http://ayende.comAyende @ RahienCopyright (C) Ayende Rahien 2004 - 2021 (c) 202660There is an interesting discussion going on between Eli Lopian , Roy Osherove and Oren Eini on whether http://ayende.com/2155/the-production-value-of-seams#comment9http://ayende.com/2155/the-production-value-of-seams#comment9Mon, 05 Mar 2007 13:09:18 GMTIn response to Ayende: " Would you think to do it at first if it was something that wasn't driven by the need for testability?" I can't answer that. I am so used to TDD so testability could be a subconscious desicion. Or it could be that TDD also taught me how to code better. A BatchFileWriter would of course be testable, but my main reason, I think, is to encapsulate the behavior. http://ayende.com/2155/the-production-value-of-seams#comment8http://ayende.com/2155/the-production-value-of-seams#comment8Mon, 05 Mar 2007 08:31:20 GMT[new post] http://ayende.com/2155/the-production-value-of-seams#comment7http://ayende.com/2155/the-production-value-of-seams#comment7Sun, 04 Mar 2007 21:57:26 GMTNope, the ability to supply an alternate implementation is a side benefit to the testability benefits. I would agree with YAGNI, except that this is something that I do need, in order to decrease coupling and enable better testing. Intersting post, I'm going to reply in a new post. http://ayende.com/2155/the-production-value-of-seams#comment6http://ayende.com/2155/the-production-value-of-seams#comment6Sun, 04 Mar 2007 20:55:25 GMTOren, "I MAY want to supply alternate implementation" Sounds Like "You Aint Gonna Need It" See here: http://www.elilopian.com/2007/03/04/design-and-testability-yagni/ http://ayende.com/2155/the-production-value-of-seams#comment5http://ayende.com/2155/the-production-value-of-seams#comment5Sun, 04 Mar 2007 20:39:14 GMTWould you think to do it at first if it was something that wasn't driven by the need for testability? http://ayende.com/2155/the-production-value-of-seams#comment4http://ayende.com/2155/the-production-value-of-seams#comment4Sun, 04 Mar 2007 20:33:33 GMTMy naive approach would be to introduce a BatchFileWriter and put all calls to File.WriteAllText in there. http://ayende.com/2155/the-production-value-of-seams#comment3http://ayende.com/2155/the-production-value-of-seams#comment3Sun, 04 Mar 2007 19:26:48 GMTI understand your point, but Helpers.CreateFile() is still not really testable, is it? And you are talking to someone who is using 4 layers of abstractions just to get a value from a text box. :-) http://ayende.com/2155/the-production-value-of-seams#comment2http://ayende.com/2155/the-production-value-of-seams#comment2Sun, 04 Mar 2007 14:17:06 GMTIt's all a question of compromise between code conciseness and the ability to inject different code or behavior. Coming from a more performance-oriented, systems-programming background, I prefer the former approach. I think that for anyone not well into TDD, the code sample above (with the IoC container) is very difficult to understand or follow. A more traditional (and not necessarily a less powerful one) would be to simply provide a factory method in a utility class; I think using(TextWriter writer = Helpers.CreateFile(filename)) is much easier to follow than using(TextWriter writer = IoC.Resolve<IFileWriterFactory>().Create(filename))). http://ayende.com/2155/the-production-value-of-seams#comment1http://ayende.com/2155/the-production-value-of-seams#comment1Sun, 04 Mar 2007 14:09:37 GMT