InfoQ Homepage Presentations Heresies and Dogmas in Software Development
Heresies and Dogmas in Software Development
Summary
Dean Wampler discusses the merits of several controversial issues: Goto, Design before Code, Design Patterns, Corba vs. REST, Object Middleware and ORMs, and Identifiers with Spaces.
Bio
Dean Wampler specializes in Scala and “big data” analytics using the Hadoop ecosystem of tools. He is a Principal Consultant at Think Big Analytics (http://thinkbiganalytics.com) and the co-author of Programming Scala and author of Functional Programming for Java Developers, both from O’Reilly Media. Dean has a Ph.D. in Physics, but he’s mostly recovered.
About the conference
Strange Loop is a multi-disciplinary conference that aims to bring together the developers and thinkers building tomorrow's technology in fields such as emerging languages, alternative databases, concurrency, distributed systems, mobile development, and the web.
Community comments
Strange Loop
by Alex Miller,
Nice talk
by Duraid Duraid,
All edges, no body
by William Cherry,
Identifiers With Spaces
by Ryan Scheel,
Strange Loop
by Alex Miller,
Your message is awaiting moderation. Thank you for participating in the discussion.
If you're interested in other upcoming videos from Strange Loop, the full release schedule is here and all slides are here. If you want to be notified about Strange Loop announcements in the future, sign up for the mailing list.
Nice talk
by Duraid Duraid,
Your message is awaiting moderation. Thank you for participating in the discussion.
The principle that I use, which I think holds for all these examples, is: a software solution lies in the middle between the problem domain (user stories, requirements, etc) and the solution domain (database, functional, OO, ORM, REST, CORBA, etc). The closer to the problem domain, the better solution it is albeit less efficient (in general) and vice versa.
It's easy to see how this applies to the goto statement.
In regard to the object middleware and the idea of functional 'wrappers', I think that's a completely wrong idea. I don't think the purpose of functional programming is to remove the need for object middleware (or object models) but the idea is to replace object models with a functional models that claim to have better composability, reusablity and closer representation of the problem domain but definitely not in order to gain efficiency. For efficiency you just need to be closer to the solution domain, or in other words, closer to the machine and not use functional nor object programming but C or assembly.
All edges, no body
by William Cherry,
Your message is awaiting moderation. Thank you for participating in the discussion.
It seemed that a lot of the advice was based on edge cases. Yes, there are edge cases where a goto makes sense, 99.5% of the time it makes code harder to read and can introduce subtle bugs (e.g. jumping over cleanup code). Yes, writing raw SQL can make your app faster. But raw SQL is harder to understand then a well defined method name. If you need to raw SQL in spots to get the speed that you need then use the ORMs ability to drop down to the native level in that one section.
And remember that these "premature optimization" are going to be maintained by the new intern and you don't want him thinking that a goto would be perfect to fix that stack overflow.
Identifiers With Spaces
by Ryan Scheel,
Your message is awaiting moderation. Thank you for participating in the discussion.
There are cases where spaces in identifiers are useful. Javascript has the same property.
var o = {
"identifier with spaces" : function () { return 0; }
};
console.log(o["identifier with spaces"]);