Search Results

Search found 6805 results on 273 pages for 'variables'.

Page 156/273 | < Previous Page | 152 153 154 155 156 157 158 159 160 161 162 163  | Next Page >

  • Pass Extra Parameters to JavaScript Callback Function

    - by BRADINO
    Here is a simple example of a function that takes a callback function as a parameter. query.send(handleQueryResponse); function handleQueryResponse(response){      alert('Processing...'); } If you wanted to pass extra variables to the callback function, you can do it like this. var param1 = 'something'; var param2 ='something else'; query.send(function(response) { handleQueryResponse(response, param1, param2) }); function handleQueryResponse(response,param1,param2){      alert('Processing...');      alert(param1);      alert(param2); }

    Read the article

  • Investigation: Can different combinations of components effect Dataflow performance?

    - by jamiet
    Introduction The Dataflow task is one of the core components (if not the core component) of SQL Server Integration Services (SSIS) and often the most misunderstood. This is not surprising, its an incredibly complicated beast and we’re abstracted away from that complexity via some boxes that go yellow red or green and that have some lines drawn between them. Example dataflow In this blog post I intend to look under that facade and get into some of the nuts and bolts of the Dataflow Task by investigating how the decisions we make when building our packages can affect performance. I will do this by comparing the performance of three dataflows that all have the same input, all produce the same output, but which all operate slightly differently by way of having different transformation components. I also want to use this blog post to challenge a common held opinion that I see perpetuated over and over again on the SSIS forum. That is, that people assume adding components to a dataflow will be detrimental to overall performance. Its not surprising that people think this –it is intuitive to think that more components means more work- however this is not a view that I share. I have always been of the opinion that there are many factors affecting dataflow duration and the number of components is actually one of the less important ones; having said that I have never proven that assertion and that is one reason for this investigation. I have actually seen evidence that some people think dataflow duration is simply a function of number of rows and number of components. I’ll happily call that one out as a myth even without any investigation!  The Setup I have a 2GB datafile which is a list of 4731904 (~4.7million) customer records with various attributes against them and it contains 2 columns that I am going to use for categorisation: [YearlyIncome] [BirthDate] The data file is a SSIS raw format file which I chose to use because it is the quickest way of getting data into a dataflow and given that I am testing the transformations, not the source or destination adapters, I want to minimise external influences as much as possible. In the test I will split the customers according to month of birth (12 of those) and whether or not their yearly income is above or below 50000 (2 of those); in other words I will be splitting them into 24 discrete categories and in order to do it I shall be using different combinations of SSIS’ Conditional Split and Derived Column transformation components. The 24 datapaths that occur will each input to a rowcount component, again because this is the least resource intensive means of terminating a datapath. The test is being carried out on a Dell XPS Studio laptop with a quad core (8 logical Procs) Intel Core i7 at 1.73GHz and Samsung SSD hard drive. Its running SQL Server 2008 R2 on Windows 7. The Variables Here are the three combinations of components that I am going to test:     One Conditional Split - A single Conditional Split component CSPL Split by Month of Birth and income category that will use expressions on [YearlyIncome] & [BirthDate] to send each row to one of 24 outputs. This next screenshot displays the expression logic in use: Derived Column & Conditional Split - A Derived Column component DER Income Category that adds a new column [IncomeCategory] which will contain one of two possible text values {“LessThan50000”,”GreaterThan50000”} and uses [YearlyIncome] to determine which value each row should get. A Conditional Split component CSPL Split by Month of Birth and Income Category then uses that new column in conjunction with [BirthDate] to determine which of the same 24 outputs to send each row to. Put more simply, I am separating the Conditional Split of #1 into a Derived Column and a Conditional Split. The next screenshots display the expression logic in use: DER Income Category         CSPL Split by Month of Birth and Income Category       Three Conditional Splits - A Conditional Split component that produces two outputs based on [YearlyIncome], one for each Income Category. Each of those outputs will go to a further Conditional Split that splits the input into 12 outputs, one for each month of birth (identical logic in each). In this case then I am separating the single Conditional Split of #1 into three Conditional Split components. The next screenshots display the expression logic in use: CSPL Split by Income Category         CSPL Split by Month of Birth 1& 2       Each of these combinations will provide an input to one of the 24 rowcount components, just the same as before. For illustration here is a screenshot of the dataflow containing three Conditional Split components: As you can these dataflows have a fair bit of work to do and remember that they’re doing that work for 4.7million rows. I will execute each dataflow 10 times and use the average for comparison. I foresee three possible outcomes: The dataflow containing just one Conditional Split (i.e. #1) will be quicker There is no significant difference between any of them One of the two dataflows containing multiple transformation components will be quicker Regardless of which of those outcomes come to pass we will have learnt something and that makes this an interesting test to carry out. Note that I will be executing the dataflows using dtexec.exe rather than hitting F5 within BIDS. The Results and Analysis The table below shows all of the executions, 10 for each dataflow. It also shows the average for each along with a standard deviation. All durations are in seconds. I’m pasting a screenshot because I frankly can’t be bothered with the faffing about needed to make a presentable HTML table. It is plain to see from the average that the dataflow containing three conditional splits is significantly faster, the other two taking 43% and 52% longer respectively. This seems strange though, right? Why does the dataflow containing the most components outperform the other two by such a big margin? The answer is actually quite logical when you put some thought into it and I’ll explain that below. Before progressing, a side note. The standard deviation for the “Three Conditional Splits” dataflow is orders of magnitude smaller – indicating that performance for this dataflow can be predicted with much greater confidence too. The Explanation I refer you to the screenshot above that shows how CSPL Split by Month of Birth and salary category in the first dataflow is setup. Observe that there is a case for each combination of Month Of Date and Income Category – 24 in total. These expressions get evaluated in the order that they appear and hence if we assume that Month of Date and Income Category are uniformly distributed in the dataset we can deduce that the expected number of expression evaluations for each row is 12.5 i.e. 1 (the minimum) + 24 (the maximum) divided by 2 = 12.5. Now take a look at the screenshots for the second dataflow. We are doing one expression evaluation in DER Income Category and we have the same 24 cases in CSPL Split by Month of Birth and Income Category as we had before, only the expression differs slightly. In this case then we have 1 + 12.5 = 13.5 expected evaluations for each row – that would account for the slightly longer average execution time for this dataflow. Now onto the third dataflow, the quick one. CSPL Split by Income Category does a maximum of 2 expression evaluations thus the expected number of evaluations per row is 1.5. CSPL Split by Month of Birth 1 & CSPL Split by Month of Birth 2 both have less work to do than the previous Conditional Split components because they only have 12 cases to test for thus the expected number of expression evaluations is 6.5 There are two of them so total expected number of expression evaluations for this dataflow is 6.5 + 6.5 + 1.5 = 14.5. 14.5 is still more than 12.5 & 13.5 though so why is the third dataflow so much quicker? Simple, the conditional expressions in the first two dataflows have two boolean predicates to evaluate – one for Income Category and one for Month of Birth; the expressions in the Conditional Split in the third dataflow however only have one predicate thus they are doing a lot less work. To sum up, the difference in execution times can be attributed to the difference between: MONTH(BirthDate) == 1 && YearlyIncome <= 50000 and MONTH(BirthDate) == 1 In the first two dataflows YearlyIncome <= 50000 gets evaluated an average of 12.5 times for every row whereas in the third dataflow it is evaluated once and once only. Multiply those 11.5 extra operations by 4.7million rows and you get a significant amount of extra CPU cycles – that’s where our duration difference comes from. The Wrap-up The obvious point here is that adding new components to a dataflow isn’t necessarily going to make it go any slower, moreover you may be able to achieve significant improvements by splitting logic over multiple components rather than one. Performance tuning is all about reducing the amount of work that needs to be done and that doesn’t necessarily mean use less components, indeed sometimes you may be able to reduce workload in ways that aren’t immediately obvious as I think I have proven here. Of course there are many variables in play here and your mileage will most definitely vary. I encourage you to download the package and see if you get similar results – let me know in the comments. The package contains all three dataflows plus a fourth dataflow that will create the 2GB raw file for you (you will also need the [AdventureWorksDW2008] sample database from which to source the data); simply disable all dataflows except the one you want to test before executing the package and remember, execute using dtexec, not within BIDS. If you want to explore dataflow performance tuning in more detail then here are some links you might want to check out: Inequality joins, Asynchronous transformations and Lookups Destination Adapter Comparison Don’t turn the dataflow into a cursor SSIS Dataflow – Designing for performance (webinar) Any comments? Let me know! @Jamiet

    Read the article

  • Solaris X86 64-bit Assembly Programming

    - by danx
    Solaris X86 64-bit Assembly Programming This is a simple example on writing, compiling, and debugging Solaris 64-bit x86 assembly language with a C program. This is also referred to as "AMD64" assembly. The term "AMD64" is used in an inclusive sense to refer to all X86 64-bit processors, whether AMD Opteron family or Intel 64 processor family. Both run Solaris x86. I'm keeping this example simple mainly to illustrate how everything comes together—compiler, assembler, linker, and debugger when using assembly language. The example I'm using here is a C program that calls an assembly language program passing a C string. The assembly language program takes the C string and calls printf() with it to print the string. AMD64 Register Usage But first let's review the use of AMD64 registers. AMD64 has several 64-bit registers, some special purpose (such as the stack pointer) and others general purpose. By convention, Solaris follows the AMD64 ABI in register usage, which is the same used by Linux, but different from Microsoft Windows in usage (such as which registers are used to pass parameters). This blog will only discuss conventions for Linux and Solaris. The following chart shows how AMD64 registers are used. The first six parameters to a function are passed through registers. If there's more than six parameters, parameter 7 and above are pushed on the stack before calling the function. The stack is also used to save temporary "stack" variables for use by a function. 64-bit Register Usage %rip Instruction Pointer points to the current instruction %rsp Stack Pointer %rbp Frame Pointer (saved stack pointer pointing to parameters on stack) %rdi Function Parameter 1 %rsi Function Parameter 2 %rdx Function Parameter 3 %rcx Function Parameter 4 %r8 Function Parameter 5 %r9 Function Parameter 6 %rax Function return value %r10, %r11 Temporary registers (need not be saved before used) %rbx, %r12, %r13, %r14, %r15 Temporary registers, but must be saved before use and restored before returning from the current function (usually with the push and pop instructions). 32-, 16-, and 8-bit registers To access the lower 32-, 16-, or 8-bits of a 64-bit register use the following: 64-bit register Least significant 32-bits Least significant 16-bits Least significant 8-bits %rax%eax%ax%al %rbx%ebx%bx%bl %rcx%ecx%cx%cl %rdx%edx%dx%dl %rsi%esi%si%sil %rdi%edi%di%axl %rbp%ebp%bp%bp %rsp%esp%sp%spl %r9%r9d%r9w%r9b %r10%r10d%r10w%r10b %r11%r11d%r11w%r11b %r12%r12d%r12w%r12b %r13%r13d%r13w%r13b %r14%r14d%r14w%r14b %r15%r15d%r15w%r15b %r16%r16d%r16w%r16b There's other registers present, such as the 64-bit %mm registers, 128-bit %xmm registers, 256-bit %ymm registers, and 512-bit %zmm registers. Except for %mm registers, these registers may not present on older AMD64 processors. Assembly Source The following is the source for a C program, helloas1.c, that calls an assembly function, hello_asm(). $ cat helloas1.c extern void hello_asm(char *s); int main(void) { hello_asm("Hello, World!"); } The assembly function called above, hello_asm(), is defined below. $ cat helloas2.s /* * helloas2.s * To build: * cc -m64 -o helloas2-cpp.s -D_ASM -E helloas2.s * cc -m64 -c -o helloas2.o helloas2-cpp.s */ #if defined(lint) || defined(__lint) /* ARGSUSED */ void hello_asm(char *s) { } #else /* lint */ #include <sys/asm_linkage.h> .extern printf ENTRY_NP(hello_asm) // Setup printf parameters on stack mov %rdi, %rsi // P2 (%rsi) is string variable lea .printf_string, %rdi // P1 (%rdi) is printf format string call printf ret SET_SIZE(hello_asm) // Read-only data .text .align 16 .type .printf_string, @object .printf_string: .ascii "The string is: %s.\n\0" #endif /* lint || __lint */ In the assembly source above, the C skeleton code under "#if defined(lint)" is optionally used for lint to check the interfaces with your C program--very useful to catch nasty interface bugs. The "asm_linkage.h" file includes some handy macros useful for assembly, such as ENTRY_NP(), used to define a program entry point, and SET_SIZE(), used to set the function size in the symbol table. The function hello_asm calls C function printf() by passing two parameters, Parameter 1 (P1) is a printf format string, and P2 is a string variable. The function begins by moving %rdi, which contains Parameter 1 (P1) passed hello_asm, to printf()'s P2, %rsi. Then it sets printf's P1, the format string, by loading the address the address of the format string in %rdi, P1. Finally it calls printf. After returning from printf, the hello_asm function returns itself. Larger, more complex assembly functions usually do more setup than the example above. If a function is returning a value, it would set %rax to the return value. Also, it's typical for a function to save the %rbp and %rsp registers of the calling function and to restore these registers before returning. %rsp contains the stack pointer and %rbp contains the frame pointer. Here is the typical function setup and return sequence for a function: ENTRY_NP(sample_assembly_function) push %rbp // save frame pointer on stack mov %rsp, %rbp // save stack pointer in frame pointer xor %rax, %r4ax // set function return value to 0. mov %rbp, %rsp // restore stack pointer pop %rbp // restore frame pointer ret // return to calling function SET_SIZE(sample_assembly_function) Compiling and Running Assembly Use the Solaris cc command to compile both C and assembly source, and to pre-process assembly source. You can also use GNU gcc instead of cc to compile, if you prefer. The "-m64" option tells the compiler to compile in 64-bit address mode (instead of 32-bit). $ cc -m64 -o helloas2-cpp.s -D_ASM -E helloas2.s $ cc -m64 -c -o helloas2.o helloas2-cpp.s $ cc -m64 -c helloas1.c $ cc -m64 -o hello-asm helloas1.o helloas2.o $ file hello-asm helloas1.o helloas2.o hello-asm: ELF 64-bit LSB executable AMD64 Version 1 [SSE FXSR FPU], dynamically linked, not stripped helloas1.o: ELF 64-bit LSB relocatable AMD64 Version 1 helloas2.o: ELF 64-bit LSB relocatable AMD64 Version 1 $ hello-asm The string is: Hello, World!. Debugging Assembly with MDB MDB is the Solaris system debugger. It can also be used to debug user programs, including assembly and C. The following example runs the above program, hello-asm, under control of the debugger. In the example below I load the program, set a breakpoint at the assembly function hello_asm, display the registers and the first parameter, step through the assembly function, and continue execution. $ mdb hello-asm # Start the debugger > hello_asm:b # Set a breakpoint > ::run # Run the program under the debugger mdb: stop at hello_asm mdb: target stopped at: hello_asm: movq %rdi,%rsi > $C # display function stack ffff80ffbffff6e0 hello_asm() ffff80ffbffff6f0 0x400adc() > $r # display registers %rax = 0x0000000000000000 %r8 = 0x0000000000000000 %rbx = 0xffff80ffbf7f8e70 %r9 = 0x0000000000000000 %rcx = 0x0000000000000000 %r10 = 0x0000000000000000 %rdx = 0xffff80ffbffff718 %r11 = 0xffff80ffbf537db8 %rsi = 0xffff80ffbffff708 %r12 = 0x0000000000000000 %rdi = 0x0000000000400cf8 %r13 = 0x0000000000000000 %r14 = 0x0000000000000000 %r15 = 0x0000000000000000 %cs = 0x0053 %fs = 0x0000 %gs = 0x0000 %ds = 0x0000 %es = 0x0000 %ss = 0x004b %rip = 0x0000000000400c70 hello_asm %rbp = 0xffff80ffbffff6e0 %rsp = 0xffff80ffbffff6c8 %rflags = 0x00000282 id=0 vip=0 vif=0 ac=0 vm=0 rf=0 nt=0 iopl=0x0 status=<of,df,IF,tf,SF,zf,af,pf,cf> %gsbase = 0x0000000000000000 %fsbase = 0xffff80ffbf782a40 %trapno = 0x3 %err = 0x0 > ::dis # disassemble the current instructions hello_asm: movq %rdi,%rsi hello_asm+3: leaq 0x400c90,%rdi hello_asm+0xb: call -0x220 <PLT:printf> hello_asm+0x10: ret 0x400c81: nop 0x400c85: nop 0x400c88: nop 0x400c8c: nop 0x400c90: pushq %rsp 0x400c91: pushq $0x74732065 0x400c96: jb +0x69 <0x400d01> > 0x0000000000400cf8/S # %rdi contains Parameter 1 0x400cf8: Hello, World! > [ # Step and execute 1 instruction mdb: target stopped at: hello_asm+3: leaq 0x400c90,%rdi > [ mdb: target stopped at: hello_asm+0xb: call -0x220 <PLT:printf> > [ The string is: Hello, World!. mdb: target stopped at: hello_asm+0x10: ret > [ mdb: target stopped at: main+0x19: movl $0x0,-0x4(%rbp) > :c # continue program execution mdb: target has terminated > $q # quit the MDB debugger $ In the example above, at the start of function hello_asm(), I display the stack contents with "$C", display the registers contents with "$r", then disassemble the current function with "::dis". The first function parameter, which is a C string, is passed by reference with the string address in %rdi (see the register usage chart above). The address is 0x400cf8, so I print the value of the string with the "/S" MDB command: "0x0000000000400cf8/S". I can also print the contents at an address in several other formats. Here's a few popular formats. For more, see the mdb(1) man page for details. address/S C string address/C ASCII character (1 byte) address/E unsigned decimal (8 bytes) address/U unsigned decimal (4 bytes) address/D signed decimal (4 bytes) address/J hexadecimal (8 bytes) address/X hexadecimal (4 bytes) address/B hexadecimal (1 bytes) address/K pointer in hexadecimal (4 or 8 bytes) address/I disassembled instruction Finally, I step through each machine instruction with the "[" command, which steps over functions. If I wanted to enter a function, I would use the "]" command. Then I continue program execution with ":c", which continues until the program terminates. MDB Basic Cheat Sheet Here's a brief cheat sheet of some of the more common MDB commands useful for assembly debugging. There's an entire set of macros and more powerful commands, especially some for debugging the Solaris kernel, but that's beyond the scope of this example. $C Display function stack with pointers $c Display function stack $e Display external function names $v Display non-zero variables and registers $r Display registers ::fpregs Display floating point (or "media" registers). Includes %st, %xmm, and %ymm registers. ::status Display program status ::run Run the program (followed by optional command line parameters) $q Quit the debugger address:b Set a breakpoint address:d Delete a breakpoint $b Display breakpoints :c Continue program execution after a breakpoint [ Step 1 instruction, but step over function calls ] Step 1 instruction address::dis Disassemble instructions at an address ::events Display events Further Information "Assembly Language Techniques for Oracle Solaris on x86 Platforms" by Paul Lowik (2004). Good tutorial on Solaris x86 optimization with assembly. The Solaris Operating System on x86 Platforms An excellent, detailed tutorial on X86 architecture, with Solaris specifics. By an ex-Sun employee, Frank Hofmann (2005). "AMD64 ABI Features", Solaris 64-bit Developer's Guide contains rules on data types and register usage for Intel 64/AMD64-class processors. (available at docs.oracle.com) Solaris X86 Assembly Language Reference Manual (available at docs.oracle.com) SPARC Assembly Language Reference Manual (available at docs.oracle.com) System V Application Binary Interface (2003) defines the AMD64 ABI for UNIX-class operating systems, including Solaris, Linux, and BSD. Google for it—the original website is gone. cc(1), gcc(1), and mdb(1) man pages.

    Read the article

  • How can I promote clean coding at my workplace?

    - by Michael
    I work with a lot of legacy Java and RPG code on an internal company application. As you might expect, a lot of the code is written in many different styles, and often is difficult to read because of poorly named variables, inconsistent formatting, and contradictory comments (if they're there at all). Also, a good amount of code is not robust. Many times code is pushed to production quickly by the more experienced programmers, while code by newer programmers is held back by "code reviews" that IMO are unsatisfactory. (They usually take the form of, "It works, must be ok," than a serious critique of the code.) We have a fair number of production issues, which I feel could be lessened by giving more thought to the original design and testing. I have been working for this company for about 4 months, and have been complimented on my coding style a couple of times. My manager is also a fan of cleaner coding than is the norm. Is it my place to try to push for better style and better defensive coding, or should I simply code in the best way I can, and hope that my example will help others see how cleaner, more robust code (as well as aggressive refactoring) will result in less debugging and change time?

    Read the article

  • What's the benefit of object-oriented programming over procedural programming?

    - by niko
    I'm trying to understand the difference between procedural languages like C and object-oriented languages like C++. I've never used C++, but I've been discussing with my friends on how to differentiate the two. I've been told C++ has object-oriented concepts as well as public and private modes for definition of variables: things C does not have. I've never had to use these for while developing programs in Visual Basic.NET: what are the benefits of these? I've also been told that if a variable is public, it can be accessed anywhere, but it's not clear how that's different from a global variable in a language like C. It's also not clear how a private variable differs from a local variable. Another thing I've heard is that, for security reasons, if a function needs to be accessed it should be inherited first. The use-case is that an administrator should only have as much rights as they need and not everything, but it seems a conditional would work as well: if ( login == "admin") { // invoke the function } Why is this not ideal? Given that there seems to be a procedural way to do everything object-oriented, why should I care about object-oriented programming?

    Read the article

  • 45° Slopes in a Tile based 2D platformer

    - by xNidhogg
    I want to have simple 45° slopes in my tile based platformer, however I just cant seem to get the algorithm down. Please take a look at the code and video, maybe I'm missing the obvious? //collisionRectangle is the collision rectangle of the player with //origin at the top left and width and height //wantedPosition is the new position the player will be set to. //this is determined elsewhere by checking the bottom center point of the players rect if(_leftSlope || _rightSlope) { //Test bottom center point var calculationPoint = new Vector2(collisionRectangle.Center.X, collisionRectangle.Bottom); //Get the collision rectangle of the tile, origin is top-left Rectangle cellRect = _tileMap.CellWorldRectangle( _tileMap.GetCellByPixel(calculationPoint)); //Calculate the new Y coordinate depending on if its a left or right slope //CellSize = 8 float newY = _leftSlope ? (calculationPoint.X % CellSize) + cellRect.Y : (-1 * (calculationPoint.X % CellSize) - CellSize) + cellRect.Y; //reset variables so we dont jump in here next frame _leftSlope = false; _rightSlope = false; //now change the players Y according to the difference of our calculation wantedPosition.Y += newY - calculationPoint.Y; } Video of what it looks like: http://youtu.be/EKOWgD2muoc

    Read the article

  • How to get the path of a file after publishing my game

    - by NDraskovic
    I made a "game" for a college project that reads data from .txt file at startup and draws some models according to the data in that file. This is the code I use using (StreamReader sr = new StreamReader(@"C:\Users\User\Desktop\Linije.txt")) { String linija; while ((linija = sr.ReadLine()) != null) { red = linija.Split(','); model = red[0]; x = red[1]; y = red[2]; z = red[3]; elementi.Add(Convert.ToInt32(model)); podatci.Add(new Vector3(Convert.ToSingle(x),Convert.ToSingle(y),Convert.ToSingle(z))); } } As you see, this code fills some variables that are then used to define the model that will be drawn and the coordinates where it will be drawn. The problem that I'm having is that I don't know how to distribute that file to other computers (obviously on another computer it would have another path)? Do you have some advices on how to do this? P.S I tried to put it in the Content and set the Build Action on None, and I can see the file in the content directory, but when I change it, nothing happens (the models don't change as they should)

    Read the article

  • Convert your Hash keys to object properties in Ruby

    - by kerry
    Being a Ruby noob (and having a background in Groovy), I was a little surprised that you can not access hash objects using the dot notation.  I am writing an application that relies heavily on XML and JSON data.  This data will need to be displayed and I would rather use book.author.first_name over book[‘author’][‘first_name’].  A quick search on google yielded this post on the subject. So, taking the DRYOO (Don’t Repeat Yourself Or Others) concept.  I came up with this: 1: class ::Hash 2:  3: # add keys to hash 4: def to_obj 5: self.each do |k,v| 6: if v.kind_of? Hash 7: v.to_obj 8: end 9: k=k.gsub(/\.|\s|-|\/|\'/, '_').downcase.to_sym 10: self.instance_variable_set("@#{k}", v) ## create and initialize an instance variable for this key/value pair 11: self.class.send(:define_method, k, proc{self.instance_variable_get("@#{k}")}) ## create the getter that returns the instance variable 12: self.class.send(:define_method, "#{k}=", proc{|v| self.instance_variable_set("@#{k}", v)}) ## create the setter that sets the instance variable 13: end 14: return self 15: end 16: end This works pretty well.  It converts each of your keys to properties of the Hash.  However, it doesn’t sit very well with me because I probably will not use 90% of the properties most of the time.  Why should I go through the performance overhead of creating instance variables for all of the unused ones? Enter the ‘magic method’ #missing_method: 1: class ::Hash 2: def method_missing(name) 3: return self[name] if key? name 4: self.each { |k,v| return v if k.to_s.to_sym == name } 5: super.method_missing name 6: end 7: end This is a much cleaner method for my purposes.  Quite simply, it checks to see if there is a key with the given symbol, and if not, loop through the keys and attempt to find one. I am a Ruby noob, so if there is something I am overlooking, please let me know.

    Read the article

  • General directions on developing a server side control system for JS/Canvas Action RPG

    - by Billy Ninja
    Well, yesterday I asked on anti-cheat JS, and confirmed what I kind of already knew that it's just not possible. Now I wanna measure roughly how hard it is to implement a server side checking that is agnostic to client input, that does not mess with the game experience so much. I don't wanna waste to much resource on this matter, since it's going to be initially a single player game, that I may or would like to introduce some kind of ranking, trading system later on. I'd rather deliver better more cool game features instead. I don't wanna have to guarantee super fast server response to keep the game going lag free. I'd rather go with more loose discrete control of key variables and instances. Like store user's action on a fifo buffer on the client, and push that actions to the server gradually. I'd love to see a elegant, generic solution that I could plug into my client game logic root (not having to scatter treatments everywhere in my client js) - and have few classes on Node.js server that could handle that - without having to mirror/describe all of my game entities a second time on the server.

    Read the article

  • Where can I find video resources of people programming?

    - by Corey
    This might be a strange question. I'm looking for videos of people actively coding something while explaining it. However, I don't want is a beginner video that delves into what variables and objects are. Nick Gravelyn's tile engine tutorial is a great example of what I'm looking for. (He actually used to host the full, unbroken video files in his site's archive, but I guess he took them down...) I tend to learn best by "action" examples; it's difficult for me to learn by reading through documentation and text tutorials, but if I see somebody actively doing a task, I can immediately register it and apply it myself. I'm hard-of-hearing, so I would really prefer that if the video has a lot of talking, it have captioning or subtitling of some sort, or at the very least, a transcript. The tile engine videos did not have captions, but the code he was writing was very self-documenting, so I understood it for the most part. I've gone through most of the relevant GoogleDevelopers and GoogleTechTalks videos on Youtube, so those need not apply. Are there any resources out there, or even websites dedicated to this kind of thing?

    Read the article

  • Fibonacci numbers in F#

    - by BobPalmer
    As you may have gathered from some of my previous posts, I've been spending some quality time at Project Euler.  Normally I do my solutions in C#, but since I have also started learning F#, it only made sense to switch over to F# to get my math coding fix. This week's post is just a small snippet - spefically, a simple function to return a fibonacci number given it's place in the sequence.  One popular example uses recursion: let rec fib n = if n < 2 then 1 else fib (n-2) + fib(n-1) While this is certainly elegant, the recursion is absolutely brutal on performance.  So I decided to spend a little time, and find an option that achieved the same functionality, but used a recursive function.  And since this is F#, I wanted to make sure I did it without the use of any mutable variables. Here's the solution I came up with: let rec fib n1 n2 c =    if c = 1 then        n2    else        fib n2 (n1+n2) (c-1);;let GetFib num =    (fib 1 1 num);;printfn "%A" (GetFib 1000);; Essentially, this function works through the sequence moving forward, passing the two most recent numbers and a counter to the recursive calls until it has achieved the desired number of iterations.  At that point, it returns the latest fibonacci number. Enjoy!

    Read the article

  • How should data be passed between client-side Javascript and C# code behind an ASP.NET app?

    - by ctck
    I'm looking for the most efficient / standard way of passing data between client-side Javascript code and C# code behind an ASP.NET application. I've been using the following methods to achieve this but they all feel a bit of a fudge. To pass data from Javascript to the C# code is by setting hidden ASP variables and triggering a postback: <asp:HiddenField ID="RandomList" runat="server" /> function SetDataField(data) { document.getElementById('<%=RandomList.ClientID%>').value = data; } Then in the C# code I collect the list: protected void GetData(object sender, EventArgs e) { var _list = RandomList.value; } Going back the other way I often use either ScriptManager to register a function and pass it data during Page_Load: ScriptManager.RegisterStartupScript(this.GetType(), "Set","get("Test();",true); or I add attributes to controls before a post back or during the initialization or pre-rendering stages: Btn.Attributes.Add("onclick", "DisplayMessage("Hello");"); These methods have served me well and do the job, but they just dont feel complete. Is there a more standard way of passing data between client side Javascript and C# backend code? Ive seen some posts like this one that describe HtmlElement class; is this something I should look into?

    Read the article

  • Why is HTML/Javascript minification beneficial

    - by Channel72
    Why is HTML/Javascript minification beneficial when the HTTP protocol already supports gzip data compression? I realize that Javascript/HTML minification has the potential to significantly reduce the size of Javascript/HTML files by removing unnecessary whitespace, and perhaps renaming variables to a few letters each, but doesn't the LZW algorithm do especially well when there are many repeated characters (e.g. lots of whitespace?) I realize that some Javascript minification tools do more than just reduce size. Google's closure compiler, for example, also tries to improve code performance by inlining functions and doing other analyses. But the primary purpose of Javascript minification is usually to reduce file size. I also realize there are other reasons you might want to minify aside from performace, such as code obfuscation. But again, that reason is not usually emphasized as much as performance gain and file size reduction. For example, Closure Compiler is not advertised as an obfuscation tool, but as a code size reducer and download-speed enhancer. So, how much performance do you really gain from Javascript/HTML minification when you're already significantly reducing file size with gzip compression?

    Read the article

  • Performance-Driven Development

    - by BuckWoody
    I was reading a blog yesterday about the evils of SELECT *. The author pointed out that it's almost always a bad idea to use SELECT * for a query, but in the case of SQL Azure (or any cloud database, for that matter) it's especially bad, since you're paying for each transmission that comes down the line. A very good point indeed. This got me to thinking - shouldn't we treat ALL programming that way? In other words, wouldn't it make sense to pretend that we are paying for every chunk of data - a little less for a bit, a lot more for a BLOB or VARCHAR(MAX), that sort of thing? In effect, we really are paying for that. Which led me to the thought of Performance-Driven Development, or the act of programming with the goal of having the fastest code from the very outset. This isn't an original title, since a quick Bing-search shows me a couple of offerings from Forrester and a professional in Israel who already used that title, but the general idea I'm thinking of is assigning a "cost" to each code round-trip, be it network, storage, trip time and other variables, and then rewarding the developers that come up with the fastest code. I wonder what kind of throughput and round-trip times you could get if your developers were paid on a scale of how fast the application performed... Share this post: email it! | bookmark it! | digg it! | reddit! | kick it! | live it!

    Read the article

  • Get Started using Build-Deploy-Test Workflow with TFS 2012

    - by Jakob Ehn
    TFS 2012 introduces a new type of Lab environment called Standard Environment. This allows you to setup a full Build Deploy Test (BDT) workflow that will build your application, deploy it to your target machine(s) and then run a set of tests on that server to verify the deployment. In TFS 2010, you had to use System Center Virtual Machine Manager and involve half of your IT department to get going. Now all you need is a server (virtual or physical) where you want to deploy and test your application. You don’t even have to install a test agent on the machine, TFS 2012 will do this for you! Although each step is rather simple, the entire process of setting it up consists of a bunch of steps. So I thought that it could be useful to run through a typical setup.I will also link to some good guidance from MSDN on each topic. High Level Steps Install and configure Visual Studio 2012 Test Controller on Target Server Create Standard Environment Create Test Plan with Test Case Run Test Case Create Coded UI Test from Test Case Associate Coded UI Test with Test Case Create Build Definition using LabDefaultTemplate 1. Install and Configure Visual Studio 2012 Test Controller on Target Server First of all, note that you do not have to have the Test Controller running on the target server. It can be running on another server, as long as the Test Agent can communicate with the test controller and the test controller can communicate with the TFS server. If you have several machines in your environment (web server, database server etc..), the test controller can be installed either on one of those machines or on a dedicated machine. To install the test controller, simply mount the Visual Studio Agents media on the server and browse to the vstf_controller.exe file located in the TestController folder. Run through the installation, you might need to reboot the server since it installs .NET 4.5. When the test controller is installed, the Test Controller configuration tool will launch automatically (if it doesn’t, you can start it from the Start menu). Here you will supply the credentials of the account running the test controller service. Note that this account will be given the necessary permissions in TFS during the configuration. Make sure that you have entered a valid account by pressing the Test link. Also, you have to register the test controller with the TFS collection where your test plan is located (and usually the code base of course) When you press Apply Settings, all the configuration will be done. You might get some warnings at the end, that might or might not cause a problem later. Be sure to read them carefully.   For more information about configuring your test controllers, see Setting Up Test Controllers and Test Agents to Manage Tests with Visual Studio 2. Create Standard Environment Now you need to create a Lab environment in Microsoft Test Manager. Since we are using an existing physical or virtual machine we will create a Standard Environment. Open MTM and go to Lab Center. Click New to create a new environment Enter a name for the environment. Since this environment will only contain one machine, we will use the machine name for the environment (TargetServer in this case) On the next page, click Add to add a machine to the environment. Enter the name of the machine (TargetServer.Domain.Com), and give it the Web Server role. The name must be reachable both from your machine during configuration and from the TFS app tier server. You also need to supply an account that is a local administration on the target server. This is needed in order to automatically install a test agent later on the machine. On the next page, you can add tags to the machine. This is not needed in this scenario so go to the next page. Here you will specify which test controller to use and that you want to run UI tests on this environment. This will in result in a Test Agent being automatically installed and configured on the target server. The name of the machine where you installed the test controller should be available on the drop down list (TargetServer in this sample). If you can’t see it, you might have selected a different TFS project collection. Press Next twice and then Verify to verify all the settings: Press finish. This will now create and prepare the environment, which means that it will remote install a test agent on the machine. As part of this installation, the remote server will be restarted. 3-5. Create Test Plan, Run Test Case, Create Coded UI Test I will not cover step 3-5 here, there are plenty of information on how you create test plans and test cases and automate them using Coded UI Tests. In this example I have a test plan called My Application and it contains among other things a test suite called Automated Tests where I plan to put test cases that should be automated and executed as part of the BDT workflow. For more information about Coded UI Tests, see Verifying Code by Using Coded User Interface Tests   6. Associate Coded UI Test with Test Case OK, so now we want to automate our Coded UI Test and have it run as part of the BDT workflow. You might think that you coded UI test already is automated, but the meaning of the term here is that you link your coded UI Test to an existing Test Case, thereby making the Test Case automated. And the test case should be part of the test suite that we will run during the BDT. Open the solution that contains the coded UI test method. Open the Test Case work item that you want to automate. Go to the Associated Automation tab and click on the “…” button. Select the coded UI test that you corresponds to the test case: Press OK and the save the test case For more information about associating an automated test case with a test case, see How to: Associate an Automated Test with a Test Case 7. Create Build Definition using LabDefaultTemplate Now we are ready to create a build definition that will implement the full BDT workflow. For this purpose we will use the LabDefaultTemplate.11.xaml that comes out of the box in TFS 2012. This build process template lets you take the output of another build and deploy it to each target machine. Since the deployment process will be running on the target server, you will have less problem with permissions and firewalls than if you were to remote deploy your solution. So, before creating a BDT workflow build definition, make sure that you have an existing build definition that produces a release build of your application. Go to the Builds hub in Team Explorer and select New Build Definition Give the build definition a meaningful name, here I called it MyApplication.Deploy Set the trigger to Manual Define a workspace for the build definition. Note that a BDT build doesn’t really need a workspace, since all it does is to launch another build definition and deploy the output of that build. But TFS doesn’t allow you to save a build definition without adding at least one mapping. On Build Defaults, select the build controller. Since this build actually won’t produce any output, you can select the “This build does not copy output files to a drop folder” option. On the process tab, select the LabDefaultTemplate.11.xaml. This is usually located at $/TeamProject/BuildProcessTemplates/LabDefaultTemplate.11.xaml. To configure it, press the … button on the Lab Process Settings property First, select the environment that you created before: Select which build that you want to deploy and test. The “Select an existing build” option is very useful when developing the BDT workflow, because you do not have to run through the target build every time, instead it will basically just run through the deployment and test steps which speeds up the process. Here I have selected to queue a new build of the MyApplication.Test build definition On the deploy tab, you need to specify how the application should be installed on the target server. You can supply a list of deployment scripts with arguments that will be executed on the target server. In this example I execute the generated web deploy command file to deploy the solution. If you for example have databases you can use sqlpackage.exe to deploy the database. If you are producing MSI installers in your build, you can run them using msiexec.exe and so on. A good practice is to create a batch file that contain the entire deployment that you can run both locally and on the target server. Then you would just execute the deployment batch file here in one single step. The workflow defines some variables that are useful when running the deployments. These variables are: $(BuildLocation) The full path to where your build files are located $(InternalComputerName_<VM Name>) The computer name for a virtual machine in a SCVMM environment $(ComputerName_<VM Name>) The fully qualified domain name of the virtual machine As you can see, I specify the path to the myapplication.deploy.cmd file using the $(BuildLocation) variable, which is the drop folder of the MyApplication.Test build. Note: The test agent account must have read permission in this drop location. You can find more information here on Building your Deployment Scripts On the last tab, we specify which tests to run after deployment. Here I select the test plan and the Automated Tests test suite that we saw before: Note that I also selected the automated test settings (called TargetServer in this case) that I have defined for my test plan. In here I define what data that should be collected as part of the test run. For more information about test settings, see Specifying Test Settings for Microsoft Test Manager Tests We are done! Queue your BDT build and wait for it to finish. If the build succeeds, your build summary should look something like this:

    Read the article

  • Output = MAXDOP 1

    - by Dave Ballantyne
    It is widely know that data modifications on table variables do not support parallelism, Peter Larsson has a good example of that here .  Whilst tracking down a performance issue,  I saw that using the OUTPUT clause also causes parallelism to not be used. By way of example,  first lets create two tables with a simple parent and child (one to one) relationship, and then populate them with 100,000 rows. Drop table ParentDrop table Childgocreate table Parent(id integer identity Primary Key,data1 char(255))Create Table Child(id integer Primary Key)goinsert into Parent(data1)Select top 1000000 NULL from sys.columns a cross join sys.columns b insert into ChildSelect id from Parentgo If we then execute update Parent set data1 =''from Parentjoin Child on Parent.Id = Child.Id where Parent.Id %100 =1 and Child.id %100 =1 We should see an execution plan using parallelism such as   However,  if the OUTPUT clause is now used update Parent set data1 =''output inserted.idfrom Parentjoin Child on Parent.Id = Child.Id where Parent.Id %100 =1 and Child.id %100 =1   The execution plan shows that Parallelism was not used Make of that what you will, but i thought that this was a pretty unexpected outcome. Update : Laurence Hoff has mailed me to note that when the OUTPUT results are captured to a temporary table using the INTO clause,  then parallelism is used.  Naturally if you use a table variable then there is still no parallelism  

    Read the article

  • Silverlight Cream for November 25, 2011 -- #1174

    - by Dave Campbell
    In this Issue: Michael Collier, Samidip Basu, Jesse Liberty, Dhananjay Kumar, and Michael Crump. Above the Fold: WP7: "31 Days of Mango | Day #16: Isolated Storage Explorer" Samidip Basu Metro/WinRT/W8: "1360x768x32 Resolution in Windows 8 in VirtualBox" Michael Crump Shoutouts: Michael Palermo's latest Desert Mountain Developers is up Michael Washington's latest Visual Studio #LightSwitch Daily is up Alex Golesh releases a Silverlight 5-friendly version of his external map manifest file tool: Utility: Extmap Maker v1.1From SilverlightCream.com:31 Days of Mango | Day #17: Using Windows AzureMichael Collier has Jeff Blankenburg's Day 17 and is talking about Azure services for your Phone apps... great discussion on this... good diagrams, code, and entire project to download31 Days of Mango | Day #16: Isolated Storage ExplorerSamidip Basu has Jeff Blankenburg's 31 Days for Day 16, and is discussing ISO, and the Isolated Storage Explorer which helps peruse ISO either in the emulator or on your deviceTest Driven Development–Testing Private ValuesJesse Liberty's got a post up discussing TDD in his latest Full Stack excerpt wherein he and Jon Galloway are building a Pomodoro timer app. He has a solution for dealing with private member variables and is looking for feedbackVideo on How to work with System Tray Progress Indicator in Windows Phone 7Dhananjay Kumar's latest video tutorial is up... covering working with the System Tray Progress Indicator in WP7, as the title says :)1360x768x32 Resolution in Windows 8 in VirtualBoxMichael Crump is using a non-standard resolution with Win8 preview and demosntrates how to make that all work with VirtualBoxMichaelStay in the 'Light!Twitter SilverlightNews | Twitter WynApse | WynApse.com | Tagged Posts | SilverlightCreamJoin me @ SilverlightCream | Phoenix Silverlight User GroupTechnorati Tags:Silverlight    Silverlight 3    Silverlight 4    Windows PhoneMIX10

    Read the article

  • What is the most time-effective way to monitor & manage threats from bots and/or humans?

    - by CheeseConQueso
    I'm usually overwhelmed by the amount of tools that hosting companies provide to track & quantify traffic data and statistics. I'm equally overwhelmed by the countless flavors of malicious 'attacks' that target any and every web site known to man. The security methods used to protect both the back and front end of a website are documented well and are straight-forward in terms of ease of implementation and application, but the army of autonomous bots knows no boundaries and will always find a niche of a website to infest. So what can be done to handle the inevitable swarm of bots that pound your domain with brute force? Whenever I look at error logs for my domains, there are always thousands of entries that look like bots trying to sneak sql code into the database by tricking the variables in the url into giving them schema information or private data within the database. My barbaric and time-consuming plan of defense is just to monitor visitor statistics for those obvious patterns of abuse and either ban the ips or range of ips accordingly. Aside from that, I don't know much else I could do to prevent all of the ping pong going on all day. Are there any good tools that automatically monitor this background activity (specifically activity that throws errors on the web & db server) and proactively deal with these source(s) of mayhem?

    Read the article

  • Designing javascript chart library

    - by coolscitist
    I started coding a chart library on top of d3js: My chart library. I read Javascript API reusability and Towards reusable charts. However, I am NOT really following the suggestions because I am not really convinced about them. This is how my library can be used to create a bubble chart: var chart = new XYBubbleChart(); chart.data = [{"xValue":200,"yValue":300},{"xValue":400,"yValue":200},{"xValue":100,"yValue":310}]; //set data chart.dataKey.x = "xValue"; chart.dataKey.y = "yValue"; chart.elementId = "#chart"; chart.createChart(); Here are my questions: It does not use chaining. Is it a big issue? Every property and function is exposed publicly. (Example: width, height are exposed in Chart.js). OOP is all about abstraction and hiding, but I don't really see the point right now. I think exposing everything gives flexibility to change property and functionality inside subclasses and objects without writing a lot of code. What could be pitfalls of such exposure? I have implemented functions like: zooming, "showing info boxes when data point is clicked" as "abilities". (example: XYZoomingAbility.js). Basically, such "abilities" accept "chart" object, play around with public variables of "chart" to add functionality. What this allows me to do is to add an ability by writing: activateZoomAbility(chartObject); My goal is to separate "visualization" from "interactivity". I want "interactivity" like: zooming to be plugged into the chart rather than built inside the chart. Like, I don't want my bubble chart to know anything about "zooming". However, I do want zoomable bubble chart. What is the best way to do this? How to test and what to test? I have written mixed tests: jasmine and actual html files so that I can test manually on browser.

    Read the article

  • Scaling background without scaling foreground in platformer?

    - by David Xu
    I'm currently developing a platform game and I've run into a problem with scaling resolutions. I want a different resolution of the game to still display the foreground unscaled (characters, tiles, etc) but I want the background to be scaled to fit into the window. To explain this better, my viewport has 4 variables: (x, y, width, height) where x and y are the top left corner and width and height are the dimensions. These can be either 800x600, 1024x768 or 1280x960. When I design my levels, I design everything for the highest resolution (1280x960) and expect the game engine to scale it down if a user is running in a lower resolution. I have tried the following to make it work but nothing I've come up with solves it so far: scale = view->width/1280; drawX = x * scale; drawY = y * scale; (this makes the translation too small for low resolution) and scale = view->width/1280; bgWidth = background->width*scale; bgHeight = background->height*scale; drawX = x + background->width/2 - bgWidth/2; drawY = y + background->height/2 - bgHeight/2; (this makes the translation completely wrong at the edges of the map) The thing is, no matter what resolution the game is run at, the map remains the same size, and the foreground is unscaled. (With a lower resolution you just see less of the foreground in the viewport) I was wondering if anyone had any idea how to solve this problem? Thank you in advance!

    Read the article

  • MySQL Policy-Based Auditing Webinar Recording Now Availabile

    - by Rob Young
    For those who missed the live event, the recording of the "How to Add Policy-Based Auditing to your MySQL Applications" webinar is now available.  You can view it here. This presentation builds on my earlier blog post on MySQL Enterprise Audit that was announced at MySQL Connect in late September.  The web presentation expands on the introductory blog and covers: The regulatory problem to be solved (internal audit, PCI, Sarbanes-Oxley, HIPAA, others) MySQL Audit solutions for both Community and Enterprise users: General Log - use the basic features of the MySQL server MySQL 5.5 open audit API - or use your time and talent to build your own solution MySQL Enterprise Audit - or use the out of the box, ready for production solution from MySQL Simple, step-by-step process for installing, enabling and configuring the MySQL Enterprise Audit plugin for use with existing apps New variables and options for tuning the MySQL Enterprise Audit plugin for your specific use case Best practices for securing and managing audit log files and archived images Roadmap for adding an integrated solution around MySQL Enterprise Audit for MySQL only and Oracle/MySQL shops You can learn all the technical details on MySQL Enterprise Audit in the MySQL docs and learn all about MySQL Enterprise Edition and Auditing here. As always, thanks for your support of MySQL!

    Read the article

  • What do you call an obfuscator that isn't an obfuscator?

    - by Alex.Davies
    SmartAssembly, formerly {smartassembly}, version 5 is now available as an Early Access Build. You can get it here: http://www.red-gate.com/MessageBoard/viewforum.php?f=116 We're having second thoughts about the name change though. It isn't that we like the curly brackets, far from it. The trouble is that the first rule of product naming is to name a product by what it does. SmartAssembly may make an assembly smarter, but that's not something people really google for. The trouble is, I can't think of a better name for it. That's because SmartAssembly really does two completely separate things: Obfuscates Sets up your assembly for the awesome exception reports which get sent to you whenever your application crashes. You may have been (un?)lucky enough to see one in reflector if you use it. This is what those exception reports look like when they arrive back with the developer: Look at all those local variables! If you ask me, this is much cooler than the obfuscation. So obviously we don't want to call it just "Red Gate Obfuscator" or something, because it doesn't do justice to the exception reporting. What would you call it?

    Read the article

  • Using ASP.NET C# and Javascript

    - by ctck
    I'm looking for the most efficient / standardized way of passing data between client javascript code and C# code behind in an ASP.NET application. Currently ive been using the following methods to achieve this but they all feel a bit like a fudge. The way i pass data from javascript to the C# code behind is by setting hidden asp variables and triggering a postback <asp:HiddenField ID="RandomList" runat="server" /> function SetDataField(data) { document.getElementById('<%=RandomList.ClientID%>').value = data; } Then in C# code i collect the list protected void GetData(object sender, EventArgs e) { var _list = RandomList.value; } Going back the other way i often use either scriptmanager to register a function and pass it data during Page_Load: ScriptManager.RegisterStartupScript(this.GetType(), "Set","get("Test();",true); or i add attributes to controls before a post back or during Initialization / pre rendering stages: Btn.Attributes.Add("onclick", "DisplayMessage("Hello");"); These methods have served me well and do the job. However they just dont feel complete. Is there a more standardized way of passing data between client side markup / javascript and backend code. Ive seen some posts like this one: Injecting JavaScrip : StackOverflow that describe HtmlElement class. Is this something is should look into? Thanks everyone for your time.

    Read the article

  • Why do you hate Java? Is it the language or the framework? [closed]

    - by zneak
    According to you all, Java is the third most-hated language here. The two other most hated languages are PHP and VBScript. (It's quite funny how they stand together on the podium.) I'd like to make it known that the question mostly addresses people who don't like Java. I assume here a number of subjective opinions as facts because they're usually considered true among people who don't like Java, and I don't want to be convinced otherwise here. If you're a Java enthusiast, you might find this question frustrating. It's never been made clear if people hate Java itself, or if they hate it because of the framework, or if it's a mixture of the two. On a side you have the language, where you have: the "everything should be an object" philosophy, even in instances where it should obviously be something else (event handlers I'm pointing you); checked exceptions; the idea that all logic should be presented as methods and properties is a big no-no; the fact that "closures" created by anonymous types only include final variables and arguments, but will allow write access to any member of the parent class; a few more. On the other side, you have the JDK, with... its load of inconsistencies and overengineering; monolithic class hierarchies; meaningless base exceptions like IOException (though other frameworks have similar exception hierarchies); sluggish responsiveness even with Swing; a few more. My question is, do you think that, if either one (Java or the JDK) was taken alone, and the other was dropped in favor of something else, the new combination would be better? For instance, if you could use the C# syntax with the JDK (adapting get*/set* methods into properties, and interfaces with only one method into delegates), or the Java syntax with the .NET Framework (doing the inverse transformations), would things get better in your opinion?

    Read the article

  • Teach Your Kid to Code (&hellip;and Vote early!)

    - by Steve Michelotti
    Next Tuesday I will be at the CMAP main meeting presenting Teach Your Kid to Code. Next Tuesday is of course Election Day so you have to make sure you vote early in order to get over to CMAP for the 7:00PM presentation. I will be co-presenting this talk with my 5th grade son. Here is the abstract: Have you ever wanted a way to teach your kid to code? For that matter, have you ever wanted to simply be able to explain to your kid what you do for a living? Putting things in a context that a kid can understand is not as easy as it sounds. If you are someone curious about these concepts, this is a “can’t miss” presentation that will be co-presented by Justin Michelotti (5th grader) and his father. Bring your kid with you to CMAP for this fun and educational session. We will show tools you may not have been aware of like SmallBasic and Kodu – we’ll even throw in a little Visual Studio and Windows 8! Concepts such as variables, conditionals, loops, and functions will be covered while we introduce object oriented concepts without any of the confusing words. Kids are not required for entry! I promise this will be an entertaining presentation! We hope to see you (and your kids) there. Click here for details.

    Read the article

< Previous Page | 152 153 154 155 156 157 158 159 160 161 162 163  | Next Page >