Introduction
Notes of a soldier from the oh-so-bloody front. Depending on the scope of your project, you may have the opportunity for code reuse. The reasons you might want to do so are two-fold.
First, you reduce duplication of efforts. If you have already created a Hyperlink enabled CellRenderer for your DataGrid once, why do it again?
Second, you create, or build upon, an ever growing utility code base. While it may not be in the “utils” package per se, you’ll soon end up with re-usable events, common GUI controls and widgets, and yes even utility classes. Whether by merely being in a different folder means the client doesn’t own it is up to you or your sales team.
Duplication Killed on Sight
On the current project I’m on, we re-use a LOT. My eagle-eye boss the architect head-shots any duplication he sees. Thus, we the developers have been trained to quickly identify something we create for re-use if possible and either plan accordingly when building it, or go in search of “spare parts” from existing classes throughout the main code-base for re-use. Our main base consists of 3 company names, each containing at least 3 individual “products”. The first question I ask my fellow developers, and they in return, when starting to build something new is, “Has this been done before?” Re-inventing the wheel has no place in a production cycle unless you can clearly point out how the original wheel design was flawed, can be done better, and done so in a reasonable time frame approved by the client.
This also helps ensure I don’t code something similar to what has already been done. To the client’s point of view, they already paid for a LoginForm… why should they have to pay for a new one when all they want is for it to be green instead of blue?
“Coworker, the login form’s blue by default, but the mockups I have here are green.”
“That’s because the View using it sets it to blue; it has a color property that is an inherited style; you can style it yourself, just do color=’#yourcolor’.”
“Oh… nice!”
Suddenly, I spend 10 minutes finding the file & including it, setting it’s color, registering for it’s login event, and building a test file to see it in action. This instead of 4 hours doing the same building a green one. Take how much you make per hour, multiply by 4, and then sub-tract what you make in 15 minutes. That difference is what you just saved the client by “asking a question”. I’ve learned the hard way to ask a lot of questions to the point of being annoying and forcing people to repeat themselves.
The same goes for building more complex things that have to be unique to a point. The most common example which I already alluded to is the CellRenderer. This is a class commonly utilized in Flex & Flash development to customize what is showing in a DataGrid column; each row will render the custom class instead of the default and pass it an item to render. Since every DataGrid on a project is unique, you inevitably end up with a multitude of cellrenderer classes. These classes are, at least for me, notoriously hard to share so you attempt to make them as generic as possible so others can either use them as is, or do the most common thing and extend your base one to customize it to their needs.
This is an important point. The 30 or so lines of code that are required to setup a cellrenderer are suddenly already written for you, and the other developers on your team. This is a great place where inheritance really works and should be exploited. It doesn’t stop there; some of the cellrenderers created could be used elsewhere as well. The only challenge is how are they designed from a visual standpoint. Styles can handle a lot, but most Designers I have had the pleasure of working with have a knack making something unique, and unique isn’t always re-usable. You can either find a happy compromise with your designer, or respect it’s artistic integrity and recognize the fact that it is truly made specifically for a certain need and shouldn’t be made re-usable.
Con of a Component Buffet
The first con to reusing already built components hits the design side hardest. One of the biggest grips people have had with any Flash / Flex component is styling and skinning. While CSS styling has come a long way in Flex 2, there are always those times where the artist (in you or beside you) goes, “It’s just not right…”. Sometimes extending the base component just for styling purposes is the best repose since the base component isn’t muddied with application specific styling routines.
The second con which can really come to a front in teams is you just don’t like the component. Any developer who isn’t apprehensive about using a component they didn’t write makes me nervous. You trust code you didn’t write? Sometimes you don’t have a choice or recognize the alternatives are unacceptable. We’re all human, and have our coding styles, and even if notation and other rules are enforced on your team, you can still dislike the implementation of something. This should be in the back of your mind when creating code for reuse as well. How will my code be perceived? The base rules such as encapsulation should be followed, but obviously there are other esoteric and styles of implementation that can drastically impact your involvement to ensure others spend very little of theirs getting acclimated to how it works for example.
Sum Greater Than Its Parts
For components that are made up of other components via Composition, I’ll ask my fellow developers if the pieces I need are already built. For example, if I’m creating a form, I’ll re-use the extended TextInput’s we have. For the above CellRenderer, if it has to display a Date, I’ll set the labelFunction to utilize the DateUtils class we have which will take a regular Date object and make it look like, “Tue 5/23/2006”. I didn’t have to write the base class of the CellRenderer and I didn’t have to write a class to format dates, I just had to use the class. I’ve been doing the same thing for a long time without really thinking about it. For example, I take for example I can just “use a DataGrid” and “extend it with a CellRenderer”. Macromedia / Adobe spent a lot of time developing one that could be reusable. Naturally, there comes a point where something is coded to business rules or a certain design, and the pragmatic in me knows when to stop trying to overdo it.
Let’s turn it up a notch, though. What if you’ve built a DataGrid that has custom cellrenders, text fields on the bottom to filter it, and is tied to a specific ValueObject it knows how to display, and display well? Can you re-use that? Absolutely! Just because you don’t right now doesn’t mean you won’t later. You should typically design with the intention to do so, but not so much you don’t ever actually complete anything beyond a pimped out skeleton. Adamantine ingots may have great potential, but are merely blocks of immobile metal until merged with, say, a regenerative Canadian supersoldier’s skeleton. Put what you write to use sooner rather than later.
This is where code reuse really shines. You’re near mini-application status component which does a lot is now re-used by your team. If the components within it follow normal styling rules and expose their innards to those who wish to extend it, your golden if certain custom styling needs to be applied. What could possibly be a con about this?
Package Structure
Package structure. Package structure, for those who don’t know, is how code is organized into folders. Folders are called “packages” because they have code and other packages in them depending on how deeply you nest your code. Code is placed into class files (.as or .mxml) and placed into folders. The typical naming scheme goes deployment type, company, project, and then regular code. This can take the form of com.adobe.utils.DateUtils where you have a com folder that contains an adobe folder which in turn contains a utils folder which contains the DateUtils.as file. You then import the class or “package path” into your code and the class DateUtils, and your code will know what folder to look in from the import statement.
Some projects do get large enough where you do in fact have more than 1 project folder. Code re-use is typically thought of through View’s of some sort client developers. There is however, no reason you cannot re-use View’s across projects. You just reference the package path. The pro’s are, work effort utilized on a project can be re-used on other projects. Sometimes you can plan for this re-use, and sometimes it’s a pleasant surprise.
Licenses
I suppose depending on license that you could use it for different companies as well, but each license has specific rules on how code is used. Creative Commons is pretty simple; just keep the author’s name in the code unless she/he says otherwise. From that point forward, you can modify and re-modify to your heart’s content with no license fee. Others are a lot weirder. Some companies require you supply them source code. Some don’t even know what source code is. Still others own everything you write while in their presence, thus preventing you from using any of your own re-usable stuff. A common tactic I’ve seen is to have your company’s name next to the client’s company name. All common view’s and utility classes are put in yours, and the custom developed work is put in the client’s package path.
Global Ramifications
That last point is another important point. Code by it’s very nature residing in your company’s folder has a very special, and important place; it’s created for re-use. On the current project, or projects, it’s re-used in many places. While all uses of it are immediately improved upon once you improve the base component, thus improving the whole project, this can have unforeseen consequences as well.
For example, you build your own List control to render extremely unique rows that animate instead of simply refreshing. In practice, you see that it is extremely slow, especially when multiple instances of it are used. You go in an re-factor it, finding many places in the code you can speed up. Suddenly, the entire app speeds up where those components are used.
One of these changes was to expose a method for updating one specific item versus all of them when a piece of data changes. This is accomplished via a new method. That is a bad thing. Suddenly, all the classes that use it now have to update themselves just to access one of the new optimizations.
Touching a point I brought up earlier, sharing views between projects. The problem with that is if you change a view in one project, you’ve just affect the other one. Code that was seemingly “working” you just broke, and you didn’t know it because you were never compiling, nor even involved in the other project. The important point here is if you are going to identify something as re-usable, make it so, and put it into a global package such as your company’s folder, or another aptly named common folder. Developers who go into the code in those packages know full well the ramifications of what their changes could do for good or ill.
Tool Shed
When I got my first apartment, I had 1 screwdriver and 1 small Philips head to my name. Now that I own a house, I have a small utility closet full of tools. I’m sure by the time I’m 80, I’ll need a tool shed to hold them all. The same can be said about your “common” code base. You’ll find over time that it’ll grow into an extremely useful and portable set of code. Remember, you don’t have to write it all yourself, nor does your team. There is a lot of free code on the net that you can test yourself, and then incorporate. This’ll save you from having to create a hammer before you can use said hammer every project; going in, you’ll be equipped to the teeth, and you’ll come out with more ammo than you started with.
Conclusions
Reusing code becomes more important that greater the size of the project, and/or in the frequency of projects. Making the most of a developer’s time spent coding is done this way, and allows many others to benefit from that work for months, even years to come. For example, I’ve been using the same preloader in Flash for 3 years, written in ActionScript 1 in Flash 5. I still customize the colors, though, every time.
With reuse comes great responsibility. If you are designating something as reusable, it can be quite frightening to realize that a lot of others are suddenly depending on this pinnacle piece of code. That is a risk worth taking, and a strength you should prey upon. As long as you recognize the dependencies, and reduce coupling, you’ll start gaining a lot more efficiency from the time you and your team spend coding.
When I first learned about reuse, I could never get it to work in practice. I either ran out of time and copy and pasted or my original design was perceived to me as being “flawed” even though it worked perfectly fine. I was just being unreasonable. Keep it simple, and you’ll do fine.
This entry in Word, Flash Paper, and PDF.