At some point in everyone's career, we all write crappy code (except my friend Mark Lucas, but he's in a class by himself). You know the kind: You go back to it six months later to add new features or better error-checking, and you have no idea what you were doing. It's "spaghetti code", and it can eat a lot of time as you try to figure things out all over again. That's where this document comes in.
A few healthy habits can make your code more robust and your workday shorter. While they may require a little self-discipline at first, if you find these practices useful and adopt them in your daily work they'll become second-nature and eventually require no extra effort at all.
There are many reasons to use good style practice when writing code, including:
The tips presented here reflect some of the best practices of many world-class scripters, and we should take a moment to acknowledge them here:
Michael Silver, Ken Ray, Mark Lucas, and Mark Hanrek are some of the best SuperCard scripters I've known, and it was while we were sharing code that we discovered that we had each independently developed pretty much the same code style. The first draft of this document reflected our common interest in SuperTalk, and was accordingly more limited in scope. It was distributed as part of a panel discussion on scripting techniques at the SuperCard Developer Conference in 1996, the year SuperCard 2.5 won the Mac Eddy award for Best New Multimedia Authoring Tool.
Since then this document has been revised many times, expanding its scope to include new tips for other similar scripting languages. Along the way I've collected valuable tips from a great many programmers, including Scott Raney, Kevin Miller, Steven McConnell, Jad Santos, and Alex Bronstein. With the things these folks have taught me, this document should hopefully be useful for scripters who prefer just about any HyperTalk dialect, and a few others besides.
As you read these tips, you may feel at times that where you code style differs you kinda like it that way. That's okay. If it works for you, it don't need fixin'.
These are only guidelines, and the more people use them the more easily people will exchange code happily, solve problems more quickly, spend more time with their families, and create an ever better world.
Having said all that, it is never worthwhile to rewrite existing code simply to conform to these suggestions. Nor is it worthwhile to stick to a guideline where the function or efficiency of the code is compromised. There is a fine line between making more maintainable code and more efficient code. It is up to you to decide where that line is in your own code.
This document is a work in progress. We use it here in the course of teaching and working with contractors, and accordingly it is always being ammended and enhanced as experience suggests and time permits. If you have ideas for tips and techniques which you would like to see included here, please email them to me at firstname.lastname@example.org.
With all that out of the way, here's the beef:
General Scripting Techniques
A few techniques for getting the spaghetti out of your code....
Some programmers will fight more ardently over whether to list their handlers top-down or bottom-up than they will over whether to order single-malt or blended. Bad priorities, IMO: pick one code listing method and be done with it, and spend your brawling time over more important things like whiskey.
Code tells a story, and both top-down and bottom-up tell the same story but each with a different emphasis. Bottom-up code has core handlers at the top, with the functions that call them below, and system messages at the bottom. Top-down puts system messages at the top, the more complex handlers that use them below, with generalized utiliy routines at the bottom.
Personally, I like to tell a story in the order in which the events occurred, and since events in a script begin with system messages, it seems a more natural reading to put those first.
Clarify Program Flow
One way to fight spaghetti code is to make sure your code's logic is self-evident. For example, if you have one long handler which takes care of a variety of startup stuff, you could break that down into multiple handlers which handle only specific related tasks. If you use descriptive names, you'll have no trouble grasping how the first handler works:
The extra time required by the interpreter to handle these separate calls is trivial, and isolating specific sets of related routines in this way can help you track down the source of errors during debugging.
One good technique for making sure your code's flow is evident is to write the comments first. Just make an outline of what the code needs to do in comments, then go back and fill in between them with the actual executable lines. This technique can also be useful in the early stages of tackling a difficult routine, helping you think through the stages needed to complete it.
Many words in scripting languages can be abbreviated. Where available, use 'em. They save space, and by being visually less significant they help your variables and handler names stand out. Plus you'll save a little typing.
You can use parentheses to group logical or arithmetic expressions or to clarify code:
if the vis of wd "myWindow" = true
Also, always using "=" for comparisons instead of "is" will help make such comparisons stand out visually.
When nesting conditions I've found it helpful to use a separate line for each condition, making use of the "\" line continuation character and putting the conjunction in all-caps to draw attention to it:
if (tWd is among the lines of the windows) \ AND (the vis of stack tWd is true) \ AND (the mode of stack tWd = 1) then ... end if
This means a few extra characters when you're typing, but the clarity is well worth the effort. Conditional expressions are the heart of complexity, and the more complex your code is the more difficult it can be to maintain and enhance. Giving them a little extra attention when you write you code can pay big dividends down the road if something breaks.
As a general rule, it's useful to comment any block of code whose functionality is not immediately self-evident. For example, it's merely distracting to see:
-- Set the cursor to watch: set the cursor to watch
But it's very helpful to have a brief note like this for more complex blocks:
-- Update each card with the current date: set the cursor to watch put the short date into tDate repeat with i = 1 to the number of cds put tDate into fld "Date" of cd i end repeat
Comments can also be used as visual guides, separating different sections of code. Here's an example of how sets of related handlers can be grouped visually using commented lines:
--=============================================-- -- WINDOW ROUTINES -- -- -- DocWindowRect() -- -- Returns the default size for new document windows -- function DocWindowRect put the screenrect into r add 4 to item 1 of r add 4 to item 2 of r subtract 4 from item 3 of r subtract 4 from item 4 of r return r end DocWindowRect -- -- UpdateAllWindows -- -- Allows each open window to refresh itself -- on UpdateAllWindows put windows() into tWdList repeat for each line tWd in tWdList send "UpdateThisWindow" to tWd end repeat end UpdateAllWindows --=================================================--
Note that the example above also includes a brief description of each handler. This makes handlers stand out more clearly when skimming, and provides a useful overview as to its purpose.
You can also use comments to separate blocks of code within a handler, like this
on MyHandler global gMyGlobal -- SomeStatementHere AnotherStatement -- StatmentForSomethingElse MoreOfThat end MyHandler
By breaking code into blocks with comments you can make groups of related statements stand out from the rest of the code. And by using blank lines only between handlers but not within then, you can skim through the code more easily.
You'll want to use comments liberally if you're not concerned about script size. SuperCard, HyperCard, and a few other scripting environments have a 32k limit on the size of a script, so if you run into this limit you make want to use only important descriptive ones and ditch those used as visual separators to save space.
Comments as Breadcrumbs: "What?" Bang!" "Hmmm"
In a perfect world all the code we write would be written with complete confidence that it's perfect the first time out. But the real world is much muckier, and often we code in ways that we know aren't optimal, but we have to get something working now so we sacrifice quality or completeness for time. Merde happens, even on the most well managed projects.
When we write suboptimal code it can be useful to leave some sort of marker identifying it as something we'll want to revisit later on. We use a convention we call "What? Bang! Hmmm".
The goal of "What? Bang! Hmmm" is to leave specific types of comment tags which can be searched for later to revisit portions of scripts which may need revision, sort of like leaving breadcrumbs on your trail through the code. The convention we use is designed to be simple to remember, and simple to type:
When you write your handlers, try to make them as generalized as possible. If you can avoid using global variables do so, since anything dependent on other routines will reduce the chances of being able to use this routine again in the future. If you only need to read a value in a global, consider passing it into the handler as an argument instead.
Another way to keep code portable is to remove references to specific objects. If you have a function which calculates the sum of two fields for example, a non-portable version might look like this:
put SumFields() function SumFields return fld "Num1" + fld "Num2" + fld "Num3" end SumFields
This handler only works if you have three fields using those specific names. You could make it more portable like this:
put SumFields(fld "Num1", fld "Num2", fld "Num3") function SumFields put 0 into tSum repeat with i = 1 to paramcount() add param(i) to tSum end repeat return tSum end SumFields
Not only can this handler be used anywhere, by using the param and paramCount functions we can now return the sum of any number of containers, making it applicable for a great many more uses.
When referring to objects, use names whenever possible. If the name cannot be known or may change, use the object's ID number. Try to avoid using the ordinal number of the object (e.g., "button 4"), as you may make changes to the layout which will cause the script to break. Descriptive names also help you understand the purpose of the object.
Name all objects (cd flds, grcs, wds, &c.). Stay away from using numbers unless your script depends upon that technique (and even so, you should still name them. I often use names like "Bookmark 1," "Bookmark 2", &c.).
Code for Speed
Because interpreted languages are generally slowed than compiled ones, speed is a more critical consideration. Here's a few tips for helping the interpeter do its job faster:
Many scripting languages tout their "English-like" syntax as one of their strongest benefits, and of course it is. But code serves a different purpose than narrative writing, so while readability is important we find that in production environments code is far more frequently skimmed than read. We don't often read an enitre script to grasp it's story; if we wrote the code we already know the general story it tells, and if we inherited the code base some someone else we'll likely read it only once. But in the course of debugging and enhancing the code base we'll frequently skim for certain elements, specific tokens or patterns, and using naming conventions can help make the elements you're most likely to be looking for stand out visually from the rest of the code.
In this section we advocate liberal use of what is commonly referred to as Hungarian notation, in honor of the famous Microsoft programmer Charles Simonyi, who is said to follow this practice obsessively. The value of Hungarian notation is that it provides a consistent method for determining the use and nature of a given container by its name alone, without having to look elsewhere in the code to find out where it came from, such as whether it's global or was passed in as an argument. To varying degrees, this style has been adopted by many scripters in recent years, as is reflected in the product documentation for Revolution, SuperCard, and others.
While it can be argued that naming conventions like "Hungarian-lite" make code less readable, it can make the code more skimmable, which is a far more frequent need over the life cycle of a code base.
Variables and Arguments
You can quickly identify whether a variable is local or global, or whether it was passed in as an argument, if you preceed the descriptive name with a lower-case letter to determine its type.
* Runtime Revolution, MetaCard, and Visual Basic only.
Associative arrays can be local, script-local, global, or passed as a parameter, so to preserve the type designation arrays are noted with a trailing "A", often in plural form to reflect its status as a collection, e.g.:
gOpenWindowsA tPasswordsA pSelectedObjectsA
In this example we use a number of these tips:
Handlers and Functions
When used in libraries, most of the handlers will be for use by other components. But you may have some handlers which are used only by other routines within the library, not intended for use by others. It can be helpful to distinguish these "private" handlers, so using a leading underscore ("_") character can help identify these readily:
Yes, we know this breaks the rule in the fourth bullet point above, but for a reason: because it's harder to copy such handler names, it draws extra attention to their more specialized use.
Object names are usually best left to the specific needs of the program you're working on, but there are some guidelines which may be useful in some cirumstances: