A rant
-
IMO, refactoring has to happen. Even if you wrote code the right way to begin with, requirements change. When requirements change, your code architecture has to change to reflect that. Often times, that's where refactoring comes in for us.
Marc Clifton wrote:
And come now, does refactoring really happen?
Absolutely. I do a little everyday at work.
Marc Clifton wrote:
Does your team ducttape (I always thought it was duck tape) the code and promise each other to refactor later?
We refactor when it's needed. I think the only exception is large refactors that would takes many, many months to complete. For example, after working in the same codebase for 2 years, we've recently started using test driven development. Much of our code up to this point wasn't designed with testing in mind, so if we really want to test a piece of code, it often has to be refactored. UI code is a big example here: I now see that most of the logic in our UI code simply cannot be tested because it's too intertwined with the UI code. So we've started to use the MVP pattern to seperate the logic and the UI code, thus allowing us to test the logic of the UI without getting entangled in UI code. But to take a timeout and refactor our hundreds of controls and forms to use the MVP pattern with unit testing would just be too huge a refactor. We simply don't have time to spend a couple months to do this. Instead, I'm migrating pieces at a time. Also, all new pieces use the MVP pattern with unit testing. So the code migrates over time. Besides the rare huge refactoring like that (maybe refactor isn't even the right word in that scenario), we do smaller refactors often, every day. Pulling interfaces out of classes, having consuming classes use interfaces (which can be mocked in unit testing) instead of concrete classes, renaming classes and namespaces to better reflect their usage (again, things evolve), making things simpler, the list goes on. In that sense, we refactor all the time. p.s. By the way Marc, [happy birthday](http://judahhimango.com/music/the beatles - birthday.mp3)[[^](http://judahhimango.com/music/the beatles - birthday.mp3)] man! Have a great, happy birthday weekend. :cool:
Tech, life, family, faith: Give me a visit. I
Judah Himango wrote:
Much of our code up to this point wasn't designed with testing in mind, so if we really want to test a piece of code, it often has to be refactored.
That kind of refactoring, sure I'm fine with that. But the "I'll refactor the compiler warnings later" problem, hmmm. And I can understand refactoring architecture to add features, improve testing, etc. I don't agree with putting off even minimal thinking about architecture and "we'll refactor that later". Some things ought be done more or less right, with the knowledge available at the time, and not just "coded".
Judah Himango wrote:
Have a great, happy birthday weekend.
:-D Thanks! Marc
People are just notoriously impossible. --DavidCrow
There's NO excuse for not commenting your code. -- John Simmons / outlaw programmer
People who say that they will refactor their code later to make it "good" don't understand refactoring, nor the art and craft of programming. -- Josh Smith -
Yes crap code happens every day. Why... lots of reasons, like crap developers (spend some time in the forums) and crap managers and crap companies.
Marc Clifton wrote:
And then I hear the "we need to get something out, and we'll refactor it later". Refactoring is abused, IMO.
Refactoring "should" occur when the design must change due to an analysis error or requirement change, not as an excuse to do crap work originally. And all of this is under the guise of "it's faster", which in my experience has never proven to be true, it's all bullshit. Anyway you don't need to believe me, see the web for Technical Debt[^]
led mike
led mike wrote:
Refactoring "should" occur when the design must change due to an analysis error or requirement change, not as an excuse to do crap work originally. And all of this is under the guise of "it's faster", which in my experience has never proven to be true, it's all bullsh*t. Anyway you don't need to believe me, see the web for Technical Debt[^]
Oh I believe you. And thank you for the link. It's great writeup. Short and to the point. Marc
People are just notoriously impossible. --DavidCrow
There's NO excuse for not commenting your code. -- John Simmons / outlaw programmer
People who say that they will refactor their code later to make it "good" don't understand refactoring, nor the art and craft of programming. -- Josh Smith -
The similar circumstances that I see all too often is "we'll skip unit testing because of time pressure and catch all the errors in integration". Firstly the project will never catch up and second the integration will take longer without unit tests. I once did a phase review on a project where the manager had just decreed that no more unit testing or code reviews would happen in order to keep on schedule. My report stated that I considered that too risky. However, the project manager asked if it were certain that integration would take longer than scheduled and I had to say no, but it was high risk. He went ahead anyway. Guess what - integration took much, much longer than scheduled because of all the bugs they had to fix. :sigh: I didn't get much satisfaction from being proved right. I did the analysis and the extra time spent in integration was more than the time saved in unit testing and code reviews. So yes, code should be reviewed frequently even in the midst of the continuous fire drill. Bad code will bite you (or someone else) in the end. As for refactoring, my initial reaction was yes I do it all the time. But reading other replies I realised that I do it as part of the natural development process when writing the code in the first place or because I'm changing a piece of old code because it now needs to meet new requirements and it's architecture is no longer sufficient. I cannot think of any time when the code has been refactored just because it is messy.
Graham My signature is not black, just a very, very dark blue
Graham Shanks wrote:
The similar circumstances that I see all too often is "we'll skip unit testing because of time pressure and catch all the errors in integration". Firstly the project will never catch up and second the integration will take longer without unit tests.
Yup. I've got a situation where the following has been said: "we'll refactor in the unit tests when we've shipped the first version of the product." :laugh: :(( :laugh: :(( Marc
People are just notoriously impossible. --DavidCrow
There's NO excuse for not commenting your code. -- John Simmons / outlaw programmer
People who say that they will refactor their code later to make it "good" don't understand refactoring, nor the art and craft of programming. -- Josh Smith -
I call it code slamming - generally. Then there is the code clearly written by someone who has no passion for the quality of their product.
Charlie Gilley Will program for food... Whoever said children were cheaper by the dozen... lied. My son's PDA is an M249 SAW. My other son commutes in an M1A2 Abrams
charlieg wrote:
I call it code slamming
That's a great phrase! Marc
People are just notoriously impossible. --DavidCrow
There's NO excuse for not commenting your code. -- John Simmons / outlaw programmer
People who say that they will refactor their code later to make it "good" don't understand refactoring, nor the art and craft of programming. -- Josh Smith -
Kevin McFarlane wrote:
Only the first is an undesirable situation. However, I think that's what you mean?
Yes, that's what I mean. Because #2 doesn't happen, even when the compiler screams warnings. Marc
Some people believe what the bible says. Literally. At least [with Wikipedia] you have the chance to correct the wiki -- Jörgen Sigvardsson
People are just notoriously impossible. --DavidCrow
There's NO excuse for not commenting your code. -- John Simmons / outlaw programmerSurely 2 does happen? Don't you ever write a method, get it working then go back and tidy it up - rename, extract methods, etc? Or do you mean that you do this yourself but most others don't?
Kevin
-
if(!connect()) return false; //continue with processing return true; That's a lot cleaner than bool ret = connect(); if(ret) { //continue with processing } return ret; And it saves a level of indenting. Maybe with 1280 x 1028 resolution this isn't such a big deal, but I still say there's nothing wrong with this style. Its just that, style.
This statement is false.
I generally aim for the single return but not religiously. I think in the case you mention where you have a return false at the top and then a single return at the bottom that's fine. Any other multiple return sceanrio is to be avoided. It's not such a big issue when methods are short, which they're supposed to be in good OO. But most programmers seem unable to write sufficiently short methods. Then multiple returns become a menace.
Kevin
-
charlieg wrote:
I call it code slamming
That's a great phrase! Marc
People are just notoriously impossible. --DavidCrow
There's NO excuse for not commenting your code. -- John Simmons / outlaw programmer
People who say that they will refactor their code later to make it "good" don't understand refactoring, nor the art and craft of programming. -- Josh SmithYeah, it tends to fit too. Analogy is "mule marks" in carpentry. This is where you miss the nail and leave a dent in the wood. Why mule marks? Because only an ass would make them... Geeze, I'm feisty today. I posted a rant in Dell's support forums, and they deleted it. Seems I put too much #$%#$^%$^ in the forum for their taste (it was all directed at Dell). Mistakes I can handle but making part of your process, pisses me off. Anyway, back to the code - the problem with code slamming, errr, it isn't good, everyone forgets the schedule pressures, so don't be *too* hard on the developer. There is usually more than enough blame to sling even though I would argue that one should get better at code slamming with more practice :mad:
Charlie Gilley Will program for food... Whoever said children were cheaper by the dozen... lied. My son's PDA is an M249 SAW. My other son commutes in an M1A2 Abrams
-
Surely 2 does happen? Don't you ever write a method, get it working then go back and tidy it up - rename, extract methods, etc? Or do you mean that you do this yourself but most others don't?
Kevin
Kevin McFarlane wrote:
Don't you ever write a method, get it working then go back and tidy it up - rename, extract methods, etc? Or do you mean that you do this yourself but most others don't?
Sure I do, but yes, it seems that on this project I'm helping out on, others are not doing that. Marc
People are just notoriously impossible. --DavidCrow
There's NO excuse for not commenting your code. -- John Simmons / outlaw programmer
People who say that they will refactor their code later to make it "good" don't understand refactoring, nor the art and craft of programming. -- Josh Smith -
Kevin McFarlane wrote:
Don't you ever write a method, get it working then go back and tidy it up - rename, extract methods, etc? Or do you mean that you do this yourself but most others don't?
Sure I do, but yes, it seems that on this project I'm helping out on, others are not doing that. Marc
People are just notoriously impossible. --DavidCrow
There's NO excuse for not commenting your code. -- John Simmons / outlaw programmer
People who say that they will refactor their code later to make it "good" don't understand refactoring, nor the art and craft of programming. -- Josh SmithMarc Clifton wrote:
Sure I do, but yes, it seems that on this project I'm helping out on, others are not doing that.
Seems to be the norm I'm afraid. It's rare that I maintain code that shows any courtesy to the maintenance programmer. :mad:
Kevin
-
Yeah, it tends to fit too. Analogy is "mule marks" in carpentry. This is where you miss the nail and leave a dent in the wood. Why mule marks? Because only an ass would make them... Geeze, I'm feisty today. I posted a rant in Dell's support forums, and they deleted it. Seems I put too much #$%#$^%$^ in the forum for their taste (it was all directed at Dell). Mistakes I can handle but making part of your process, pisses me off. Anyway, back to the code - the problem with code slamming, errr, it isn't good, everyone forgets the schedule pressures, so don't be *too* hard on the developer. There is usually more than enough blame to sling even though I would argue that one should get better at code slamming with more practice :mad:
Charlie Gilley Will program for food... Whoever said children were cheaper by the dozen... lied. My son's PDA is an M249 SAW. My other son commutes in an M1A2 Abrams
charlieg wrote:
the problem with code slamming,
LOL! I thought you meant that as a phrase for how coding is done, not how code is criticized.
charlieg wrote:
everyone forgets the schedule pressures, so don't be *too* hard on the developer.
True, but somewhere there's the line between doing something to save time and doing something, erm, dumb. Marc
People are just notoriously impossible. --DavidCrow
There's NO excuse for not commenting your code. -- John Simmons / outlaw programmer
People who say that they will refactor their code later to make it "good" don't understand refactoring, nor the art and craft of programming. -- Josh Smith -
charlieg wrote:
the problem with code slamming,
LOL! I thought you meant that as a phrase for how coding is done, not how code is criticized.
charlieg wrote:
everyone forgets the schedule pressures, so don't be *too* hard on the developer.
True, but somewhere there's the line between doing something to save time and doing something, erm, dumb. Marc
People are just notoriously impossible. --DavidCrow
There's NO excuse for not commenting your code. -- John Simmons / outlaw programmer
People who say that they will refactor their code later to make it "good" don't understand refactoring, nor the art and craft of programming. -- Josh Smithoh this is priceless :) I *do* mean how coding is done :doh: the fact that it might be interpreted either way makes it that much funnier...
Charlie Gilley Will program for food... Whoever said children were cheaper by the dozen... lied. My son's PDA is an M249 SAW. My other son commutes in an M1A2 Abrams
-
Chris S Kaiser wrote:
That's a lot cleaner than
Actually, I'll disagree. Ironically, you'll see exactly your example of "not cleaner" in my code. And the reason is because at some point, I actually needed to code an else statement when connection failed, which didn't require my having to check if there's a return statement earlier. Trivial example, but still. Marc
People are just notoriously impossible. --DavidCrow
There's NO excuse for not commenting your code. -- John Simmons / outlaw programmer
People who say that they will refactor their code later to make it "good" don't understand refactoring, nor the art and craft of programming. -- Josh SmithYeah, it was a trivial example. I've had to work on code where the method isn't broken out into multiple methods, and the ret val has to be checked again and again for validity for further processing. Which could get deep with regard to if blocks. So really I use it in the case where I didn't use exception handling. So the "correct" method in my case would be to throw an exception, but for some stuff I just return instead. But I don't think its a valid representation of a senior developer that doesn't care enough. Its situational.
This statement is false.
-
I generally aim for the single return but not religiously. I think in the case you mention where you have a return false at the top and then a single return at the bottom that's fine. Any other multiple return sceanrio is to be avoided. It's not such a big issue when methods are short, which they're supposed to be in good OO. But most programmers seem unable to write sufficiently short methods. Then multiple returns become a menace.
Kevin
Yeah, I won't disagree with that. Mostly though in a case where if block indentation goes more than three levels deep I'll use this mechanism if possible. But really as I said in my reply to Marc, I probably should be using exceptions in the extreme cases. I think this is bad if its not responsibly done. Blanket statements tend to grate on me. I don't write them but I still see the occaisonal goto used this way, and its argued that for errors that need to be further processing regardless of where they occured its "ok". Well, I guess we have the finally block to argue with now. ;)
This statement is false.
-
Graham Shanks wrote:
The similar circumstances that I see all too often is "we'll skip unit testing because of time pressure and catch all the errors in integration". Firstly the project will never catch up and second the integration will take longer without unit tests.
Yup. I've got a situation where the following has been said: "we'll refactor in the unit tests when we've shipped the first version of the product." :laugh: :(( :laugh: :(( Marc
People are just notoriously impossible. --DavidCrow
There's NO excuse for not commenting your code. -- John Simmons / outlaw programmer
People who say that they will refactor their code later to make it "good" don't understand refactoring, nor the art and craft of programming. -- Josh SmithMarc Clifton wrote:
I've got a situation where the following has been said: "we'll refactor in the unit tests when we've shipped the first version of the product."
Let me guess - it never happened :)
Graham My signature is not black, just a very, very dark blue
-
Yeah, I won't disagree with that. Mostly though in a case where if block indentation goes more than three levels deep I'll use this mechanism if possible. But really as I said in my reply to Marc, I probably should be using exceptions in the extreme cases. I think this is bad if its not responsibly done. Blanket statements tend to grate on me. I don't write them but I still see the occaisonal goto used this way, and its argued that for errors that need to be further processing regardless of where they occured its "ok". Well, I guess we have the finally block to argue with now. ;)
This statement is false.
Yes, a lot of these rules are guidelines which you should follow unless you have a reasonable reason not to. There are often good exceptions. Also a comment explaining why you've violated the guideline is often worthwhile, e.g., a typical case might be a swallowed exception. At least this tells the code maintainer that you've thought about what you were doing. Re: goto, there's an example in Steve McConnell's Code Complete 1st edition (dunno whether it's in the 2nd ed) where there's some code with a goto in it and virtually everyoe who tries to rewrite it without the goto does so incorrectly!
Kevin
-
Yes, a lot of these rules are guidelines which you should follow unless you have a reasonable reason not to. There are often good exceptions. Also a comment explaining why you've violated the guideline is often worthwhile, e.g., a typical case might be a swallowed exception. At least this tells the code maintainer that you've thought about what you were doing. Re: goto, there's an example in Steve McConnell's Code Complete 1st edition (dunno whether it's in the 2nd ed) where there's some code with a goto in it and virtually everyoe who tries to rewrite it without the goto does so incorrectly!
Kevin
Another example would be in an assignment operator in C++:
SomeObject &operator=(const SomeObject &rhs)
{
if(&rhs == this)
return *this;//assign other members return \*this;}
This is recommended by Scott Meyers. So I still say its a matter of intelligent use and style. I worked with people from both camps. Some prefer one, as it makes it more readable, and others prefer the second as its safer as a habit. But that's what I'm getting at. People consider it bad style because of irresponsible usage. And this brings up another point. For greenhorns this might be a good rule. But to say that a senior developer doesn't care about his code because of multiple returns, is a bit brash and not qualified as a blanket statement. -- modified at 13:10 Saturday 19th August, 2006 ~Added pre tag to code section.
This statement is false.
-
Another example would be in an assignment operator in C++:
SomeObject &operator=(const SomeObject &rhs)
{
if(&rhs == this)
return *this;//assign other members return \*this;}
This is recommended by Scott Meyers. So I still say its a matter of intelligent use and style. I worked with people from both camps. Some prefer one, as it makes it more readable, and others prefer the second as its safer as a habit. But that's what I'm getting at. People consider it bad style because of irresponsible usage. And this brings up another point. For greenhorns this might be a good rule. But to say that a senior developer doesn't care about his code because of multiple returns, is a bit brash and not qualified as a blanket statement. -- modified at 13:10 Saturday 19th August, 2006 ~Added pre tag to code section.
This statement is false.
Chris, you've just reminded me of some code I was maintaining several years ago that was generally well-written. However, the author did religiously follow the single return rule and in one case it led to a bug which would have been avoided had he relaxed the rule. A bit similar to McConnell's goto example.
Kevin