Shortly after my recent post of a realpath implementation in bash my friend David Kendal suggested a C implementation would be preferable. I have to admit, it’s far simpler code, it wraps around the actual realpath() so no work to make it functionally equivalent, and it’ll be immensely faster (although I haven’t benchmarked it). See his example implementation. I really can’t agree more.

That said, I do see use for the bash implementation.

Why bash?

Most of my bash scripts are are exercises to keep my bash-fu honed, so this one certainly started that way and did so as a part of tools-osx. While not a modern, object-oriented language, I find bash to be there for me on nearly every platform I use (excluding the obvious like pre-Mac OS X, Newton OS, etc.) without needing to build or install anything else. There are feature differences between versions, but they are minor.

It’s very much in the spirit of UNIX and forces you to use the other command-line tools at your disposal, in conjunction with each other. So, I find it a great way to master more command-line tools and also to become a better command-line user. The more I learn in bash the more I’m able to do efficiently on the command-line, esp. complex tasks. As a server admin, this is immensely helpful. And, the server admin in me is probably why I find it greatly comforting to not have to install or maintain another language and its tools.

It’s a relatively minimalist language and lacks a lot of niceties, but I consider it a bit of a survivalist language at this point. It’s no Perl, Python, or Ruby and it may be frustrating to work with arrays or return anything but integers, but if you can master it you’ll be able to be tossed into any random server and come out alive.

So, Why realpath in bash?

The real benefit for a realpath implementation in bash is the portability without the need to compile for new platforms. Simply copy to the appropriate directory and you’re done. This is ideal for situations where you cannot guarantee a compiler will be available (the use-case for tools-osx) or cannot have a compiler installed (one I commonly encounter in my admin work due to various security requirements, but also one people with shared hosts often run into).

I was also recently contacted by someone who is using my realpath implementation in part of the process for launching a Java app on Linux and Mac OS X, so in this case it’s ideal to keep everything extremely portable with only one codebase and no build process required. Exactly the kind of thing I had in mind.

In Conclusion

So, the bash implementation of realpath has a couple of use-cases, mostly if you’re shooting for no compile process for installation. If you’ve already got a compile process, you’ll probably be better off with the C implementation for performance and to reduce your codebase. If you’re Linux-only, or platforms where you already have readlink available, just use readlink. Naturally, this is all with bash scripting in mind, most higher languages have a built-in realpath function.