Theory Question

Question

Here is one of those fun conversation starters... I'm looking for differing opinions on programming theory...here is the synopsis... ---------------------- I am building a Web Service to act as the communication liasion between all of our corporate apps and our corporate databases. Would it be more beneficial to: Build one Method to connect to the DB and get each peice of data seperately (reducing the amount of data that gets passes and only returning what is actually needed but requiring more programming) or Building a few GetAllMYData type methods that return lots of values whether they are needed or not for the given instance. (which would require less programming and maintenance but would often contain data I don't need)

marvin7 · Answer

What's the easiest to maintain? That should be your primary concern. I can almost guarantee you that this function(s) are going to change in the future. If you have ONE method you have a very good chance of breaking something when you update it. That being said you definitely want to compartmentalize your code so that it's easier for new developers that look at your code to debug understand maintain and extend without breaking something... and be able to do it quickly. Don't worry about more programming or less programming. When you try to save on programming someone pays the interest later. Just do it right. It sounds like this is pretty big so I'm leaning towards the latter (a few GetAllMyData type methods). Optimization will be necessary no matter what so less maintenance is almost always the deciding factor. Optimizing monolithic code can be a real nightmare especially after it's gone through dozens of revisions over the life cycle of the program.

John_Boals · Answer

Generally speaking I'd go with option A though there'll be exceptions (always is in I.T. right?). The problem I see with GetAllMyData() is that it doesn't really reduce the programming effort; it just moves it from the business object to the client where it doesn't belong. Example. You've just assigned Developer1 to write a client app that displays the data either by customer or by month. So he makes a screen that lets user enter customer id or month and then he calls GetAllMyData(). The client app now has to have the intelligence to cull through the results and display for the given month or customer. This obviously is done a lot more efficiently on the server than on the client so it'd be better if the client could call a couple different methods depending on what the user wants e.g. GetDataForCustomer(int ACustomerID) or GetDataForMonth(int AMonth). It can get worse. Let's say there's a business rule that the month ends on the last Sunday of the month. So when Developer1 wrote the client app above he had to program in that logic say into a C# windows app. Now users in the field want to get at that data via the web. So you assign Developer2 and he builds a web client say using Java. In the best case he calls GetAllMyData() and essentially re-writes the same algorithms in Java that had been written in C#. Worst case is he's not aware of the eom rule and writes the algorithm for a calendar month. Now the web users are seeing different data than the window app users and boy are we in trouble. These problems are minimized by having a GetDataForMonth(int AMonth) method in the business class that knows what a month really is. All clients calling GetDataForMonth get the same results. All that being said I think there are circumstances when it might be advantageous to bring back more data than the client might need but this message is long enough. As this thread progresses I may toss out some other ideas. My main immediate concern is your statement about more or less programming being required between A and B. I think you're going to have to program it somewhere it just depends where.