This is part eighteen 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.
Last time, we went over all of the steps in bash command processing. This time, let’s talk about one of the steps in particular, word splitting of expansions.
Breaking Up Is Hard To Do
Word splitting is a feature of the shell which applies after certain expansions. Such expansions include variable expansion, command substitution and process substitution (and technically arithmetic expansion as well, although it’s unclear to me how that could ever result in multiple words).
Any of these expansions can result in text with whitespace. Bash decides that this text should not be taken together as one entity, but rather should be split into separate words before invoking the resulting command.
For example, consider the assignment myfiles=”file1.txt file2.txt”. The following command copies both files into another directory:
After expanding myfiles into a string, bash takes it upon itself to split the resulting string into the two words based on the whitespace between them, resulting in file1.txt and file2.txt.
If it hadn’t done so, then the cp command would try to find a single file with the name “file1.txt file2.txt” and copy it instead.
That’s pretty much the idea behind word splitting. In some cases, it can make it easy to store multiple arguments to a command in a single string since they’ll be split back into individual arguments before the command is invoked. It can also be used to iterate in a loop such as this one, which accomplishes the same thing as the command above:
Here myfiles is working like an array would in other languages, allowing file to iterate through its values.
Really, though, word splitting is only of use if you choose to rely on it, and there’s no need to do so. Bash has had arrays for quite some time now, and they not only supersede whitespace-based lists, they are better at what they do. There’s almost no occasion to use a poor-man’s array.
Here’s how the same loop would work with an actual array. This one works with spaces in filenames:
It may be tempting to write the earlier code since it looks a bit cleaner and is easier to write, since you aren’t thinking about quotation marks, brackets or @ symbols.
Don’t be fooled, though. First of all, it simply doesn’t work for valid inputs with spaces. That’s not an edge case.
Second, the quotation marks are because of the automatic word splitting. Turn it off and the need for them goes away. In fact, all of our code will look better.
If you’ve been following my code so far, you’ll have noticed fairly liberal usage of quotation marks already. In fact, I’ve used them on every expansion (with a couple exceptions for statements which don’t do word splitting).
The reason for all the quotes is word splitting. There’s a big problem with it. It simply shouldn’t be done automatically.
You see, wherever there might be a space in an variable expansion, you have to disable word splitting manually, or else you won’t get be getting the value stored in the variable. You’ll be getting two or three or more things instead.
That’s what the double-quotes do, they disable word splitting on the expansion result.
Take pathnames with spaces as an example again. If you want to store such values in a variable, you can’t use that expansion without disabling word splitting with double quotes. In fact, it’s unpredictable enough that the standard practice is to double-quote every expansion, all the time. It’s a waste of time and a ton of visual noise in your code.
Most people who write scripts either don’t know or don’t care, preferring not to litter their code with quotes. That’s understandable. But it introduces a class of bugs which bite you in the edge case and are pervasive throughout your code. All because of a feature you’re probably not even using anyway.
So what to do? Turn it off, and if you happen to need it, toggle it on and back off again. Trust me, you won’t need it much.
Continue with part 19 - disabling word splitting