This is part 22.5 of a series on how to approach bash programming in a way that’s safer and more structured than your basic script.
See part 1 if you want to catch the series from the start.
Would Smell as Sweet
My bash scripts primarily deal with three namespaces:
While it’s possible to export functions in bash, the function namespace is not inherited by default. Therefore all you have to worry about with function names is your own names and anything you source.
Aliases are their own namespace, although of course they will mask functions of the same name. Still, if you unset an alias, it will not affect the function and the function will be available once more. Aliases can’t be exported and are never inherited.
We’ll be mostly focusing on variable names for the rest of this post.
An Aside on Unsetting Variables and Functions
Before we do though, there is actually one place that the variable and function namespaces do overlap, and that’s when you’re unsetting a name.
The unset command is used to unset both variables and functions. If you unset a function and a variable with the same name, it will unset the variable first. If you do it again, only then will it unset the function.
This makes it easy unset a variable when you meant to unset a function, or to unset a function because you forgot you already unset the variable name.
For that reason, you should always use an option to unset which applies for what you’re trying to do. If you’re unsetting a variable, use unset -v. If you’re unsetting a function, use unset -f.
I’m primarily concerned with the variable namespace, which includes the environment.
I’ll start by highlighting the fact that most environment variables use all-caps for their names. We should follow the same convention with any variables we export in our programs.
But why are environment variables named in all caps? The basic idea is to reduce the potential for naming conflicts with your script’s variables, since environment variables are like squatters in your scripts variable namespace.
Shells are a special programming environment because they share some of their namespace (the environment) with other instances of themselves. In most programming environments, you start with a clean variable namespace…perhaps there are a few special builtin variables, but for the most part, the space is yours.
With bash, however, it starts with any defined environment variables. Any login user can define whatever variables they want in their environment. Many packages intended for interactive use at the command line create environment variables that no other program cares about, but every program still receives. The more environment variables that are defined, the more that are inherited, and it can turn into a big heap of random names.
If you were to reuse one of those names in your program, you would inadvertently export your value for the variable. That’s because when you modify an existing environment variable, it remains exported. Your code wouldn’t even know.
So, getting back to the answer to our question, environment variables use all caps because that is a poor-man’s version of namespacing. If all environment variables use all-caps, and all regular shell variables don’t use all-caps, then they will never conflict.
That’s what namespacing is about, making sure that variable names don’t accidentally overlap. By making sure of that, we eliminate an entire class of possible programming mistake.
Namespaces on the Cheap
Really, all it takes to make a poor man’s namespace is a convention. Using all-caps for names is one such convention. The thing is, the convention has to be must be unique enough that no one will use it accidentally.
For example, if you had a library module you were writing, a short name or id for it could serve as a prefix for your variable names. A module named “mymodule” could have a “mym_” prefix for its global variables. That’s sufficient for a cheap namespace.
Of course, you only need to worry about global variables if you’re developing code for use as third-party code. It’s a good practice for environment variables in all cases, however. Unique prefixes make for good cheap namespaces.
By the same token, postfixes work just as well as a convention for cheap namespacing.
Bash Variable Identifiers
Acceptable variable names in bash are fairly proscribed. Variable identifiers consist of alphanumeric characters, uppercase or lowercase, and the underscore character. The first character of the id cannot be a number, however, leaving underscore and alphabetical characters.
By convention, most programmers use snake_case for bash variable names since it’s easy (no pressing shift) and underscore is a valid character for identifiers. It’s also very readable and doesn’t conflict with environment variable names.
You can make this work for yourself as well. Any time you break that convention, you’re necessarily creating a new namespace.
Here are the conventions I follow in general for scripts.
If I’m writing a module for use in other scripts, I create a prefix that all globals and environment variables can use, as mentioned. The following are for when I don’t have to worry about that, but want to ensure that variables within my program don’t conflict.
For some time, I tried making all global names separate from local variable names, but sometimes it doesn’t make sense. It’s occasionally useful to override a global variable with a local one. Dynamic scoping makes it possible to mask globals with locals. In that case, they need to be in the same namespace.
Constants, however, are a special kind of global. Constants don’t change. They should never need to be overridden by a local, so they can have their own namespace. We can avoid overlap with this type of global at the least.
My method is to stick with snake_case, but to capitalize the first letter of the constant, for example: My_constant.
Locals with References
We’ll discuss indirection later, but it’s the other case where I use a convention to namespace my variables. The reason being that if you pass a variable name to a function and want to set (or read) that variable in the function, you can’t use local variables of the same name within that function.
Because you can’t know what that variable name will be, you have to namespace your local variables. I use a trailing underscore, such as my_local_ because a single underscore is minimally obtrusive and still very readable, which is important when all of your variables are using it.
Continue with part 23 - passing arguments