Search Results

Search found 15233 results on 610 pages for 'ssis design patterns'.

Page 16/610 | < Previous Page | 12 13 14 15 16 17 18 19 20 21 22 23  | Next Page >

  • SSIS (missing) Pre-Build and Post-Build

    - by Raj More
    For the warehouse work under progress, we have a single solution with multiple projects in it OLTP Database Project Warehouse Database Project SSIS ETL project After the SSIS project is built, I want to move the binaries (XML, really) from the Bin folder to "C:\AutomatedTasks\ETL.Warehouse\" and "C:\AutomatedTasks\ETL" I cannot find the Post-Build events to do that for the SSIS project. Where are they? If they aren't available, how do I achieve this?

    Read the article

  • Should I pass the BrainBench Design patterns certification?

    - by Fedyashev Nikita
    I have found Design patterns certification at the Brainbehch. I have heard from people who passed it, that there are many Language-specific patterns questions, mostly from Java and C++. I think that this certification can: force me to improve my skills on Object oriented design and design patterns; improve and structure my knowledge of the domain; give real estimate of my knowledge, which is useful issue itself The only confusion I have about this certification, is that I have to learn C++/Java language specific design patterns, while I mostly do PHP development and don't want to switch to C++/Java. I'm familiar with Java & C++ syntax, read lots of books about different subjects with code snippets in this programming languages. I think, that if I pass well all concepts except language specific patterns at certification, it won't be very good, because this concepts will gain quite low results. What would you recommend in this particular circumstance?

    Read the article

  • Design suggestion for expression tree evaluation with time-series data

    - by Lirik
    I have a (C#) genetic program that uses financial time-series data and it's currently working but I want to re-design the architecture to be more robust. My main goals are: sequentially present the time-series data to the expression trees. allow expression trees to access previous data rows when needed. to optimize performance of the data access while evaluating the expression trees. keep a common interface so various types of data can be used. Here are the possible approaches I've thought about: I can evaluate the expression tree by passing in a data row into the root node and let each child node use the same data row. I can evaluate the expression tree by passing in the data row index and letting each node get the data row from a shared DataSet (currently I'm passing the row index and going to multiple synchronized arrays to get the data). Hybrid: an immutable data set is accessible by all of the expression trees and each expression tree is evaluated by passing in a data row. The benefit of the first approach is that the data row is being passed into the expression tree and there is no further query done on the data set (which should increase performance in a multithreaded environment). The drawback is that the expression tree does not have access to the rest of the data (in case some of the functions need to do calculations using previous data rows). The benefit of the second approach is that the expression trees can access any data up to the latest data row, but unless I specify what that row is, I'll have to iterate through the rows and figure out which one is the last one. The benefit of the hybrid is that it should generally perform better and still provide access to the earlier data. It supports two basic "views" of data: the latest row and the previous rows. Do you guys know of any design patterns or do you have any tips that can help me build this type of system? Should I use a DataSet to hold and present the data, or are there more efficient ways to present rows of data while maintaining a simple interface? FYI: All of my code is written in C#.

    Read the article

  • 'Forward-Compatible' Program Design

    - by Jeffrey Kern
    The majority of my questions I've asked here so far on StackOverflow have been how to implement individual concepts and techniques towards developing a software-based NES clone via the XNA environment. The small samples that I've thrown together on my PC work relatively great and everything. Except I hit a brick wall. How do I merge all of these samples together. Having proof-of-concept is amazing, except when you need it to go beyond just that. I now have samples strewn about that I'm trying to merge, some of them incomplete. And now I'm stuck with the chicken-and-the-egg situation of where I would like to incorporate these samples together, to make sure they work, but I cannot without test data. And I don't have tools to create test data, because they'd need to be based off of the individual pieces that need to be put together. In my mind, I'm having nightmares with circular reference. For my sample data, I am hoping to save it in XML and write a specification - and then make sample data by hand - but I'm too paranoid of manually creating an XML file full of incorrect data and blaming it on my code, or vice-versa. It doesn't help that the end-result of my work is graphic-oriented, which makes it interseting how a graphic on the screen can be visualized in XML Nodes. I guess, my question is this: What design patterns and disciplines exist in the coding world that address this type of concern? I've always relied on brute-force coding and restarting a project with a whole new code base in attempts to further along my goals, but I doubt that would be the best way to do so. Within my college career, the majority of my programming was to work on simple projects that came out of a book, or with a given correct data set and a verifyable result. I don't have that, as my own design documents that I am going by could be terribly wrong.

    Read the article

  • Design for a Debate club assignment application

    - by Amir Rachum
    Hi all, For my university's debate club, I was asked to create an application to assign debate sessions and I'm having some difficulties as to come up with a good design for it. I will do it in Java. Here's what's needed: What you need to know about BP debates: There are four teams of 2 debaters each and a judge. The four groups are assigned a specific position: gov1, gov2, op1, op2. There is no significance to the order within a team. The goal of the application is to get as input the debaters who are present (for example, if there are 20 people, we will hold 2 debates) and assign them to teams and roles with regards to the history of each debater so that: Each debater should debate with (be on the same team) as many people as possible. Each debater should uniformly debate in different positions. The debate should be fair - debaters have different levels of experience and this should be as even as possible - i.e., there shouldn't be a team of two very experienced debaters and a team of junior debaters. There should be an option for the user to restrict the assignment in various ways, such as: Specifying that two people should debate together, in a specific position or not. Specifying that a single debater should be in a specific position, regardless of the partner. etc... If anyone can try to give me some pointers for a design for this application, I'll be so thankful! Also, I've never implemented a GUI before, so I'd appreciate some pointers on that as well, but it's not the major issue right now.

    Read the article

  • Which design pattern is most appropriate?

    - by Anon
    Hello, I want to create a class that can use one of four algorithms (and the algorithm to use is only known at run-time). I was thinking that the Strategy design pattern sounds appropriate, but my problem is that each algorithm requires slightly different parameters. Would it be a bad design to use strategy, but pass in the relevant parameters into the constructor?. Here is an example (for simplicity, let's say there are only two possible algorithms) ... class Foo { private: // At run-time the correct algorithm is used, e.g. a = new Algorithm1(1); AlgorithmInterface* a; }; class AlgorithmInterface { public: virtual void DoSomething = 0; }; class Algorithm1 : public AlgorithmInterface { public: Algorithm1( int i ) : value(i) {} virtual void DoSomething(){ // Does something with int value }; int value; }; class Algorithm2 : public AlgorithmInterface { public: Algorithm2( bool b ) : value(b) {} virtual void DoSomething(){ // Do something with bool value }; bool value; };

    Read the article

  • Which is the better C# class design for dealing with read+write versus readonly

    - by DanM
    I'm contemplating two different class designs for handling a situation where some repositories are read-only while others are read-write. (I don't foresee any need to a write-only repository.) Class Design 1 -- provide all functionality in a base class, then expose applicable functionality publicly in sub classes public abstract class RepositoryBase { protected virtual void SelectBase() { // implementation... } protected virtual void InsertBase() { // implementation... } protected virtual void UpdateBase() { // implementation... } protected virtual void DeleteBase() { // implementation... } } public class ReadOnlyRepository : RepositoryBase { public void Select() { SelectBase(); } } public class ReadWriteRepository : RepositoryBase { public void Select() { SelectBase(); } public void Insert() { InsertBase(); } public void Update() { UpdateBase(); } public void Delete() { DeleteBase(); } } Class Design 2 - read-write class inherits from read-only class public class ReadOnlyRepository { public void Select() { // implementation... } } public class ReadWriteRepository : ReadOnlyRepository { public void Insert() { // implementation... } public void Update() { // implementation... } public void Delete() { // implementation... } } Is one of these designs clearly stronger than the other? If so, which one and why? P.S. If this sounds like a homework question, it's not, but feel free to use it as one if you want :)

    Read the article

  • Unsure how to come up with a good design

    - by Mewzer
    Hello there, I am having trouble coming up with a good design for a group of classes and was hoping that someone could give me some guidance on best practices. I have kept the classes and member functions generic to make the problem simpler. Essentially, I have three classes (lets call them A, B, and C) as follows: class A { ... int GetX( void ) const { return x; }; int GetY( void ) const { return y; }; private: B b; // NOTE: A "has-a" B int x; int y; }; class B { ... void SetZ( int value ) { z = value }; private: int z; C c; // NOTE: B "has-a" C }; class C { private: ... void DoSomething(int x, int y){ ... }; void DoSomethingElse( int z ){ ... }; }; My problem is as follows: Class A uses its member variables "x" and "y" a lot internally. Class B uses its member variable "z" a lot internally. Class B needs to call C::DoSomething(), but C::DoSomething() needs the values of X and Y in class A passed in as arguments. C::DoSomethingElse() is called from say another class (e.g. D), but it needs to invoke SetZ() in class B!. As you can see, it is a bit of a mess as all the classes need information from one another!. Are there any design patterns I can use?. Any ideas would be much appreciated ....

    Read the article

  • Design pattern for parsing data that will be grouped to two different ways and flipped

    - by lewisblackfan
    I'm looking for an easily maintainable and extendable design model for a script to parse an excel workbook into two separate workbooks after pulling data from other locations like the command line, and a database. The high level details are as follows. I need to parse an excel workbook containing a sheet that lists unique question names, the only reliable information that can be parsed from the question name is the book code that identifies the title and edition of the textbook the question is associated with, the rest of the question name is not standardized well enough to be reliably parsed by computer. The general form of the question name is best described by the following regular expression. '^(\w+)\s(\w{1,2})\.(\w{1,2})\.(\w{1,3})\.(\w{1,3}\.)*$' The first sub-pattern is the book code, the second sub-pattern is 90% of the time the chapter, and the rest of the sub-patterns could be section, problem type, problem number, or question type information. There is no simple logic, at least not one I can find. There will be a minimum of three other columns in this spreadsheet; one column will be the chapter the question is associated with, the second will be the section within the chapter the question is associated with, and the third will be some kind of asset indicated by a uniform resource locator. 1 | 1 | qname1 | url | description | url | description ... 1 | 1 | qname2 | url | description 1 | 1 | qname3 | url | description | url | description | url | The asset can be indicated by a full or partial uniform resource locator, the partial url will need to be completed before it can be fed into the application. There theoretically could be no limit to the number of asset columns, the assets will be grouped in columns by type. Some times additional data will have to be retrieved from a database or combined with the book code before the asset url is complete and can be understood by the application that will be using the asset. The type is an abstraction, there are eight types right now, each with their own logic in how the uniform resource locator is handled and or completed, and I have to add a new type and its logic every three or four months. For each asset url there is the possibility of a description column, a character string for display in the application, but not always. (I've already worked out validating the description text, and squashing MSs obscure code page down to something 7-bit ascii can handle.) Now that all the details are filled-in I can get to the actual problem of parsing the file. I need to split the information in this excel workbook into two separate workbooks. The first workbook will group all the questions by section in rows. With the first cell being the section doublet and the rest of the cells in the row are the question names. 1.1 | qname1 | qname2 | qname3 | qname4 | 1.2 | qname1 | qname2 | qname3 | 1.3 | qname1 | qname2 | qname3 | qname4 | qname5 There is no set number of questions for each section as you can see from the above example. The second workbook is more complicated, there is one row per asset, and question names that have more than one asset will be duplicated. There will be four or five columns on this sheet. The first is the question name for the asset, the second is a media type used to select the correct icon for the asset in the application, the third is string representing the asset type, the four is the full and complete uniform resource locator for the asset, and the fifth columns is the optional text description for the asset. q1 | mtype1 | atype1 | url | description q1 | mtype2 | atype2 | url | description q1 | mtype2 | atype3 | url | description q2 | mtype1 | atype1 | url | description q2 | mtype2 | atype3 | url | description For the original six types I did have a script that parsed the source excel workbook into the other two excel workbooks, and I was able to add two more types until I ran aground on the implementation of the ninth type and tenth types. What broke my script was the fact that the ninth type is actually a sub-type of one of the original six, but with entirely different logic, and my mostly procedural script could not accommodate without duplicating a lot of code. I also had a lot of bugs in the script and will be writing the test first on this time around. I'm stuck with the format for the resulting two workbooks, this script is glue code, development went ahead with the project without bothering to get a complete spec from the sponsor. I work for the same company as the developers but in the editorial department, editorial is co-sponsor of the project, and am expected to fix pesky details like this (I'm foaming at the mouth as I type this). I've tried factories, I've tried different object models, but each resulting workbook is so different when I find a design that works for generating one workbook the code is not really usable for generating the other. What I would really like are ideas about a maintainable and extensible design for parsing the source workbook into both workbooks with maximum code reuse, and or sympathy.

    Read the article

  • Which book should I choose?

    - by sebastianlarsson
    Hi guys, I'm looking for a good read on object oriented design. The two books I'm currently looking Head First Design Patterns and Head First Object object-oriented analysis & design. They seem very similar when looking at the contents and browsing through available sample text. Which one would be the best choice? About myself: I have a bachelor in computer science and I am currently studying Msc. Software Quality Engineering (read Software Engineering with focus on Quality). I am already confident in object-oriented design and have a lot of programming courses in my backpack. I have done games in c++, courses in advanced java programming (I am SCJP certified), but my preferred language is C#. I have also worked with Java for the last 7 months while studying. I am currently also studying for certificates in C# (apart from my usual studies). So I believe I have the prerequisites of actually understanding the contents of both books. Reason: I just want to be better and keep evolving as a programmer. I think it is fun. I believe Bert Bates and Kathy Sierra are involved in both these books and I have previously read their SCJP preparation book in java. I really do enjoy their style of writing. Other books which I am considering are: Clean Code: A Handbook Of Agile Software Craftsmanship Thx in advance Sebastian

    Read the article

  • Improvements to Joshua Bloch's Builder Design Pattern?

    - by Jason Fotinatos
    Back in 2007, I read an article about Joshua Blochs take on the "builder pattern" and how it could be modified to improve the overuse of constructors and setters, especially when an object has a large number of properties, most of which are optional. A brief summary of this design pattern is articled here [http://rwhansen.blogspot.com/2007/07/theres-builder-pattern-that-joshua.html]. I liked the idea, and have been using it since. The problem with it, while it is very clean and nice to use from the client perspective, implementing it can be a pain in the bum! There are so many different places in the object where a single property is reference, and thus creating the object, and adding a new property takes a lot of time. So...I had an idea. First, an example object in Joshua Bloch's style: Josh Bloch Style: public class OptionsJoshBlochStyle { private final String option1; private final int option2; // ...other options here <<<< public String getOption1() { return option1; } public int getOption2() { return option2; } public static class Builder { private String option1; private int option2; // other options here <<<<< public Builder option1(String option1) { this.option1 = option1; return this; } public Builder option2(int option2) { this.option2 = option2; return this; } public OptionsJoshBlochStyle build() { return new OptionsJoshBlochStyle(this); } } private OptionsJoshBlochStyle(Builder builder) { this.option1 = builder.option1; this.option2 = builder.option2; // other options here <<<<<< } public static void main(String[] args) { OptionsJoshBlochStyle optionsVariation1 = new OptionsJoshBlochStyle.Builder().option1("firefox").option2(1).build(); OptionsJoshBlochStyle optionsVariation2 = new OptionsJoshBlochStyle.Builder().option1("chrome").option2(2).build(); } } Now my "improved" version: public class Options { // note that these are not final private String option1; private int option2; // ...other options here public String getOption1() { return option1; } public int getOption2() { return option2; } public static class Builder { private final Options options = new Options(); public Builder option1(String option1) { this.options.option1 = option1; return this; } public Builder option2(int option2) { this.options.option2 = option2; return this; } public Options build() { return options; } } private Options() { } public static void main(String[] args) { Options optionsVariation1 = new Options.Builder().option1("firefox").option2(1).build(); Options optionsVariation2 = new Options.Builder().option1("chrome").option2(2).build(); } } As you can see in my "improved version", there are 2 less places in which we need to add code about any addition properties (or options, in this case)! The only negative that I can see is that the instance variables of the outer class are not able to be final. But, the class is still immutable without this. Is there really any downside to this improvement in maintainability? There has to be a reason which he repeated the properties within the nested class that I'm not seeing?

    Read the article

  • Design pattern for logging changes in parent/child objects saved to database

    - by andrew
    I’ve got a 2 database tables in parent/child relationship as one-many. I’ve got three classes representing the data in these two tables: Parent Class { Public int ID {get; set;} .. other properties } Child Class { Public int ID {get;set;} Public int ParentID {get; set;} .. other properties } TogetherClass { Public Parent Parent; Public List<Child> ChildList; } Lastly I’ve got a client and server application – I’m in control of both ends so can make changes to both programs as I need to. Client makes a request for ParentID and receives a Together Class for the matching parent, and all of the child records. The client app may make changes to the children – add new children, remove or modify existing ones. Client app then sends the Together Class back to the server app. Server app needs to update the parent and child records in the database. In addition I would like to be able to log the changes – I’m doing this by having 2 separate tables one for Parent, one for child; each containing the same columns as the original plus date time modified, by whom and a list of the changes. I’m unsure as to the best approach to detect the changes in records – new records, records to be deleted, records with no fields changed, records with some fields changed. I figure I need to read the parent & children records and compare those to the ones in the Together Class. Strategy A: If Together class’s child record has an ID of say 0, that indicates a new record; insert. Any deleted child records are no longer in the Together Class; see if any of the comparison child records are not found in the Together class and delete if not found (Compare using ID). Check each child record for changes and if changed log. Strategy B: Make a new Updated TogetherClass UpdatedClass { Public Parent Parent {get; set} Public List<Child> ListNewChild {get;set;} Public List<Child> DeletedChild {get;set;} Public List<Child> ExistingChild {get;set;} // used for no changes and modified rows } And then process as per the list. The reason why I’m asking for ideas is that both of these solutions don’t seem optimal to me and I suspect this problem has been solved already – some kind of design pattern ? I am aware of one potential problem in this general approach – that where Client App A requests a record; App B requests same record; A then saves changes; B then saves changes which may overwrite changes A made. This is a separate locking issue which I’ll raise a separate question for if I’ve got trouble implementing. The actual implementation is c#, SQL Server and WCF between client and server - sharing a library containing the class implementations. Apologies if this is a duplicate post – I tried searching various terms without finding a match though.

    Read the article

  • Speaking at SQL Saturday #39 in NYC!

    - by andyleonard
    I am honored to present Applied SSIS Design Patterns and Introduction to Incremental Loads at SQL Saturday #39 in New York City! If you're there and you read this blog, be sure to stop by and introduce yourself! :{> Andy Share this post: email it! | bookmark it! | digg it! | reddit! | kick it! | live it!...(read more)

    Read the article

  • Why can't we capture the design of software more effectively?

    - by Ira Baxter
    As engineers, we all "design" artifacts (buildings, programs, circuits, molecules...). That's an activity (design-the-verb) that produces some kind of result (design-the-noun). I think we all agree that design-the-noun is a different entity than the artifact itself. A key activity in the software business (indeed, in any business where the resulting product artifact needs to be enhanced) is to understand the "design (the-noun)". Yet we seem, as a community, to be pretty much complete failures at recording it, as evidenced by the amount of effort people put into rediscovering facts about their code base. Ask somebody to show you the design of their code and see what you get. I think of a design for software as having: An explicit specification for what the software is supposed to do and how well it does it An explicit version of the code (this part is easy, everybody has it) An explanation for how each part of the code serves to achieve the specification A rationale as to why the code is the way it is (e.g., why a particualr choice rather than another) What is NOT a design is a particular perspective on the code. For example [not to pick specifically on] UML diagrams are not designs. Rather, they are properties you can derive from the code, or arguably, properties you wish you could derive from the code. But as a general rule, you can't derive the code from UML. Why is it that after 50+ years of building software, why don't we have regular ways to express this? My personal opinion is that we don't have good ways to express this. Even if we do, most of the community seems so focused on getting "code" that design-the-noun gets lost anyway. (IMHO, until design becomes the purpose of engineering, with the artifact extracted from the design, we're not going to get around this). What have you seen as means for recording designs (in the sense I have described it)? Explicit references to papers would be good. Why do you think specific and general means have not been succesful? How can we change this?

    Read the article

  • How to implement Restricted access to application features

    - by DroidUser
    I'm currently developing a web application, that provides some 'service' to the user. The user will have to select a 'plan' according to which she/he will be allowed to perform application specific actions but up to a limit defined by the plan. A Plan will also limit access to certain features, which will not be available at all for some plans. As an example : say there are 3 plans, 2 actions throughout the application users in plan-1 can perform action-1 3 times, and they can't perform action-2 at all users in plan-2 can perform action-1 10 times, action-2 5 times users in plan-3 can perform action-1 20 times, action-2 10 times So i'm looking for the best way to get this done, and my main concerns besides implementing it, are the following(in no particular order) maintainability/changeability : the number of plans, and type of features/actions will change in the final product industry standard/best practice : for future readiness!! efficiency : ofcourse, i want fast code!! I have never done anything like this before, so i have no clue about how do i go about implementing these functionalities. Any tips/guides/patterns/resources/examples? I did read a little about ACL, RBAC, are they the patterns that i need to follow? Really any sort of feedback will help.

    Read the article

  • (Database Design - products attributes): What is better option for product attribute database design

    - by meyosef
    Hi, I new in database design. What is better option for product attribute database design for cms?(Please suggest other options also). option 1: 1 table products{ id product_name color price attribute_name1 attribute_value1 attribute_name2 attribute_value2 attribute_name3 attribute_value3 } option 2: 3 tables products{ id product_name color price } attribute{ id name value } products_attribute{ products_id attribute_id } Thanks, Yosef

    Read the article

  • Help me to find a better approach-Design Pattern

    - by DJay
    I am working on an ASP.Net web application in which several WCF services are being used. At client level, I am creating channel factory mechanism to invoke service operations. Right now, I have created an assembly having classes used for channel factory creation code for every service. As per my assumption this is some sort of facade pattern. Please help me to find a better approach or any design pattern, which I can use here.

    Read the article

  • SQL SERVER – SSIS Parameters in Parent-Child ETL Architectures – Notes from the Field #040

    - by Pinal Dave
    [Notes from Pinal]: SSIS is very well explored subject, however, there are so many interesting elements when we read, we learn something new. A similar concept has been Parent-Child ETL architecture’s relationship in SSIS. Linchpin People are database coaches and wellness experts for a data driven world. In this 40th episode of the Notes from the Fields series database expert Tim Mitchell (partner at Linchpin People) shares very interesting conversation related to how to understand SSIS Parameters in Parent-Child ETL Architectures. In this brief Notes from the Field post, I will review the use of SSIS parameters in parent-child ETL architectures. A very common design pattern used in SQL Server Integration Services is one I call the parent-child pattern.  Simply put, this is a pattern in which packages are executed by other packages.  An ETL infrastructure built using small, single-purpose packages is very often easier to develop, debug, and troubleshoot than large, monolithic packages.  For a more in-depth look at parent-child architectures, check out my earlier blog post on this topic. When using the parent-child design pattern, you will frequently need to pass values from the calling (parent) package to the called (child) package.  In older versions of SSIS, this process was possible but not necessarily simple.  When using SSIS 2005 or 2008, or even when using SSIS 2012 or 2014 in package deployment mode, you would have to create package configurations to pass values from parent to child packages.  Package configurations, while effective, were not the easiest tool to work with.  Fortunately, starting with SSIS in SQL Server 2012, you can now use package parameters for this purpose. In the example I will use for this demonstration, I’ll create two packages: one intended for use as a child package, and the other configured to execute said child package.  In the parent package I’m going to build a for each loop container in SSIS, and use package parameters to pass in a value – specifically, a ClientID – for each iteration of the loop.  The child package will be executed from within the for each loop, and will create one output file for each client, with the source query and filename dependent on the ClientID received from the parent package. Configuring the Child and Parent Packages When you create a new package, you’ll see the Parameters tab at the package level.  Clicking over to that tab allows you to add, edit, or delete package parameters. As shown above, the sample package has two parameters.  Note that I’ve set the name, data type, and default value for each of these.  Also note the column entitled Required: this allows me to specify whether the parameter value is optional (the default behavior) or required for package execution.  In this example, I have one parameter that is required, and the other is not. Let’s shift over to the parent package briefly, and demonstrate how to supply values to these parameters in the child package.  Using the execute package task, you can easily map variable values in the parent package to parameters in the child package. The execute package task in the parent package, shown above, has the variable vThisClient from the parent package mapped to the pClientID parameter shown earlier in the child package.  Note that there is no value mapped to the child package parameter named pOutputFolder.  Since this parameter has the Required property set to False, we don’t have to specify a value for it, which will cause that parameter to use the default value we supplied when designing the child pacakge. The last step in the parent package is to create the for each loop container I mentioned earlier, and place the execute package task inside it.  I’m using an object variable to store the distinct client ID values, and I use that as the iterator for the loop (I describe how to do this more in depth here).  For each iteration of the loop, a different client ID value will be passed into the child package parameter. The final step is to configure the child package to actually do something meaningful with the parameter values passed into it.  In this case, I’ve modified the OleDB source query to use the pClientID value in the WHERE clause of the query to restrict results for each iteration to a single client’s data.  Additionally, I’ll use both the pClientID and pOutputFolder parameters to dynamically build the output filename. As shown, the pClientID is used in the WHERE clause, so we only get the current client’s invoices for each iteration of the loop. For the flat file connection, I’m setting the Connection String property using an expression that engages both of the parameters for this package, as shown above. Parting Thoughts There are many uses for package parameters beyond a simple parent-child design pattern.  For example, you can create standalone packages (those not intended to be used as a child package) and still use parameters.  Parameter values may be supplied to a package directly at runtime by a SQL Server Agent job, through the command line (via dtexec.exe), or through T-SQL. Also, you can also have project parameters as well as package parameters.  Project parameters work in much the same way as package parameters, but the parameters apply to all packages in a project, not just a single package. Conclusion Of the numerous advantages of using catalog deployment model in SSIS 2012 and beyond, package parameters are near the top of the list.  Parameters allow you to easily share values from parent to child packages, enabling more dynamic behavior and better code encapsulation. If you want me to take a look at your server and its settings, or if your server is facing any issue we can Fix Your SQL Server. Reference: Pinal Dave (http://blog.sqlauthority.com)Filed under: Notes from the Field, PostADay, SQL, SQL Authority, SQL Query, SQL Server, SQL Tips and Tricks, T SQL

    Read the article

  • How to program for constraints/rules

    - by Gaurav
    First the background, during interviews in the past, many times I have been asked to design some or other variation of card game as programming puzzle, and I have tried to design it in OO way, but I have never been satisfied with my solutions. However it was not until recently that I realized that I had been approaching the problem from the wrong direction. Specifically I was trying to solve the problem by modeling individual card as an object. Problem with this is individual cards don't have any non-trivial intrinsic behavior and therefore are not suitable (or primary) candidate as objects. What is interesting and important about cards are rules and constraints, such as there could be only four suits, or only thirteen cards in each suit. Of course, then there are any number of rules for games. So my questions are Are there any idioms/constructs/patterns to program for rules & constraints. How many in 1 can be applied in conjunction with OO paradigm.

    Read the article

  • What are the software design essentials? [closed]

    - by Craig Schwarze
    I've decided to create a 1 page "cheat sheet" of essential software design principles for my programmers. It doesn't explain the principles in any great depth, but is simply there as a reference and a reminder. Here's what I've come up with - I would welcome your comments. What have I left out? What have I explained poorly? What is there that shouldn't be? Basic Design Principles The Principle of Least Surprise – your solution should be obvious, predictable and consistent. Keep It Simple Stupid (KISS) - the simplest solution is usually the best one. You Ain’t Gonna Need It (YAGNI) - create a solution for the current problem rather than what might happen in the future. Don’t Repeat Yourself (DRY) - rigorously remove duplication from your design and code. Advanced Design Principles Program to an interface, not an implementation – Don’t declare variables to be of a particular concrete class. Rather, declare them to an interface, and instantiate them using a creational pattern. Favour composition over inheritance – Don’t overuse inheritance. In most cases, rich behaviour is best added by instantiating objects, rather than inheriting from classes. Strive for loosely coupled designs – Minimise the interdependencies between objects. They should be able to interact with minimal knowledge of each other via small, tightly defined interfaces. Principle of Least Knowledge – Also called the “Law of Demeter”, and is colloquially summarised as “Only talk to your friends”. Specifically, a method in an object should only invoke methods on the object itself, objects passed as a parameter to the method, any object the method creates, any components of the object. SOLID Design Principles Single Responsibility Principle – Each class should have one well defined purpose, and only one reason to change. This reduces the fragility of your code, and makes it much more maintainable. Open/Close Principle – A class should be open to extension, but closed to modification. In practice, this means extracting the code that is most likely to change to another class, and then injecting it as required via an appropriate pattern. Liskov Substitution Principle – Subtypes must be substitutable for their base types. Essentially, get your inheritance right. In the classic example, type square should not inherit from type rectangle, as they have different properties (you can independently set the sides of a rectangle). Instead, both should inherit from type shape. Interface Segregation Principle – Clients should not be forced to depend upon methods they do not use. Don’t have fat interfaces, rather split them up into smaller, behaviour centric interfaces. Dependency Inversion Principle – There are two parts to this principle: High-level modules should not depend on low-level modules. Both should depend on abstractions. Abstractions should not depend on details. Details should depend on abstractions. In modern development, this is often handled by an IoC (Inversion of Control) container.

    Read the article

  • Algorithm to Solve Most of a Problem

    - by Mike G
    I need an Algorithm/Design Pattern that allows me to try to get the maximum number of rules followed. So I have a couple teams and I need to pair them with a referee and against each other into a round robin. There a rules on who can compete with who and who can judge who so I need to find the configuration that satisfies the most of these. Some rules are more important than others and are "worth more" when evaluating "what satisfies the most of them" There probably isn't a algorithm for this, but is there a design pattern that could help me maximize my chances of finding this configuration?

    Read the article

  • Exposing warnings\errors from data objects (that are also list returned)

    - by Oren Schwartz
    I'm exposing Data objects via service oriented assembly (which on future usages might become a WCF service). The data object is tree designed, as well as formed from allot of properties.Moreover, some services return one objects, others retrieve a list of them (thus disables throwing exceptions). I now want to expose data flow warnings and wondering what's the best way to do it having to things to consider: (1) seperation (2) ease of access. On the one hand, i want the UI team to be able to access a fields warnings (or errors) without having them mapping the field names to an external source but on the other hand, i don't want the warnings "hanged" on the object itself (as i don't see it a correct design). I tought of creating a new type of wrapper for each field, that'll expose events and they'll have to register the one's they care about (but totally not sure) I'll be happy to hear your thoughts. Could you please direct me to a respectful design pattern ? what dp will do best here ? Thank you very much!

    Read the article

  • How to avoid code duplication for a system which has logic that may change year wise?

    - by aravind
    What would be the way to design a system which has logic that may change year wise? There is an application which conducts online exams. There are five questions for a particular subject. The questions may (or may not) change year wise. As per my current design, the questions in database are stored year wise. There are some year specific code logic as well. In order to enable the application for another year, the year specific database records and code will be copied or duplicated. How to avoid this code duplication?

    Read the article

  • Using packages (gems, eggs, etc.) to create decoupled architectures

    - by Juan Carlos Coto
    The main issue Seeing the good support most modern programming platforms have for package management (think gem, npm, pip, etc), does it make sense to design an application or system be composed of internally developed packages, so as to promote and create a loosely coupled architecture? Example An example of this would be to create packages for database access, as well as for authentication and other components of the system. These, of course, use external packages as well. Then, your system imports and uses these packages - instead of including their code within its own code base. Considerations To me, it seems that this would promote code decoupling and help maintainability, almost in a Web-based-vs.-desktop-application kind of way (updates are applied almost automatically, single code base for single functionality, etc.). Does this seem like a rational and sane design concept? Is this actually used as a standard way of structuring applications today? Thanks very much!

    Read the article

< Previous Page | 12 13 14 15 16 17 18 19 20 21 22 23  | Next Page >