Summary While Timeout.Infinite and Timeout.InfiniteTimeSpan are not earth-shattering classes in terms of functionality, they do give you very handy and readable constant values that you can use in your programs to help increase readability and maintainability when specifying infinite timeouts for various timeouts in the BCL and your own applications. Technorati Tags: C#,CSharp,.NET,Little Wonders,Timeout,Task

    Read the article

  • C++ Little Wonders: The C++11 auto keyword redux

    - by James Michael Hare
    I’ve decided to create a sub-series of my Little Wonders posts to focus on C++.  Just like their C# counterparts, these posts will focus on those features of the C++ language that can help improve code by making it easier to write and maintain.  The index of the C# Little Wonders can be found here. This has been a busy week with a rollout of some new website features here at my work, so I don’t have a big post for this week.  But I wanted to write something up, and since lately I’ve been renewing my C++ skills in a separate project, it seemed like a good opportunity to start a C++ Little Wonders series.  Most of my development work still tends to focus on C#, but it was great to get back into the saddle and renew my C++ knowledge.  Today I’m going to focus on a new feature in C++11 (formerly known as C++0x, which is a major move forward in the C++ language standard).  While this small keyword can seem so trivial, I feel it is a big step forward in improving readability in C++ programs. The auto keyword If you’ve worked on C++ for a long time, you probably have some passing familiarity with the old auto keyword as one of those rarely used C++ keywords that was almost never used because it was the default. That is, in the code below (before C++11): 1: int foo() 2: { 3: // automatic variables (allocated and deallocated on stack) 4: int x; 5: auto int y; 6:  7: // static variables (retain their value across calls) 8: static int z; 9:  10: return 0; 11: } The variable x is assumed to be auto because that is the default, thus it is unnecessary to specify it explicitly as in the declaration of y below that.  Basically, an auto variable is one that is allocated and de-allocated on the stack automatically.  Contrast this to static variables, that are allocated statically and exist across the lifetime of the program. Because auto was so rarely (if ever) used since it is the norm, they decided to remove it for this purpose and give it new meaning in C++11.  The new meaning of auto: implicit typing Now, if your compiler supports C++ 11 (or at least a good subset of C++11 or 0x) you can take advantage of type inference in C++.  For those of you from the C# world, this means that the auto keyword in C++ now behaves a lot like the var keyword in C#! For example, many of us have had to declare those massive type declarations for an iterator before.  Let’s say we have a std::map of std::string to int which will map names to ages: 1: std::map<std::string, int> myMap; And then let’s say we want to find the age of a given person: 1: // Egad that's a long type... 2: std::map<std::string, int>::const_iterator pos = myMap.find(targetName); Notice that big ugly type definition to declare variable pos?  Sure, we could shorten this by creating a typedef of our specific map type if we wanted, but now with the auto keyword there’s no need: 1: // much shorter! 2: auto pos = myMap.find(targetName); The auto now tells the compiler to determine what type pos should be based on what it’s being assigned to.  This is not dynamic typing, it still determines the type as if it were explicitly declared and once declared that type cannot be changed.  That is, this is invalid: 1: // x is type int 2: auto x = 42; 3:  4: // can't assign string to int 5: x = "Hello"; Once the compiler determines x is type int it is exactly as if we typed int x = 42; instead, so don’t' confuse it with dynamic typing, it’s still very type-safe. An interesting feature of the auto keyword is that you can modify the inferred type: 1: // declare method that returns int* 2: int* GetPointer(); 3:  4: // p1 is int*, auto inferred type is int 5: auto *p1 = GetPointer(); 6:  7: // ps is int*, auto inferred type is int* 8: auto p2 = GetPointer(); Notice in both of these cases, p1 and p2 are determined to be int* but in each case the inferred type was different.  because we declared p1 as auto *p1 and GetPointer() returns int*, it inferred the type int was needed to complete the declaration.  In the second case, however, we declared p2 as auto p2 which means the inferred type was int*.  Ultimately, this make p1 and p2 the same type, but which type is inferred makes a difference, if you are chaining multiple inferred declarations together.  In these cases, the inferred type of each must match the first: 1: // Type inferred is int 2: // p1 is int* 3: // p2 is int 4: // p3 is int& 5: auto *p1 = GetPointer(), p2 = 42, &p3 = p2; Note that this works because the inferred type was int, if the inferred type was int* instead: 1: // syntax error, p1 was inferred to be int* so p2 and p3 don't make sense 2: auto p1 = GetPointer(), p2 = 42, &p3 = p2; You could also use const or static to modify the inferred type: 1: // inferred type is an int, theAnswer is a const int 2: const auto theAnswer = 42; 3:  4: // inferred type is double, Pi is a static double 5: static auto Pi = 3.1415927; Thus in the examples above it inferred the types int and double respectively, which were then modified to const and static. Summary The auto keyword has gotten new life in C++11 to allow you to infer the type of a variable from it’s initialization.  This simple little keyword can be used to cut down large declarations for complex types into a much more readable form, where appropriate.   Technorati Tags: C++, C++11, Little Wonders, auto

    Read the article

  • MDX using EXISTING, AGGREGATE, CROSSJOIN and WHERE

    - by James Rogers
    It is a well-published approach to using the EXISTING function to decode AGGREGATE members and nested sub-query filters.  Mosha wrote a good blog on it here and a more recent one here.  The use of EXISTING in these scenarios is very useful and sometimes the only option when dealing with multi-select filters.  However, there are some limitations I have run across when using the EXISTING function against an AGGREGATE member:   The AGGREGATE member must be assigned to the Dimension.Hierarchy being detected by the EXISTING function in the calculated measure. The AGGREGATE member cannot contain a crossjoin from any other dimension or hierarchy or EXISTING will not be able to detect the members in the AGGREGATE member.   Take the following query (from Adventure Works DW 2008):   With   member [Week Count] as 'count(existing([Date].[Fiscal Weeks].[Fiscal Week].members))'    member [Date].[Fiscal Weeks].[CM] as 'AGGREGATE({[Date].[Fiscal Weeks].[Fiscal Week].&[47]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[48]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[49]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[50]&[2004]})'   select   {[Week Count]} on columns from   [Adventure Works]     where   [Date].[Fiscal Weeks].[CM]   Here we are attempting to count the existing fiscal weeks in slicer.  This is useful to get a per-week average for another member. Many applications generate queries in this manner (such as Oracle OBIEE).  This query returns the correct result of (4) weeks. Now let's put a twist in it.  What if the querying application submits the query in the following manner:   With   member [Week Count] as 'count(existing([Date].[Fiscal Weeks].[Fiscal Week].members))'    member [Customer].[Customer Geography].[CM] as 'AGGREGATE({[Date].[Fiscal Weeks].[Fiscal Week].&[47]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[48]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[49]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[50]&[2004]})'   select   {[Week Count]} on columns from   [Adventure Works]     where   [Customer].[Customer Geography].[CM]   Here we are attempting to count the existing fiscal weeks in slicer.  However, the AGGREGATE member is built on a different dimension (in name) than the one EXISTING is trying to detect.  In this case the query returns (174) which is the total number of [Date].[Fiscal Weeks].[Fiscal Week].members defined in the dimension.   Now another twist, the AGGREGATE member will be named appropriately and contain the hierarchy we are trying to detect with EXISTING but it will be cross-joined with another hierarchy:   With   member [Week Count] as 'count(existing([Date].[Fiscal Weeks].[Fiscal Week].members))'    member [Date].[Fiscal Weeks].[CM] as 'AGGREGATE({[Date].[Fiscal Weeks].[Fiscal Week].&[47]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[48]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[49]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[50]&[2004]}*    {[Customer].[Customer Geography].[Country].&[Australia],[Customer].[Customer Geography].[Country].&[United States]})'  select   {[Week Count]} on columns from   [Adventure Works]    where   [Date].[Fiscal Weeks].[CM]   Once again, we are attempting to count the existing fiscal weeks in slicer.  Again, in this case the query returns (174) which is the total number of [Date].[Fiscal Weeks].[Fiscal Week].members defined in the dimension. However, in 2008 R2 this query returns the correct result of 4 and additionally , the following will return the count of existing countries as well (2):   With   member [Week Count] as 'count(existing([Date].[Fiscal Weeks].[Fiscal Week].members))'   member [Country Count] as 'count(existing([Customer].[Customer Geography].[Country].members))'  member [Date].[Fiscal Weeks].[CM] as 'AGGREGATE({[Date].[Fiscal Weeks].[Fiscal Week].&[47]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[48]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[49]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[50]&[2004]}*    {[Customer].[Customer Geography].[Country].&[Australia],[Customer].[Customer Geography].[Country].&[United States]})'  select   {[Week Count]} on columns from   [Adventure Works]    where   [Date].[Fiscal Weeks].[CM]   2008 R2 seems to work as long as the AGGREGATE member is on at least one of the hierarchies attempting to be detected (i.e. [Date].[Fiscal Weeks] or [Customer].[Customer Geography]). If not, it seems that the engine cannot find a "point of entry" into the aggregate member and ignores it for calculated members.   One way around this would be to put the sets from the AGGREGATE member explicitly in the WHERE clause (slicer).  I realize this is only supported in SSAS 2005 and 2008.  However, after talking with Chris Webb (his blog is here and I highly recommend following his efforts and musings) it is a far more efficient way to filter/slice a query:   With   member [Week Count] as 'count(existing([Date].[Fiscal Weeks].[Fiscal Week].members))'    select   {[Week Count]} on columns from   [Adventure Works]    where   ({[Date].[Fiscal Weeks].[Fiscal Week].&[47]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[48]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[49]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[50]&[2004]}   ,{[Customer].[Customer Geography].[Country].&[Australia],[Customer].[Customer Geography].[Country].&[United States]})   This query returns the correct result of (4) weeks.  Additionally, we can count the cross-join members of the two hierarchies in the slicer:   With   member [Week Count] as 'count(existing([Date].[Fiscal Weeks].[Fiscal Week].members)*existing([Customer].[Customer Geography].[Country].members))'    select   {[Week Count]} on columns from   [Adventure Works]    where   ({[Date].[Fiscal Weeks].[Fiscal Week].&[47]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[48]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[49]&[2004],[Date].[Fiscal Weeks].[Fiscal Week].&[50]&[2004]}   ,{[Customer].[Customer Geography].[Country].&[Australia],[Customer].[Customer Geography].[Country].&[United States]})   We get the correct number of (8) here.

    Read the article

  • Oracle Supply Chain Sessions at Collaborate 2010

    - by Paul Homchick
    Oracle SCM executives will be presenting more than thirty sessions at the Collaborate 2010 Oracle Users' Group Conference in Las Vegas, Nevada, April 18-22.Charles Phillips, President of Oracle will give a keynote address on Monday "Transforming Customer Value -- Delivering Highest Customer Service."A list of the Oracle SCM sessions to be presented at Collaborate 2010 is presented after the break...

    Read the article

  • ext4 jbd2 journaling active even on empty filesystem

    - by Paul
    I have been having several issues with my ext4 filesystems that seem to be due to jbd2 journaling. I made a related post here and am rephrasing it with the hope that someone may be able to help. For a minimal example, I start with an empty 8gb USB stick and use gparted to create one ext4 partition. The command used by gparted when creating the ext4 file system is: mkfs.ext4 -j -O extent -L DataTraveler8gb /dev/sde1 I check the filesystem with gparted: e2fsck -f -y -v /dev/sde1 and I mount it: sudo mount /dev/sde1 /media/test The disk is empty, but the journaling is very active on this disk (/dev/sde1). The other disks are ext4 SSDs formatted similarly. A snapshot of iotop: % sudo iotop -oPa Total DISK READ: 0.00 B/s | Total DISK WRITE: 2027.21 K/s PID PRIO USER DISK READ DISK WRITE SWAPIN IO COMMAND 262 be/3 root 0.00 B 56.00 K 0.00 % 0.18 % [jbd2/sda1-8] 29069 be/3 root 0.00 B 0.00 B 0.00 % 0.16 % [jbd2/sde1-8] 891 be/3 root 0.00 B 4.00 K 0.00 % 0.03 % [jbd2/sdc1-8] What is jbd2 doing with /dev/sde1? If I follow the same steps with a larger 2Tb disk, iotop indicates this empty disk is constantly being written to by jbd2 at the rate of Mb/s as soon as I mount it. On the other disks, which have the OS and /home, I have tried to find if any files are being modified by processes to cause this behavior but couldn't find any. I also moved many of the disk intensive process to use a tmpfs. And used noatime. I have another non-SSD hard disk on this machine, /dev/sdb, that is also ext4 but was not formatted by gparted (given to me by a coworker). It does not appear in iotop. So I am assuming there an issue with gparted. Any suggestions are appreciated. Also any tips on how to modify existing partitions to fix the issue without having to start from scratch would be great. There are some posts related to jbd2 but they didn't help (eg. here).

    Read the article

  • An issue with tessellation a model with DirectX11

    - by Paul Ske
    I took the hardware tessellation tutorial from Rastertek and implemended texturing instead of color. This is great, so I wanted to implemended the same techique to a model inside my game editor and I noticed it doesn't draw anything. I compared the detailed tessellation from DirectX SDK sample. Inside the shader file - if I replace the HullInputType with PixelInputType it draws. So, I think because when I compiled the shaders inside the program it compiles VertexShader, PixelShader, HullShader then DomainShader. Isn't it suppose to be VertexShader, HullSHader, DomainShader then PixelShader or does it really not matter? I am just curious why wouldn't the model even be drawn when HullInputType but renders fine with PixelInputType. Shader Code: [code] cbuffer ConstantBuffer { float4x4 WVP; float4x4 World; // the rotation matrix float3 lightvec; // the light's vector float4 lightcol; // the light's color float4 ambientcol; // the ambient light's color bool isSelected; } cbuffer cameraBuffer { float3 cameraDirection; float padding; } cbuffer TessellationBuffer { float tessellationAmount; float3 padding2; } struct ConstantOutputType { float edges[3] : SV_TessFactor; float inside : SV_InsideTessFactor; }; Texture2D Texture; Texture2D NormalTexture; SamplerState ss { MinLOD = 5.0f; MipLODBias = 0.0f; }; struct HullOutputType { float3 position : POSITION; float2 texcoord : TEXCOORD0; float3 normal : NORMAL; float3 tangent : TANGENT; }; struct HullInputType { float4 position : POSITION; float2 texcoord : TEXCOORD0; float3 normal : NORMAL; float3 tangent : TANGENT; }; struct VertexInputType { float4 position : POSITION; float2 texcoord : TEXCOORD; float3 normal : NORMAL; float3 tangent : TANGENT; uint uVertexID : SV_VERTEXID; }; struct PixelInputType { float4 position : SV_POSITION; float2 texcoord : TEXCOORD0; // texture coordinates float3 normal : NORMAL; float3 tangent : TANGENT; float4 color : COLOR; float3 viewDirection : TEXCOORD1; float4 depthBuffer : TEXTURE0; }; HullInputType VShader(VertexInputType input) { HullInputType output; output.position.w = 1.0f; output.position = mul(input.position,WVP); output.texcoord = input.texcoord; output.normal = input.normal; output.tangent = input.tangent; //output.normal = mul(normal,World); //output.tangent = mul(tangent,World); //output.color = output.color; //output.texcoord = texcoord; // set the texture coordinates, unmodified return output; } ConstantOutputType TexturePatchConstantFunction(InputPatch inputPatch,uint patchID : SV_PrimitiveID) { ConstantOutputType output; output.edges[0] = tessellationAmount; output.edges[1] = tessellationAmount; output.edges[2] = tessellationAmount; output.inside = tessellationAmount; return output; } [domain("tri")] [partitioning("integer")] [outputtopology("triangle_cw")] [outputcontrolpoints(3)] [patchconstantfunc("TexturePatchConstantFunction")] HullOutputType HShader(InputPatch patch, uint pointId : SV_OutputControlPointID, uint patchId : SV_PrimitiveID) { HullOutputType output; // Set the position for this control point as the output position. output.position = patch[pointId].position; // Set the input color as the output color. output.texcoord = patch[pointId].texcoord; output.normal = patch[pointId].normal; output.tangent = patch[pointId].tangent; return output; } [domain("tri")] PixelInputType DShader(ConstantOutputType input, float3 uvwCoord : SV_DomainLocation, const OutputPatch patch) { float3 vertexPosition; float2 uvPosition; float4 worldposition; PixelInputType output; // Interpolate world space position with barycentric coordinates float3 vWorldPos = uvwCoord.x * patch[0].position + uvwCoord.y * patch[1].position + uvwCoord.z * patch[2].position; // Determine the position of the new vertex. vertexPosition = vWorldPos; // Calculate the position of the new vertex against the world, view, and projection matrices. output.position = mul(float4(vertexPosition, 1.0f),WVP); // Send the input color into the pixel shader. output.texcoord = uvwCoord.x * patch[0].position + uvwCoord.y * patch[1].position + uvwCoord.z * patch[2].position; output.normal = uvwCoord.x * patch[0].position + uvwCoord.y * patch[1].position + uvwCoord.z * patch[2].position; output.tangent = uvwCoord.x * patch[0].position + uvwCoord.y * patch[1].position + uvwCoord.z * patch[2].position; //output.depthBuffer = output.position; //output.depthBuffer.w = 1.0f; //worldposition = mul(output.position,WVP); //output.viewDirection = cameraDirection.xyz - worldposition.xyz; //output.viewDirection = normalize(output.viewDirection); return output; } [/code] Somethings are commented out but will be in place when fixed. I'm probably not connecting something correctly.

    Read the article

  • C#/.NET Little Wonders: ConcurrentBag and BlockingCollection

    - by James Michael Hare
    In the first week of concurrent collections, began with a general introduction and discussed the ConcurrentStack<T> and ConcurrentQueue<T>.  The last post discussed the ConcurrentDictionary<T> .  Finally this week, we shall close with a discussion of the ConcurrentBag<T> and BlockingCollection<T>. For more of the "Little Wonders" posts, see C#/.NET Little Wonders: A Redux. Recap As you'll recall from the previous posts, the original collections were object-based containers that accomplished synchronization through a Synchronized member.  With the advent of .NET 2.0, the original collections were succeeded by the generic collections which are fully type-safe, but eschew automatic synchronization.  With .NET 4.0, a new breed of collections was born in the System.Collections.Concurrent namespace.  Of these, the final concurrent collection we will examine is the ConcurrentBag and a very useful wrapper class called the BlockingCollection. For some excellent information on the performance of the concurrent collections and how they perform compared to a traditional brute-force locking strategy, see this informative whitepaper by the Microsoft Parallel Computing Platform team here. ConcurrentBag<T> – Thread-safe unordered collection. Unlike the other concurrent collections, the ConcurrentBag<T> has no non-concurrent counterpart in the .NET collections libraries.  Items can be added and removed from a bag just like any other collection, but unlike the other collections, the items are not maintained in any order.  This makes the bag handy for those cases when all you care about is that the data be consumed eventually, without regard for order of consumption or even fairness – that is, it’s possible new items could be consumed before older items given the right circumstances for a period of time. So why would you ever want a container that can be unfair?  Well, to look at it another way, you can use a ConcurrentQueue and get the fairness, but it comes at a cost in that the ordering rules and synchronization required to maintain that ordering can affect scalability a bit.  Thus sometimes the bag is great when you want the fastest way to get the next item to process, and don’t care what item it is or how long its been waiting. The way that the ConcurrentBag works is to take advantage of the new ThreadLocal<T> type (new in System.Threading for .NET 4.0) so that each thread using the bag has a list local to just that thread.  This means that adding or removing to a thread-local list requires very low synchronization.  The problem comes in where a thread goes to consume an item but it’s local list is empty.  In this case the bag performs “work-stealing” where it will rob an item from another thread that has items in its list.  This requires a higher level of synchronization which adds a bit of overhead to the take operation. So, as you can imagine, this makes the ConcurrentBag good for situations where each thread both produces and consumes items from the bag, but it would be less-than-idea in situations where some threads are dedicated producers and the other threads are dedicated consumers because the work-stealing synchronization would outweigh the thread-local optimization for a thread taking its own items. Like the other concurrent collections, there are some curiosities to keep in mind: IsEmpty(), Count, ToArray(), and GetEnumerator() lock collection Each of these needs to take a snapshot of whole bag to determine if empty, thus they tend to be more expensive and cause Add() and Take() operations to block. ToArray() and GetEnumerator() are static snapshots Because it is based on a snapshot, will not show subsequent updates after snapshot. Add() is lightweight Since adding to the thread-local list, there is very little overhead on Add. TryTake() is lightweight if items in thread-local list As long as items are in the thread-local list, TryTake() is very lightweight, much more so than ConcurrentStack() and ConcurrentQueue(), however if the local thread list is empty, it must steal work from another thread, which is more expensive. Remember, a bag is not ideal for all situations, it is mainly ideal for situations where a process consumes an item and either decomposes it into more items to be processed, or handles the item partially and places it back to be processed again until some point when it will complete.  The main point is that the bag works best when each thread both takes and adds items. For example, we could create a totally contrived example where perhaps we want to see the largest power of a number before it crosses a certain threshold.  Yes, obviously we could easily do this with a log function, but bare with me while I use this contrived example for simplicity. So let’s say we have a work function that will take a Tuple out of a bag, this Tuple will contain two ints.  The first int is the original number, and the second int is the last multiple of that number.  So we could load our bag with the initial values (let’s say we want to know the last multiple of each of 2, 3, 5, and 7 under 100. 1: var bag = new ConcurrentBag<Tuple<int, int>> 2: { 3: Tuple.Create(2, 1), 4: Tuple.Create(3, 1), 5: Tuple.Create(5, 1), 6: Tuple.Create(7, 1) 7: }; Then we can create a method that given the bag, will take out an item, apply the multiplier again, 1: public static void FindHighestPowerUnder(ConcurrentBag<Tuple<int,int>> bag, int threshold) 2: { 3: Tuple<int,int> pair; 4:  5: // while there are items to take, this will prefer local first, then steal if no local 6: while (bag.TryTake(out pair)) 7: { 8: // look at next power 9: var result = Math.Pow(pair.Item1, pair.Item2 + 1); 10:  11: if (result < threshold) 12: { 13: // if smaller than threshold bump power by 1 14: bag.Add(Tuple.Create(pair.Item1, pair.Item2 + 1)); 15: } 16: else 17: { 18: // otherwise, we're done 19: Console.WriteLine("Highest power of {0} under {3} is {0}^{1} = {2}.", 20: pair.Item1, pair.Item2, Math.Pow(pair.Item1, pair.Item2), threshold); 21: } 22: } 23: } Now that we have this, we can load up this method as an Action into our Tasks and run it: 1: // create array of tasks, start all, wait for all 2: var tasks = new[] 3: { 4: new Task(() => FindHighestPowerUnder(bag, 100)), 5: new Task(() => FindHighestPowerUnder(bag, 100)), 6: }; 7:  8: Array.ForEach(tasks, t => t.Start()); 9:  10: Task.WaitAll(tasks); Totally contrived, I know, but keep in mind the main point!  When you have a thread or task that operates on an item, and then puts it back for further consumption – or decomposes an item into further sub-items to be processed – you should consider a ConcurrentBag as the thread-local lists will allow for quick processing.  However, if you need ordering or if your processes are dedicated producers or consumers, this collection is not ideal.  As with anything, you should performance test as your mileage will vary depending on your situation! BlockingCollection<T> – A producers & consumers pattern collection The BlockingCollection<T> can be treated like a collection in its own right, but in reality it adds a producers and consumers paradigm to any collection that implements the interface IProducerConsumerCollection<T>.  If you don’t specify one at the time of construction, it will use a ConcurrentQueue<T> as its underlying store. If you don’t want to use the ConcurrentQueue, the ConcurrentStack and ConcurrentBag also implement the interface (though ConcurrentDictionary does not).  In addition, you are of course free to create your own implementation of the interface. So, for those who don’t remember the producers and consumers classical computer-science problem, the gist of it is that you have one (or more) processes that are creating items (producers) and one (or more) processes that are consuming these items (consumers).  Now, the crux of the problem is that there is a bin (queue) where the produced items are placed, and typically that bin has a limited size.  Thus if a producer creates an item, but there is no space to store it, it must wait until an item is consumed.  Also if a consumer goes to consume an item and none exists, it must wait until an item is produced. The BlockingCollection makes it trivial to implement any standard producers/consumers process set by providing that “bin” where the items can be produced into and consumed from with the appropriate blocking operations.  In addition, you can specify whether the bin should have a limited size or can be (theoretically) unbounded, and you can specify timeouts on the blocking operations. As far as your choice of “bin”, for the most part the ConcurrentQueue is the right choice because it is fairly light and maximizes fairness by ordering items so that they are consumed in the same order they are produced.  You can use the concurrent bag or stack, of course, but your ordering would be random-ish in the case of the former and LIFO in the case of the latter. So let’s look at some of the methods of note in BlockingCollection: BoundedCapacity returns capacity of the “bin” If the bin is unbounded, the capacity is int.MaxValue. Count returns an internally-kept count of items This makes it O(1), but if you modify underlying collection directly (not recommended) it is unreliable. CompleteAdding() is used to cut off further adds. This sets IsAddingCompleted and begins to wind down consumers once empty. IsAddingCompleted is true when producers are “done”. Once you are done producing, should complete the add process to alert consumers. IsCompleted is true when producers are “done” and “bin” is empty. Once you mark the producers done, and all items removed, this will be true. Add() is a blocking add to collection. If bin is full, will wait till space frees up Take() is a blocking remove from collection. If bin is empty, will wait until item is produced or adding is completed. GetConsumingEnumerable() is used to iterate and consume items. Unlike the standard enumerator, this one consumes the items instead of iteration. TryAdd() attempts add but does not block completely If adding would block, returns false instead, can specify TimeSpan to wait before stopping. TryTake() attempts to take but does not block completely Like TryAdd(), if taking would block, returns false instead, can specify TimeSpan to wait. Note the use of CompleteAdding() to signal the BlockingCollection that nothing else should be added.  This means that any attempts to TryAdd() or Add() after marked completed will throw an InvalidOperationException.  In addition, once adding is complete you can still continue to TryTake() and Take() until the bin is empty, and then Take() will throw the InvalidOperationException and TryTake() will return false. So let’s create a simple program to try this out.  Let’s say that you have one process that will be producing items, but a slower consumer process that handles them.  This gives us a chance to peek inside what happens when the bin is bounded (by default, the bin is NOT bounded). 1: var bin = new BlockingCollection<int>(5); Now, we create a method to produce items: 1: public static void ProduceItems(BlockingCollection<int> bin, int numToProduce) 2: { 3: for (int i = 0; i < numToProduce; i++) 4: { 5: // try for 10 ms to add an item 6: while (!bin.TryAdd(i, TimeSpan.FromMilliseconds(10))) 7: { 8: Console.WriteLine("Bin is full, retrying..."); 9: } 10: } 11:  12: // once done producing, call CompleteAdding() 13: Console.WriteLine("Adding is completed."); 14: bin.CompleteAdding(); 15: } And one to consume them: 1: public static void ConsumeItems(BlockingCollection<int> bin) 2: { 3: // This will only be true if CompleteAdding() was called AND the bin is empty. 4: while (!bin.IsCompleted) 5: { 6: int item; 7:  8: if (!bin.TryTake(out item, TimeSpan.FromMilliseconds(10))) 9: { 10: Console.WriteLine("Bin is empty, retrying..."); 11: } 12: else 13: { 14: Console.WriteLine("Consuming item {0}.", item); 15: Thread.Sleep(TimeSpan.FromMilliseconds(20)); 16: } 17: } 18: } Then we can fire them off: 1: // create one producer and two consumers 2: var tasks = new[] 3: { 4: new Task(() => ProduceItems(bin, 20)), 5: new Task(() => ConsumeItems(bin)), 6: new Task(() => ConsumeItems(bin)), 7: }; 8:  9: Array.ForEach(tasks, t => t.Start()); 10:  11: Task.WaitAll(tasks); Notice that the producer is faster than the consumer, thus it should be hitting a full bin often and displaying the message after it times out on TryAdd(). 1: Consuming item 0. 2: Consuming item 1. 3: Bin is full, retrying... 4: Bin is full, retrying... 5: Consuming item 3. 6: Consuming item 2. 7: Bin is full, retrying... 8: Consuming item 4. 9: Consuming item 5. 10: Bin is full, retrying... 11: Consuming item 6. 12: Consuming item 7. 13: Bin is full, retrying... 14: Consuming item 8. 15: Consuming item 9. 16: Bin is full, retrying... 17: Consuming item 10. 18: Consuming item 11. 19: Bin is full, retrying... 20: Consuming item 12. 21: Consuming item 13. 22: Bin is full, retrying... 23: Bin is full, retrying... 24: Consuming item 14. 25: Adding is completed. 26: Consuming item 15. 27: Consuming item 16. 28: Consuming item 17. 29: Consuming item 19. 30: Consuming item 18. Also notice that once CompleteAdding() is called and the bin is empty, the IsCompleted property returns true, and the consumers will exit. Summary The ConcurrentBag is an interesting collection that can be used to optimize concurrency scenarios where tasks or threads both produce and consume items.  In this way, it will choose to consume its own work if available, and then steal if not.  However, in situations where you want fair consumption or ordering, or in situations where the producers and consumers are distinct processes, the bag is not optimal. The BlockingCollection is a great wrapper around all of the concurrent queue, stack, and bag that allows you to add producer and consumer semantics easily including waiting when the bin is full or empty. That’s the end of my dive into the concurrent collections.  I’d also strongly recommend, once again, you read this excellent Microsoft white paper that goes into much greater detail on the efficiencies you can gain using these collections judiciously (here). Tweet Technorati Tags: C#,.NET,Concurrent Collections,Little Wonders

    Read the article

  • WebLogic Server JMS WLST Script – Who is Connected To My Server

    - by james.bayer
    Ever want to know who was connected to your WebLogic Server instance for troubleshooting?  An email exchange about this topic and JMS came up this week, and I’ve heard it come up once or twice before too.  Sometimes it’s interesting or helpful to know the list of JMS clients (IP Addresses, JMS Destinations, message counts) that are connected to a particular JMS server.  This can be helpful for troubleshooting.  Tom Barnes from the WebLogic Server JMS team provided some helpful advice: The JMS connection runtime mbean has “getHostAddress”, which returns the host address of the connecting client JVM as a string.  A connection runtime can contain session runtimes, which in turn can contain consumer runtimes.  The consumer runtime, in turn has a “getDestinationName” and “getMemberDestinationName”.  I think that this means you could write a WLST script, for example, to dump all consumers, their destinations, plus their parent session’s parent connection’s host addresses.    Note that the client runtime mbeans (connection, session, and consumer) won’t necessarily be hosted on the same JVM as a destination that’s in the same cluster (client messages route from their connection host to their ultimate destination in the same cluster). Writing the Script So armed with this information, I decided to take the challenge and see if I could write a WLST script to do this.  It’s always helpful to have the WebLogic Server MBean Reference handy for activities like this.  This one is focused on JMS Consumers and I only took a subset of the information available, but it could be modified easily to do Producers.  I haven’t tried this on a more complex environment, but it works in my simple sandbox case, so it should give you the general idea. # Better to use Secure Config File approach for login as shown here http://buttso.blogspot.com/2011/02/using-secure-config-files-with-weblogic.html connect('weblogic','welcome1','t3://localhost:7001')   # Navigate to the Server Runtime and get the Server Name serverRuntime() serverName = cmo.getName()   # Multiple JMS Servers could be hosted by a single WLS server cd('JMSRuntime/' + serverName + '.jms' ) jmsServers=cmo.getJMSServers()   # Find the list of all JMSServers for this server namesOfJMSServers = '' for jmsServer in jmsServers: namesOfJMSServers = jmsServer.getName() + ' '   # Count the number of connections jmsConnections=cmo.getConnections() print str(len(jmsConnections)) + ' JMS Connections found for ' + serverName + ' with JMSServers ' + namesOfJMSServers   # Recurse the MBean tree for each connection and pull out some information about consumers for jmsConnection in jmsConnections: try: print 'JMS Connection:' print ' Host Address = ' + jmsConnection.getHostAddress() print ' ClientID = ' + str( jmsConnection.getClientID() ) print ' Sessions Current = ' + str( jmsConnection.getSessionsCurrentCount() ) jmsSessions = jmsConnection.getSessions() for jmsSession in jmsSessions: jmsConsumers = jmsSession.getConsumers() for jmsConsumer in jmsConsumers: print ' Consumer:' print ' Name = ' + jmsConsumer.getName() print ' Messages Received = ' + str(jmsConsumer.getMessagesReceivedCount()) print ' Member Destination Name = ' + jmsConsumer.getMemberDestinationName() except: print 'Error retrieving JMS Consumer Information' dumpStack() # Cleanup disconnect() exit() Example Output I expect the output to look something like this and loop through all the connections, this is just the first one: 1 JMS Connections found for AdminServer with JMSServers myJMSServer JMS Connection:   Host Address = 127.0.0.1   ClientID = None   Sessions Current = 16    Consumer:      Name = consumer40      Messages Received = 1      Member Destination Name = myJMSModule!myQueue Notice that it has the IP Address of the client.  There are 16 Sessions open because I’m using an MDB, which defaults to 16 connections, so this matches what I expect.  Let’s see what the full output actually looks like: D:\Oracle\fmw11gr1ps3\user_projects\domains\offline_domain>java weblogic.WLST d:\temp\jms.py   Initializing WebLogic Scripting Tool (WLST) ...   Welcome to WebLogic Server Administration Scripting Shell   Type help() for help on available commands   Connecting to t3://localhost:7001 with userid weblogic ... Successfully connected to Admin Server 'AdminServer' that belongs to domain 'offline_domain'.   Warning: An insecure protocol was used to connect to the server. To ensure on-the-wire security, the SSL port or Admin port should be used instead.   Location changed to serverRuntime tree. This is a read-only tree with ServerRuntimeMBean as the root. For more help, use help(serverRuntime)   1 JMS Connections found for AdminServer with JMSServers myJMSServer JMS Connection: Host Address = 127.0.0.1 ClientID = None Sessions Current = 16 Consumer: Name = consumer40 Messages Received = 2 Member Destination Name = myJMSModule!myQueue Consumer: Name = consumer34 Messages Received = 2 Member Destination Name = myJMSModule!myQueue Consumer: Name = consumer37 Messages Received = 2 Member Destination Name = myJMSModule!myQueue Consumer: Name = consumer16 Messages Received = 2 Member Destination Name = myJMSModule!myQueue Consumer: Name = consumer46 Messages Received = 2 Member Destination Name = myJMSModule!myQueue Consumer: Name = consumer49 Messages Received = 2 Member Destination Name = myJMSModule!myQueue Consumer: Name = consumer43 Messages Received = 1 Member Destination Name = myJMSModule!myQueue Consumer: Name = consumer55 Messages Received = 1 Member Destination Name = myJMSModule!myQueue Consumer: Name = consumer25 Messages Received = 1 Member Destination Name = myJMSModule!myQueue Consumer: Name = consumer22 Messages Received = 1 Member Destination Name = myJMSModule!myQueue Consumer: Name = consumer19 Messages Received = 1 Member Destination Name = myJMSModule!myQueue Consumer: Name = consumer52 Messages Received = 1 Member Destination Name = myJMSModule!myQueue Consumer: Name = consumer31 Messages Received = 1 Member Destination Name = myJMSModule!myQueue Consumer: Name = consumer58 Messages Received = 1 Member Destination Name = myJMSModule!myQueue Consumer: Name = consumer28 Messages Received = 1 Member Destination Name = myJMSModule!myQueue Consumer: Name = consumer61 Messages Received = 1 Member Destination Name = myJMSModule!myQueue Disconnected from weblogic server: AdminServer     Exiting WebLogic Scripting Tool. Thanks to Tom Barnes for the hints and the inspiration to write this up. Image of telephone switchboard courtesy of http://www.JoeTourist.net/ JoeTourist InfoSystems

    Read the article

  • Managing Personal Projects As Solo Developer - Getting out of depth and failing projects

    - by James Jeffery
    I need some advice on project management. I start a project, and often times it will a large project for a solo developer. Usually its a web project. I handle everything from the UI, to the JS, PHP, server management etc. Half way in I feel out of my depth. I lose where I am, so I spend a couple of days away from the project to avoid the stress and before you know it, it becomes another unfinished project. I try to use frameworks and code libraries to make my developments easier on myself. Sometimes I will complete a project so it "works" and then go back and handle errors, design the UI properly and stuff. But without fail I will always end up out of my depth. I've though about outsourcing tasks such as the UI, and the behaviour, and focusing just on the PHP - which I feel is my strong point. But then pride kicks in, and I don't feel at one with a project I haven't completed myself. Does this make sense? I am sure there are many others who have felt like this either at home, or at work, and I would love some advice on managing my projects better.

    Read the article

  • C#: Optional Parameters - Pros and Pitfalls

    - by James Michael Hare
    When Microsoft rolled out Visual Studio 2010 with C# 4, I was very excited to learn how I could apply all the new features and enhancements to help make me and my team more productive developers. Default parameters have been around forever in C++, and were intentionally omitted in Java in favor of using overloading to satisfy that need as it was though that having too many default parameters could introduce code safety issues.  To some extent I can understand that move, as I’ve been bitten by default parameter pitfalls before, but at the same time I feel like Java threw out the baby with the bathwater in that move and I’m glad to see C# now has them. This post briefly discusses the pros and pitfalls of using default parameters.  I’m avoiding saying cons, because I really don’t believe using default parameters is a negative thing, I just think there are things you must watch for and guard against to avoid abuses that can cause code safety issues. Pro: Default Parameters Can Simplify Code Let’s start out with positives.  Consider how much cleaner it is to reduce all the overloads in methods or constructors that simply exist to give the semblance of optional parameters.  For example, we could have a Message class defined which allows for all possible initializations of a Message: 1: public class Message 2: { 3: // can either cascade these like this or duplicate the defaults (which can introduce risk) 4: public Message() 5: : this(string.Empty) 6: { 7: } 8:  9: public Message(string text) 10: : this(text, null) 11: { 12: } 13:  14: public Message(string text, IDictionary<string, string> properties) 15: : this(text, properties, -1) 16: { 17: } 18:  19: public Message(string text, IDictionary<string, string> properties, long timeToLive) 20: { 21: // ... 22: } 23: }   Now consider the same code with default parameters: 1: public class Message 2: { 3: // can either cascade these like this or duplicate the defaults (which can introduce risk) 4: public Message(string text = "", IDictionary<string, string> properties = null, long timeToLive = -1) 5: { 6: // ... 7: } 8: }   Much more clean and concise and no repetitive coding!  In addition, in the past if you wanted to be able to cleanly supply timeToLive and accept the default on text and properties above, you would need to either create another overload, or pass in the defaults explicitly.  With named parameters, though, we can do this easily: 1: var msg = new Message(timeToLive: 100);   Pro: Named Parameters can Improve Readability I must say one of my favorite things with the default parameters addition in C# is the named parameters.  It lets code be a lot easier to understand visually with no comments.  Think how many times you’ve run across a TimeSpan declaration with 4 arguments and wondered if they were passing in days/hours/minutes/seconds or hours/minutes/seconds/milliseconds.  A novice running through your code may wonder what it is.  Named arguments can help resolve the visual ambiguity: 1: // is this days/hours/minutes/seconds (no) or hours/minutes/seconds/milliseconds (yes) 2: var ts = new TimeSpan(1, 2, 3, 4); 3:  4: // this however is visually very explicit 5: var ts = new TimeSpan(days: 1, hours: 2, minutes: 3, seconds: 4);   Or think of the times you’ve run across something passing a Boolean literal and wondered what it was: 1: // what is false here? 2: var sub = CreateSubscriber(hostname, port, false); 3:  4: // aha! Much more visibly clear 5: var sub = CreateSubscriber(hostname, port, isBuffered: false);   Pitfall: Don't Insert new Default Parameters In Between Existing Defaults Now let’s consider a two potential pitfalls.  The first is really an abuse.  It’s not really a fault of the default parameters themselves, but a fault in the use of them.  Let’s consider that Message constructor again with defaults.  Let’s say you want to add a messagePriority to the message and you think this is more important than a timeToLive value, so you decide to put messagePriority before it in the default, this gives you: 1: public class Message 2: { 3: public Message(string text = "", IDictionary<string, string> properties = null, int priority = 5, long timeToLive = -1) 4: { 5: // ... 6: } 7: }   Oh boy have we set ourselves up for failure!  Why?  Think of all the code out there that could already be using the library that already specified the timeToLive, such as this possible call: 1: var msg = new Message(“An error occurred”, myProperties, 1000);   Before this specified a message with a TTL of 1000, now it specifies a message with a priority of 1000 and a time to live of -1 (infinite).  All of this with NO compiler errors or warnings. So the rule to take away is if you are adding new default parameters to a method that’s currently in use, make sure you add them to the end of the list or create a brand new method or overload. Pitfall: Beware of Default Parameters in Inheritance and Interface Implementation Now, the second potential pitfalls has to do with inheritance and interface implementation.  I’ll illustrate with a puzzle: 1: public interface ITag 2: { 3: void WriteTag(string tagName = "ITag"); 4: } 5:  6: public class BaseTag : ITag 7: { 8: public virtual void WriteTag(string tagName = "BaseTag") { Console.WriteLine(tagName); } 9: } 10:  11: public class SubTag : BaseTag 12: { 13: public override void WriteTag(string tagName = "SubTag") { Console.WriteLine(tagName); } 14: } 15:  16: public static class Program 17: { 18: public static void Main() 19: { 20: SubTag subTag = new SubTag(); 21: BaseTag subByBaseTag = subTag; 22: ITag subByInterfaceTag = subTag; 23:  24: // what happens here? 25: subTag.WriteTag(); 26: subByBaseTag.WriteTag(); 27: subByInterfaceTag.WriteTag(); 28: } 29: }   What happens?  Well, even though the object in each case is SubTag whose tag is “SubTag”, you will get: 1: SubTag 2: BaseTag 3: ITag   Why?  Because default parameter are resolved at compile time, not runtime!  This means that the default does not belong to the object being called, but by the reference type it’s being called through.  Since the SubTag instance is being called through an ITag reference, it will use the default specified in ITag. So the moral of the story here is to be very careful how you specify defaults in interfaces or inheritance hierarchies.  I would suggest avoiding repeating them, and instead concentrating on the layer of classes or interfaces you must likely expect your caller to be calling from. For example, if you have a messaging factory that returns an IMessage which can be either an MsmqMessage or JmsMessage, it only makes since to put the defaults at the IMessage level since chances are your user will be using the interface only. So let’s sum up.  In general, I really love default and named parameters in C# 4.0.  I think they’re a great tool to help make your code easier to read and maintain when used correctly. On the plus side, default parameters: Reduce redundant overloading for the sake of providing optional calling structures. Improve readability by being able to name an ambiguous argument. But remember to make sure you: Do not insert new default parameters in the middle of an existing set of default parameters, this may cause unpredictable behavior that may not necessarily throw a syntax error – add to end of list or create new method. Be extremely careful how you use default parameters in inheritance hierarchies and interfaces – choose the most appropriate level to add the defaults based on expected usage. Technorati Tags: C#,.NET,Software,Default Parameters

    Read the article

  • FAQ: Reshipping a Sun Certification

    - by Paul Sorensen
    If you obtained a Sun Certification before September 1, 2010, your success kit was shipped to the mailing address on record in your profile at certmanager.net/sun. At this time, if you require a reshipment for any reason of your original Sun Certification Certificate (it got damaged in the mail, you did not receive it, it needs to be sent to a different address, etc...), you will now receive access to an electronic reproduction of your original Certificate called an eCertificate. The Sun Certification ID card is no longer available and cannot be ordered from Oracle. This is to allow us to streamline this reshipping process and allow candidates to receive these certificates as quickly as possible.Candidates who earned a Java, Oracle Solaris, MySQL, NetBeans or OpenOffice.org certification on or after September 1, 2010 will receive their success kits in the mail within 6-8 weeks of completing the final certification requirement.

    Read the article

  • Business/Development Liaison Wanting To Enhance Understanding In Programming

    - by James Alexander
    I lead software development for a team of of about 20 devs and we're primarily a .net/sql server shop. We've recently created a new role in our organization for a more business like role to assist in prioritization of development and this business liaison has asked me if there are any books or resources he could use to better understand software concepts in a meaningful way. Any suggestions or advice would be greatly appreciated.

    Read the article

< Previous Page | 13 14 15 16 17 18 19 20 21 22 23 24  | Next Page >