New-age Transactional Systems - Not Your Grandpa's OLTP
John Hugg discusses high volume transaction processing applications with high and low frequency profiles, and how VoltDB can be used for that purpose.
News
The content has been bookmarked!
There was an error bookmarking this content! Please retry.
Presentation: 10 Ways to Improve Your Code
Posted by Abel Avram on Apr 16, 2009
In this presentation recorded during QCon SF 2008, Neal Ford, an architect at ThoughtWorks, shows 10 ways to write better code. This is practical advice for developers, but application architects can benefit from it too.
Watch: 10 Ways to Improve Your Code (1 hour)
Composed Method
- Divide your program into methods that perform one identifiable task.
- Keep all of the operations in a method at the same level of abstraction.
- This will naturally result in programs with many small methods, each a few lines long.
TDD - stands mostly for test-driven development but can also be viewed as test-driven design.
- When one uses TDD, he starts thinking about the consumer of the code he is writing. This brings awareness on how his code is going to be used by another.
- It forces the developer to mock dependent objects, resulting in a clearer picture of the relationship between the objects.
- TDD encourages the creation of composed methods.
Static Analysis
- It is quite useful to use tools like FindBugs which analyzes the bytecode to discover bugs.
Good Citizenship – refers to the way classes should react to one another
- Singletons are a bad choice because they have mixed responsibilities and are not testable. They are the objectual correspondents of global variables.
- The proposed solution is to create a regular POJO and use a factory to make sure that only one instance gets created.
YAGNI – You Ain’t Gonna Need It
- One should write only the code that is used now.
- Avoid speculative development which increases the software entropy of a code.
- Top 10 Corporate Code Smells:
1. There is a reason that WSAD isn’t called WHAPPY.
2. The initial estimate must be within 15% of the final cost, the post-analysis estimate must be within 10%, and the post-design estimate must be with 5%
3. We don’t have time to write unit tests (we’re spending too much time debugging)
4. We keep all of our business logic in stored procedures
5. The only JavaDoc is the Eclipse message explaining how to change your default JavaDoc template.
6. We have an Architect who reviews all code precheckin and decides whether or not to allow it into version control.
7. We can’t use any open source code because our lawyers say we can’t.
8. We use WebSphere because...(I always stop listening at this point)
9. We bought the entire tool suite (even though we only needed about 10% of it) because it was cheaper than buying the individual tools.
10. We invented our own web/persistence/messaging/caching framework because none of the existing ones was good enough.
Question Authority - Some things should not be done just because that is the custom.
Single Level of Abstraction Principle – everything should be at the same level of abstraction
- All lines of code in a method should be at the same level of abstraction
Polyglot Programming – use the best language for the problem while maintaining the same platform (JVM/.NET)
Learn Every Nuance – developers benefit from learning nuances of their language
Use the Anti-Objects Pattern – “an anti-object is a kind of object that appears to essentially do the opposite of what we generally think the object should be doing.”
RelatedVendorContent
agility@scale eKit: 10 Principles, Scaling Model, Metrics, Collaboration
Agile Development: A Manager's Roadmap for Success
Excelent !!
by
Erick Pimienta
Posted
Sharing Presentation?
by
Erick Pimienta
Posted
Re: Sharing Presentation?
by
Erick Pimienta
Posted
Re: Sharing Presentation?
by
Prodis a.k.a. Fernando Hamasak...
Posted
Not sure I agree Singleton's are Global Variables.
by
Porter Woodward
Posted
Re: Not sure I agree Singleton's are Global Variables.
by
Dmitry Tsygankov
Posted
Re: Not sure I agree Singleton's are Global Variables.
by
Pete Kirkham
Posted
Re: Not sure I agree Singleton's are Global Variables.
by
Manjunath Bhat
Posted
Re: Not sure I agree Singleton's are Global Variables.
by
Kevin Wong
Posted
Interesting Placement
by
James Moline
Posted
Slide and Presentation Timing issue
by
Adam Zimmerman
Posted
Re: Slide and Presentation Timing issue
by
jonty davis
Posted
Excellent
by
DSK Chakravarthy
Posted
Like all of this except using reflection of private constructors
by
Declan Whelan
Posted
Re: Like all of this except using reflection of private constructors
by
Stephan Kennedy
Posted
Singleton can be useful if used correctly
by
Jean-Simon LaRochelle
Posted
Very Good
by
Marcio Geovani Jasinski
Posted
10 Ways to Improve Your Code;Posted by Neal Ford on Apr 15, 2009 03:14 AM
by
Pradipta Dash
Posted
Thanks
by
Kamal Mettananda
Posted
-
Thanks Neal on sharing this 10 points and experience.
-
Is this presentation available for sharing? If so where I can download.
-
In practice - they are often used as such. In theory - they're not intended as such.
What I mean is that a Singleton can represent a resource of which there is a single instance. Often a logger or file system object, or some I/O port of which there is a single, physical instance of it. A singleton is thus a valuable way to model a real-world behavior.
Saying singletons are bad because they have mixed responsibilities, and are not testable - is the same as saying objects are bad because of the same reasons. Essentially you're dealing with bad singletons - which are also _bad_ objects. When they end up being used as a big library of static utility functions - that's when they're bad. At that point you're not dealing with object oriented code anymore, you've gone over to procedural code.
Properly engineered, and object oriented singletons are fine. Poorly engineered, not object oriented static function libraries are not. -
I find it funny that #10 under Coporate Code Smells:
"We invented our own web/persistence/messaging/caching framework because none of the existing ones was good enough"
is immediately followed by
"Question Authority - Some things should not be done just because that is the custom."
Seems a little incongruous. -
a logger or file system object, or some I/O port of which there is a single, physical instance of it.
But file system objects (and, therefore, loggers) are essentially the same old global variables! Global variables are in memory, files are on disk. Are there any other major differences? Both can be accessed by any thread if you have pointers to them (or filenames, or port numbers), locked in some way, concurrent access can lead to deadlocks, modularity can suffer whenever any of those things is used, one always has to access those resources in a certain order etc. An IO port is a global variable of type string, with one thread appending data to the end while another thread removes data from the beginning.
Of course, one has to do something in the end - write to a file or a database or something. But if you try to keep the core of the program clean of those things - you get a more maintainable program in the end. At least that's what I believe in. Some people don't... -
The code example given was the classic singleton anti-pattern - a hardwired, globally accessible single object which cannot be changed. What you seem to be talking about are objects of which there happens to be only one instance in the application (move to a different machine, you get an extra serial port; converting code which assumes a singleton to code which doesn't is a waste of effort).
If you hard wire the singleton as in the example, you get problems both in testing and when the singleton assumption is violated. If you create single instances of some types in your configuration, then you don't get these problems. You might say that YAGNI for two serial ports, but you do need it whenever you want to mock the object for tests, so even if there is one object in the domain of your first iteration, making it not a singleton gives an advantage. -
I like this presentation and found it confusing by the end with the timing of the slides and the content in the presentation.
I think there is a slide and timing issue at about minute 46. Anybody else seeing this? The slide changes to Swing + Ruby when dsl's are being talked about. Then the rest of the presentation is ~1 slide off.
Is this something that can be fixed? -
That is a great work and keen observation from SmallTalk stories. Surprised and unbelievable that these so called modular coding is first identified during 70s and documented by SmallTalk authoers.
Anyway, is there any possibility to download this presentation and play it offline with our dev folks!!
Apart of all.. thanks for the knowledge share -
Awesome presentation! I do have a quibble though.
I think creating private constructors and using reflection to invoke them is a questionable practice. Constructors should be appropriately visible and reflection should not be used to bypass language visibility when you have other options. Using reflection makes the code more complex, harder to maintain and limits the effectiveness of refactoring tools. And it still does not guarantee the singleton behaviour because reflection can be used anywhere.
Best to avoid the reflection in this case and make the constructor appropriately visible. -
This is certainly a pattern that can be misused. However, if you keep in mind the dangers of this pattern I think it can be used with good results. The best Singleton is stateless and of course read-only (the main problem of global variables is that they can be modified anywhere any-time so a read-only Singleton is not a store of global variables in the traditional sense). You can find good (constructive) critical analysis of the Singleton pattern on the Web.
-
Porter, Couldnt agree more with you. You put it very nicely. Singletons should be put to right use, not "abused". That's when it gets messy.
-
I agree with Declan.
I liked the presentation too, except the bit about using reflection to work around the private constructor.
I would create a public interface for the configuration class and hide the implementation class (which would have a public constructor) by making it visible only within its own package or even within the factory class. -
I think you're misinterpreting his point. He's not saying that there are not cases where there should only be one instance of an object (although if you're unit testing, there aren't), he's saying it's not the object's responsibility to manage its instances.
Specifically, I think he's referring to the pattern of a private constructor and a method/field to access the single instance. THAT "Singleton" is evil, indeed. That code using this pattern is horrible to test is solid evidence of excessive coupling, as well as a damning con in its own right.
Manage your object instances in a central place, e.g., DI container (Spring, Guice). -
Very nice presentation! Funny and goes smoothly.
Of course some topic are hard to talk in short time like singleton one...
I think Singleton becomes bad when it's misused and this happen with all languages, paradigms, etc... When someone learn a new language or a new pattern it's quite common start to use it every where and this is a bad programmer behaviour, doesn't mean the pattern/language is bad ;) -
Yeah I find this also at the same place, shame as it is such a great presentation, what a help!!
-
The link above is wrong.
The correct presentation can be downloaded in this link:
qconsf.com/sf2008/file?path=/qcon-sanfran-2008/... -
Excellent Presentation & Guide.
-
Thanks a lot for this nice presentation.
As most of the others, I also feel bad about the use of reflection on singleton.
19 comments
Watch Thread Reply