Search Results

Search found 40870 results on 1635 pages for 'database design'.

Page 334/1635 | < Previous Page | 330 331 332 333 334 335 336 337 338 339 340 341  | Next Page >

  • Iterator performance contract (and use on non-collections)

    - by polygenelubricants
    If all that you're doing is a simple one-pass iteration (i.e. only hasNext() and next(), no remove()), are you guaranteed linear time performance and/or amortized constant cost per operation? Is this specified in the Iterator contract anywhere? Are there data structures/Java Collection which cannot be iterated in linear time? java.util.Scanner implements Iterator<String>. A Scanner is hardly a data structure (e.g. remove() makes absolutely no sense). Is this considered a design blunder? Is something like PrimeGenerator implements Iterator<Integer> considered bad design, or is this exactly what Iterator is for? (hasNext() always returns true, next() computes the next number on demand, remove() makes no sense). Similarly, would it have made sense for java.util.Random implements Iterator<Double>? Should a type really implement Iterator if it's effectively only using one-third of its API? (i.e. no remove(), always hasNext())

    Read the article

  • Maintain List of Active Users for Web

    - by Bryan Marble
    Problem Statement - Would like to know if particular web app user is active (i.e. logged in and using site) and be able to query for list of active users or determine a user's activity status. Constraints - Doesn't need to be exact (i.e. if a user was active within a certain timeframe, that's ok to say that they're active even if they've closed their browser). I feel like there should be a design pattern for this type of problem but haven't been able to find anything here or elsewhere on the web. Approaches I'm considering: Maintain a table that is updated any time a user performs an action (or some subset of actions). Would then query for users that have performed an action within some threshold of time. Try to monitor session information and maintain a table that lists logged in users and times out after a certain period of time. Some other more standard way of doing this? How would you approach this problem (again, from a design pattern perspective)? Thanks!

    Read the article

  • DDD and MVC: Difference between 'Model' and 'Entity'

    - by Nathan Loding
    I'm seriously confused about the concept of the 'Model' in MVC. Most frameworks that exist today put the Model between the Controller and the database, and the Model almost acts like a database abstraction layer. The concept of 'Fat Model Skinny Controller' is lost as the Controller starts doing more and more logic. In DDD, there is also the concept of a Domain Entity, which has a unique identity to it. As I understand it, a user is a good example of an Entity (unique userid, for instance). The Entity has a life-cycle -- it's values can change throughout the course of the action -- and then it's saved or discarded. The Entity I describe above is what I thought Model was supposed to be in MVC? How off-base am I? To clutter things more, you throw in other patterns, such as the Repository pattern (maybe putting a Service in there). It's pretty clear how the Repository would interact with an Entity -- how does it with a Model? Controllers can have multiple Models, which makes it seem like a Model is less a "database table" than it is a unique Entity. So, in very rough terms, which is better? No "Model" really ... class MyController { public function index() { $repo = new PostRepository(); $posts = $repo->findAllByDateRange('within 30 days'); foreach($posts as $post) { echo $post->Author; } } } Or this, which has a Model as the DAO? class MyController { public function index() { $model = new PostModel(); // maybe this returns a PostRepository? $posts = $model->findAllByDateRange('within 30 days'); while($posts->getNext()) { echo $posts->Post->Author; } } } Both those examples didn't even do what I was describing above. I'm clearly lost. Any input?

    Read the article

  • Which Code Should Go Where in MVC Structure

    - by Oguz
    My problem is in somewhere between model and controller.Everything works perfect for me when I use MVC just for crud (create, read, update, delete).I have separate models for each database table .I access these models from controller , to crud them . For example , in contacts application,I have actions (create, read, update, delete) in controller(contact) to use model's (contact) methods (create, read, update, delete). The problem starts when I try to do something more complicated. There are some complex processes which I do not know where should I put them. For example , in registering user process. I can not just finish this process in user model because , I have to use other models too (sending mails , creating other records for user via other models) and do lots of complex validations via other models. For example , in some complex searching processes , I have to access lots of models (articles, videos, images etc.) Or, sometimes , I have to use apis to decide what I will do next or which database model I will use to record data So where is the place to do this complicated processes. I do not want to do them in controllers , Because sometimes I should use these processes in other controllers too. And I do not want to put these process in models because , I use models as database access layers .May be I am wrong,I want to know . Thank you for your answer .

    Read the article

  • desing pattern for related inputs

    - by curiousMo
    My question is a design question : let's say i have a data entry web page with 4 drop down lists, each depending on the previous one, and a bunch of text boxes. country (ddl), state (ddl), city (ddl), boro (ddl), address (txtBox), zipcode(txtbox). and an object that represents a datarow with a value for each. naturally the country, state, city and boro values will be values of primary keys of some lookup tables. when the user chooses to edits that record, i would load it from database and load it into the page. the issue that I have is how to streamline loading the ddls. i have some code that would grab the object, look thru its values and move them to their corresponding input controls in one shot. but in this case i will have to load possible values of country, then assign values, then load values of state, then assign value ans so on. I guess i am looking for an elegant solution. i am using asp.net, but i think it is irrelevant to the question. i am looking more into a design pattern. thanks

    Read the article

  • What do I name this class whose sole purpose is to report failure?

    - by Blair Holloway
    In our system, we have a number of classes whose construction must happen asynchronously. We wrap the construction process in another class that derives from an IConstructor class: class IConstructor { public: virtual void Update() = 0; virtual Status GetStatus() = 0; virtual int GetLastError() = 0; }; There's an issue with the design of the current system - the functions that create the IConstructor-derived classes are often doing additional work which can also fail. At that point, instead of getting a constructor which can be queried for an error, a NULL pointer is returned. Restructuring the code to avoid this is possible, but time-consuming. In the meantime, I decided to create a constructor class which we create and return in case of error, instead of a NULL pointer: class FailedConstructor : public IConstructor public: virtual void Update() {} virtual Status GetStatus() { return STATUS_ERROR; } virtual int GetLastError() { return m_errorCode; } private: int m_errorCode; }; All of the above this the setup for a mundane question: what do I name the FailedConstructor class? In our current system, FailedConstructor would indicate "a class which constructs an instance of Failed", not "a class which represents a failed attempt to construct another class". I feel like it should be named for one of the design patterns, like Proxy or Adapter, but I'm not sure which.

    Read the article

  • C# reference collection for storing reference types

    - by ivo s
    I like to implement a collection (something like List<T>) which would hold all my objects that I have created in the entire life span of my application as if its an array of pointers in C++. The idea is that when my process starts I can use a central factory to create all objects and then periodically validate/invalidate their state. Basically I want to make sure that my process only deals with valid instances and I don't re-fetch information I already fetched from the database. So all my objects will basically be in one place - my collection. A cool thing I can do with this is avoid database calls to get data from the database if I already got it (even if I updated it after retrieval its still up-to-date if of course some other process didn't update it but that a different concern). I don't want to be calling new Customer("James Thomas"); again if I initted James Thomas already sometime in the past. Currently I will end up with multiple copies of the same object across the appdomain - some out of sync other in sync and even though I deal with this using timestamp field on the MSSQL server I'd like to keep only one copy per customer in my appdomain (if possible process would be better). I can't use regular collections like List or ArrayList for example because I cannot pass parameters by their real local reference to the their existing Add() methods where I'm creating them using ref so that's not to good I think. So how can this be implemented/can it be implemented at all ? A 'linked list' type of class with all methods working with ref & out params is what I'm thinking now but it may get ugly pretty quickly. Is there another way to implement such collection like RefList<T>.Add(ref T obj)? So bottom line is: I don't want re-create an object if I've already created it before during the entire application life unless I decide to re-create it explicitly (maybe its out-of-date or something so I have to fetch it again from the db). Is there alternatives maybe ?

    Read the article

  • how do simple SQLAlchemy relationships work?

    - by Carson Myers
    I'm no database expert -- I just know the basics, really. I've picked up SQLAlchemy for a small project, and I'm using the declarative base configuration rather than the "normal" way. This way seems a lot simpler. However, while setting up my database schema, I realized I don't understand some database relationship concepts. If I had a many-to-one relationship before, for example, articles by authors (where each article could be written by only a single author), I would put an author_id field in my articles column. But SQLAlchemy has this ForeignKey object, and a relationship function with a backref kwarg, and I have no idea what any of it MEANS. I'm scared to find out what a many-to-many relationship with an intermediate table looks like (when I need additional data about each relationship). Can someone demystify this for me? Right now I'm setting up to allow openID auth for my application. So I've got this: from __init__ import Base from sqlalchemy.schema import Column from sqlalchemy.types import Integer, String class Users(Base): __tablename__ = 'users' id = Column(Integer, primary_key=True) username = Column(String, unique=True) email = Column(String) password = Column(String) salt = Column(String) class OpenID(Base): __tablename__ = 'openid' url = Column(String, primary_key=True) user_id = #? I think the ? should be replaced by Column(Integer, ForeignKey('users.id')), but I'm not sure -- and do I need to put openids = relationship("OpenID", backref="users") in the Users class? Why? What does it do? What is a backref?

    Read the article

  • When is a good time to start thinking about scaling?

    - by Slokun
    I've been designing a site over the past couple days, and been doing some research into different aspects of scaling a site horizontally. If things go as planned, in a few months (years?) I know I'd need to worry about scaling the site up and out, since the resources it would end up consuming would be huge. So, this got me to thinking, when is the best time to start thinking about, and designing for, scalability? If you start too early on, you could easily over complicate your design, and make it impossible to actually build. You could also get too caught up in the details, the architecture, whatever, and wind up getting nothing done. Also, if you do get it working, but the site never takes off, you may have wasted a good chunk of extra effort. On the other hand, you could be saving yourself a ton of effort down the road. Designing it from the ground up to be big would make it much easier later on to let it grow big, with very little rewriting going on. I know for what I'm working on, I've decided to make at least a few choices now on the side of scaling, but I'm not going to do a complete change of thinking to get it to scale completely. Notably, I've redesigned my database from a conventional relational design to one similar to what was suggested on the Reddit site linked below, and I'm going to give memcache a try. So, the basic question, when is a good time to start thinking or worrying about scaling, and what are some good designs, tips, etc. for when doing so? A couple of things I've been reading, for those who are interested: http://www.codinghorror.com/blog/2009/06/scaling-up-vs-scaling-out-hidden-costs.html http://highscalability.com/blog/2010/5/17/7-lessons-learned-while-building-reddit-to-270-million-page.html http://developer.yahoo.com/performance/rules.html

    Read the article

  • Is it wise to use temporary tables?

    - by Industrial
    Hi guys, We have a mySQL database table for products. We are utilizing a cache layer to reduce database load, but we think that it's a good idea to minimize the actual data needed to be stored in the cache layer to speed up the application further. All the products in the database, that is visible to visitors have a price attached to them: The prices are stored in a different table, called prices . There are multiple price categories depending on which discount level each visitor (customer) applies to. From time to time, there are campaigns which means that a special price for each product is available. The special prices are stored in a table called specials. Is it a bad to make a temp table that binds the tables together? It would only have the neccessary information and would ofcourse be cached. -------------|-------------|------------ | productId | hasPrice | hasSpecial -------------|-------------|------------ 1 | 1 | 0 2 | 1 | 1 By doing such, it would be super easy to know if the specific product really has a price, without having to iterate through the complete prices or specials table each time a product should be listed or presented. Are temp tables a common thing for web applications or is it just bad design?

    Read the article

  • OO vs Simplicity when it comes to user interaction

    - by Oetzi
    Firstly, sorry if this question is rather vague but it's something I'd really like an answer to. As a project over summer while I have some downtime from Uni I am going to build a monopoly game. This question is more about the general idea of the problem however, rather than the specific task I'm trying to carry out. I decided to build this with a bottom up approach, creating just movement around a forty space board and then moving on to interaction with spaces. I realised that I was quite unsure of the best way of proceeding with this and I am torn between two design ideas: Giving every space its own object, all sub-classes of a Space object so the interaction can be defined by the space object itself. I could do this by implementing different land() methods for each type of space. Only giving the Properties and Utilities (as each property has unique features) objects and creating methods for dealing with the buying/renting etc in the main class of the program (or Board as I'm calling it). Spaces like go and super tax could be implemented by a small set of conditionals checking to see if player is on a special space. Option 1 is obviously the OO (and I feel the correct) way of doing things but I'd like to only have to handle user interaction from the programs main class. In other words, I don't want the space objects to be interacting with the player. Why? Errr. A lot of the coding I've done thus far has had this simplicity but I'm not sure if this is a pipe dream or not for larger projects. Should I really be handling user interaction in an entirely separate class? As you can see I am quite confused about this situation. Is there some way round this? And, does anyone have any advice on practical OO design that could help in general?

    Read the article

  • Understanding MongoDB (and NoSQL in general) and how to make the best use of it

    - by Earlz
    Hello, I am beginning to think that my next project I am wanting to do would work better with a NoSQL solution. The project would either involve a ton of 2-column tables or a ton of dynamic queries with dynamically generated columns in a traditional SQL database. So I feel a NoSQL database would be much cleaner. I'm looking at MongoDB and it looks pretty promising. Anyway, I'm attempting to make sense of it all. Also, I will be using MongoMapper in Ruby. Anyway though, I'm confused as to how to layout things in such a freeform database. I've read http://stackoverflow.com/questions/2170152/nosql-best-practices and the answer there says that normalization is usually bad in a NoSQL DB. So how would be the best way of laying out say a simple blog with users, posts, and comments? My natural thought was to have three collections for each and then link them by a unique ID. But this apparently is wrong? So, what are some of the ways to lay out such a thing? My concern with the answer given in the other question is, what if the author's name changed? You'd have to go through updating a ton of posts and comments. But is this an okay thing to do with NoSQL?

    Read the article

  • Understanding MongoDB(and NoSQL in general) and How to make the best use of it

    - by Earlz
    Hello, I am beginning to think that my next project I am wanting to do would work better with a NoSQL solution. The project would either involve a ton of 2-column tables or a ton of dynamic queries with dynamically generated columns in a traditional SQL database. So I feel a NoSQL database would be much cleaner. I'm looking at MongoDB and it looks pretty promising. Anyway, I'm attempting to make sense of it all. Also, I will be using MongoMapper in Ruby. Anyway though, I'm confused as to how to layout things in such a freeform database. I've read http://stackoverflow.com/questions/2170152/nosql-best-practices and the answer there says that normalization is usually bad in a NoSQL DB. So how would be the best way of laying out say a simple blog with users, posts, and comments? My natural thought was to have 3 collections for each and then link them by a unique ID. But this apparently is wrong? So, what are some of the ways to lay out such a thing? My concern with the answer given in the other question is what if the author's name changed. You'd have to go through updating a ton of posts and comments. But is this an ok thing to do with NoSQL?

    Read the article

  • C++ interpreter conceptual problem

    - by Jan Wilkins
    I've built an interpreter in C++ for a language created by me. One main problem in the design was that I had two different types in the language: number and string. So I have to pass around a struct like: class myInterpreterValue { myInterpreterType type; int intValue; string strValue; } Objects of this class are passed around million times a second during e.g.: a countdown loop in my language. Profiling pointed out: 85% of the performance is eaten by the allocation function of the string template. This is pretty clear to me: My interpreter has bad design and doesn't use pointers enough. Yet, I don't have an option: I can't use pointers in most cases as I just have to make copies. How to do something against this? Is a class like this a better idea? vector<string> strTable; vector<int> intTable; class myInterpreterValue { myInterpreterType type; int locationInTable; } So the class only knows what type it represents and the position in the table This however again has disadvantages: I'd have to add temporary values to the string/int vector table and then remove them again, this would eat a lot of performance again. Help, how do interpreters of languages like Python or Ruby do that? They somehow need a struct that represents a value in the language like something that can either be int or string.

    Read the article

  • Question about decorator pattern and the abstract decorator class?

    - by es11
    This question was asked already here, but rather than answering the specific question, descriptions of how the decorator pattern works were given instead. I'd like to ask it again because the answer is not immediately evident to me just by reading how the decorator pattern works (I've read the wikipedia article and the section in the book Head First Design Patterns). Basically, I want to know why an abstract decorator class must be created which implements (or extends) some interface (or abstract class). Why can't all the new "decorated classes" simply implement (or extend) the base abstract object themselves (instead of extending the abstract decorator class)? To make this more concrete I'll use the example from the design patterns book dealing with coffee beverages: There is an abstract component class called Beverage Simple beverage types such as HouseBlend simply extend Beverage To decorate beverage, an abstract CondimentDecorator class is created which extends Beverage and has an instance of Beverage Say we want to add a "milk" condiment, a class Milk is created which extends CondimentDecorator I'd like to understand why we needed the CondimentDecorator class and why the class Milk couldn't have simply extended the Beverage class itself and been passed an instance of Beverage in its constructor. Hopefully this is clear...if not I'd simply like to know why is the abstract decorator class necessary for this pattern? Thanks. Edit: I tried to implement this, omitting the abstract decorator class, and it seems to still work. Is this abstract class present in all descriptions of this pattern simply because it provides a standard interface for all of the new decorated classes?

    Read the article

  • Avoid loading unnecessary data from db into objects (web pages)

    - by GmGr
    Really newbie question coming up. Is there a standard (or good) way to deal with not needing all of the information that a database table contains loaded into every associated object. I'm thinking in the context of web pages where you're only going to use the objects to build a single page rather than an application with longer lived objects. For example, lets say you have an Article table containing id, title, author, date, summary and fullContents fields. You don't need the fullContents to be loaded into the associated objects if you're just showing a page containing a list of articles with their summaries. On the other hand if you're displaying a specific article you might want every field loaded for that one article and maybe just the titles for the other articles (e.g. for display in a recent articles sidebar). Some techniques I can think of: Don't worry about it, just load everything from the database every time. Have several different, possibly inherited, classes for each table and create the appropriate one for the situation (e.g. SummaryArticle, FullArticle). Use one class but set unused properties to null at creation if that field is not needed and be careful. Give the objects access to the database so they can load some fields on demand. Something else? All of the above seem to have fairly major disadvantages. I'm fairly new to programming, very new to OOP and totally new to databases so I might be completely missing the obvious answer here. :)

    Read the article

  • Populating a foreign key table with variable user input

    - by Vincent
    I'm working on a website that will be based on user contributed data, submitted using a regular HTML form. To simplify my question, let's say that there will be two fields in the form: "User Name" and "Country" (this is just an example, not the actual site). There will be two tables in the database : "countries" and "users," with "users.country_id" being a foreign key to the "countries" table (one-to-many). The initial database will be empty. Users from all over the world will submit their names and the countries they live in and eventually the "countries" table will get filled out with all of the country names in the world. Since one country can have several alternative names, input like Chile, Chili, Chilli will generate 3 different records in the countries table, but in fact there is only one country. When I search for records from Chile, Chili and Chilli will not be included. So my question is - what would be the best way to deal with a situation like this, with conditions such that the initial database is empty, no other resources are available and everything is based on user input? How can I organize it in such way that Chile, Chili and Chilli would be treated as one country, with minimum manual interference. What are the best practices when it comes to normalizing user submitted data and is there a scientific term for this? I'm sure this is a common problem. Again, I used country names just to simplify my question, it can be anything that has possible different spellings.

    Read the article

  • How to manage data access / preloading efficiently using web services in C# ?

    - by Amadeus45
    Hello all, Ok, this is very "generic" question. We currently have a SQL Server database for which we need to develop an application in ASP.NET with will contain all the business logic in C# Web Services. The thing is that, architecturally speaking, I'm not sure how to design the web service and the data management. There are many things to consider : We need have very rapid access to data. Right now, we have over a million "sales" and "purchases" record from which we need to often calculate and load the current stock for a given day according to a serie of parameter. I'm not sure how we should preload the data and keep the data in the Web Service. Doing a stock calculation within a SQL query will be very lengthy. They currently have a stock calculation application that preloads all sales and purchases for the day and afterwards calculate the stock on the code-side. We want to develop powerful reporting tools. We want to implement a "pivot table" but not sure how to implement it and have good performances. For the reasons above, I'm not sure how to design the data model. Anybody can give me any guidelines on how to start, or from their personnal experiences (what have you done in the past ?) I'm not sure if it's possible to make a bounty even though the question is new (I'd put 300 rep on it, since I really need something). If you know how, let me know. Thanks

    Read the article

  • how to model a follower stream in appengine?

    - by molicule
    I am trying to design tables to buildout a follower relationship. Say I have a stream of 140char records that have user, hashtag and other text. Users follow other users, and can also follow hashtags. I am outlining the way I've designed this below, but there are two limitaions in my design. I was wondering if others had smarter ways to accomplish the same goal. The issues with this are The list of followers is copied in for each record If a new follower is added or one removed, 'all' the records have to be updated. The code class HashtagFollowers(db.Model): """ This table contains the followers for each hashtag """ hashtag = db.StringProperty() followers = db.StringListProperty() class UserFollowers(db.Model): """ This table contains the followers for each user """ username = db.StringProperty() followers = db.StringListProperty() class stream(db.Model): """ This table contains the data stream """ username = db.StringProperty() hashtag = db.StringProperty() text = db.TextProperty() def save(self): """ On each save all the followers for each hashtag and user are added into a another table with this record as the parent """ super(stream, self).save() hfs = HashtagFollowers.all().filter("hashtag =", self.hashtag).fetch(10) for hf in hfs: sh = streamHashtags(parent=self, followers=hf.followers) sh.save() ufs = UserFollowers.all().filter("username =", self.username).fetch(10) for uf in ufs: uh = streamUsers(parent=self, followers=uf.followers) uh.save() class streamHashtags(db.Model): """ The stream record is the parent of this record """ followers = db.StringListProperty() class streamUsers(db.Model): """ The stream record is the parent of this record """ followers = db.StringListProperty() Now, to get the stream of followed hastags indexes = db.GqlQuery("""SELECT __key__ from streamHashtags where followers = 'myusername'""") keys = [k,parent() for k in indexes[offset:numresults]] return db.get(keys) Is there a smarter way to do this?

    Read the article

  • Unity in C# for Platform Specific Implementations

    - by DxCK
    My program has heavy interaction with the operating system through Win32API functions. now i want to migrate my program to run under Mono under Linux (No wine), and this requires different implementations to the interaction with the operating system. I started designing a code that can have different implementation for difference platform and is extensible for new future platforms. public interface ISomeInterface { void SomePlatformSpecificOperation(); } [PlatformSpecific(PlatformID.Unix)] public class SomeImplementation : ISomeInterface { #region ISomeInterface Members public void SomePlatformSpecificOperation() { Console.WriteLine("From SomeImplementation"); } #endregion } public class PlatformSpecificAttribute : Attribute { private PlatformID _platform; public PlatformSpecificAttribute(PlatformID platform) { _platform = platform; } public PlatformID Platform { get { return _platform; } } } public static class PlatformSpecificUtils { public static IEnumerable<Type> GetImplementationTypes<T>() { foreach (Assembly assembly in AppDomain.CurrentDomain.GetAssemblies()) { foreach (Type type in assembly.GetTypes()) { if (typeof(T).IsAssignableFrom(type) && type != typeof(T) && IsPlatformMatch(type)) { yield return type; } } } } private static bool IsPlatformMatch(Type type) { return GetPlatforms(type).Any(platform => platform == Environment.OSVersion.Platform); } private static IEnumerable<PlatformID> GetPlatforms(Type type) { return type.GetCustomAttributes(typeof(PlatformSpecificAttribute), false) .Select(obj => ((PlatformSpecificAttribute)obj).Platform); } } class Program { static void Main(string[] args) { Type first = PlatformSpecificUtils.GetImplementationTypes<ISomeInterface>().FirstOrDefault(); } } I see two problems with this design: I can't force the implementations of ISomeInterface to have a PlatformSpecificAttribute. Multiple implementations can be marked with the same PlatformID, and i dont know witch to use in the Main. Using the first one is ummm ugly. How to solve those problems? Can you suggest another design?

    Read the article

  • How to justify using a scripting language as part of a project

    - by sylvanaar
    I have a specific project in which I want to use either a scripting language + C, or as an alternative a 100% Java solution. The program adapts a legacy system for use with other moderns systems. Basically, I have few choices as to what language I can use. I have C/C++, Java 1.4, and I have also compiled the Lua for this environment. The program does 'screen scraping' and has to deal with alot of strings. That part of the code is highly variable. Most of the developers at my company use C, so - my original design was to write some portions in C, and use Lua for the part that dealt with strings and changed freqently. I was told 'You have to justify your use of the scripting language.' So i reworked my design using 100% Java, and was told - Java wont have enough performance. You should do the whole thing in C. I'm not controlling lasers or doing image processing - just some screen scraping. I still have to provide justification for using anything but C - so what justification can I provide?

    Read the article

  • Is this WPF error handling a good idea ?

    - by Adiel
    I have a multi threaded wpf application with various HW interfaces. I want to react to several HW failures that can happen. For example : one of the interfaces is a temperature sensor and i want that from a certain temp. a meesage would appear and notify the user that it happened. i came up with the follwing design : /// <summary> /// This logic reacts to errors that occur during the system run. /// The reaction is set by the component that raised the error. /// </summary> public class ErrorHandlingLogic : Logic { } the above class would consume ErrorEventData that holds all the information about the error that occurred. public class ErrorEventData : IEventData { #region public enum public enum ErrorReaction { } #endregion public enum #region Private Data Memebers and props private ErrorReaction m_ErrorReaction; public ErrorReaction ErrorReactionValue { get { return m_ErrorReaction; } set { m_ErrorReaction = value; } } private string m_Msg; public string Msg { get { return m_Msg; } set { m_Msg = value; } } private string m_ComponentName; public string ComponentName { get { return m_ComponentName; } set { m_ComponentName = value; } } #endregion Private Data Memebers and props public ErrorEventData(ErrorReaction reaction, string msg, string componenetName) { m_ErrorReaction = reaction; m_Msg = msg; m_ComponentName = componenetName; } } the above ErrorHandlingLogic would decide what to do with the ErrorEventData sent to him from various components of the application. if needed it would be forwarded to the GUI to display a message to the user. so what do you think is it a good design ? thanks, Adiel.

    Read the article

  • C# Virtual method call in constructor - how to refactor?

    - by Cristi Diaconescu
    I have an abstract class for database-agnostic cursor actions. Derived from that, there are classes that implement the abstract methods for handling database-specific stuff. The problem is, the base class ctor needs to call an abstract method - when the ctor is called, it needs to initialize the database-specific cursor. I know why this shouldn't be done, I don't need that explanation! This is my first implementation, that obviously doesn't work - it's the textbook "wrong way" of doing it. The overridden method accesses a field from the derived class, which is not yet instantiated: public abstract class CursorReader { private readonly int m_rowCount; protected CursorReader() { m_rowCount = CreateCursor(sqlCmd); //virtual call ! } protected abstract int CreateCursor(string sqlCmd); } public class SqlCursorReader : CursorReader { private SqlConnection m_sqlConnection; public SqlCursorReader(string sqlCmd, SqlConnection sqlConnection) { m_sqlConnection = sqlConnection; //field initialized here } protected override int CreateCursor(string sqlCmd) { //uses not-yet-initialized member *m_sqlConnection* //so this throws a NullReferenceException var cursor = new CustomCursor(sqlCmd, m_sqlConnection); return cursor.Count(); } } I will follow up with an answer on my attempts to fix this...

    Read the article

  • Modeling many-to-one with constraints?

    - by Greg Beech
    I'm attempting to create a database model for movie classifications, where each movie could have a single classification from each of one of multiple rating systems (e.g. BBFC, MPAA). This is the current design, with all implied PKs and FKs: TABLE Movie ( MovieId INT ) TABLE ClassificationSystem ( ClassificationSystemId TINYINT ) TABLE Classification ( ClassificationId INT, ClassificationSystemId TINYINT ) TABLE MovieClassification ( MovieId INT, ClassificationId INT, Advice NVARCHAR(250) -- description of why the classification was given ) The problem is with the MovieClassification table whose constraints would allow multiple classifications from the same system, whereas it should ideally only permit either zero or one classifications from a given system. Is there any reasonable way to restructure this so that a movie having exactly zero or one classifications from any given system is enforced by database constraints, given the following requirements? Do not duplicate information that could be looked up (i.e. duplicating ClassificationSystemId in the MovieClassification table is not a good solution because this could get out of sync with the value in the Classification table) Remain extensible to multiple classification systems (i.e. a new classification system does not require any changes to the table structure)? Note also the Advice column - each mapping of a movie to a classification needs to have a textual description of why that classification was given to that movie. Any design would need to support this.

    Read the article

  • Customer wants some data to appear after you later delete rows. System giant / not my creation. Fast

    - by John Sullivan
    This is a fairly common problem, it probably has a name, I just don't know what it is. A.) User sees obscure piece of information in Row B of L_OBSCURE_INFO displayed on some screen at a certain point. It is in table L_Obscure_info. B.) Under certain circumstances we want to correctly delete data in L_OBSCURE_INFO. Unfortunately, nobody accounted for the fact that the user might want to backtrack and see some random piece of information that was most recently in L_OBSCURE_INFO. C.) The system is enormous and L_OBSCURE_INFO is used all the time. You have no idea what the ramifications are of implementing some kind of hack and whatever you do, you don't want to introduce more bugs. I think the best approach would be to create an L_OBSCURE_INFO_HISTORY table and record a record in there every time you change data. But god help your ensuring it's accurate in this system where L_OBSCURE_INFO is being touched everywhere and you don't have time to implement L_OBSCURE_INFO_HISTORY. Is there a particularly easy, clever design solution for this kind of problem -- basically an elegant database hack? If not, is this kind of design problem under a particular class of problems or have a name?

    Read the article

< Previous Page | 330 331 332 333 334 335 336 337 338 339 340 341  | Next Page >