Where Bash Fits for Me

 

My friend Mirek wrote an interesting post recently: Where Rust fits for me. In it, he made a hierarchy of the programming languages he reaches for on a regular basis and why he picks a particular one for any specific task. I'd summarise it crudely like this:

• Shell: only for very basic setup. Anything with non-trivial logic belongs somewhere else.
• Python: his go-to but with caveats about the expected lifetime of the code: if it needs to persist without regular attention then Python won't do.
• Rust: Whatever doesn't fit in the other two categories. 

I enjoy seeing people be thoughtful about their tools and reflect on the way that they work. Understanding what you do and why is a helpful first step on improving what you do and why, should you want to.

And I always want to.

So I thought I'd attempt the same kind of analysis and started listing languages that I use regularly:

Groovy: I use Groovy exclusively in Jenkins pipelines. It's useful to have a little familiarity with it but I would not choose to start any project in it. I have written a how-to guide for debugging pipelines because the whole ecosystem is so messy and unhelpful. 

Java and Kotlin: most of the codebases I work on are Kotlin, Java, or both. I don't feel fluent in either of these languages but I have written code from scratch in them and would reach for them again if I was writing tooling that I expect to be maintained by a team that's already using them.

Python: When I need a reasonable amount of logic, if I'm prototyping, if I think what I'm building will be collaborative, then Python is a good choice. I don't like or understand deeply the complete mess that is Python package management and the numerous different packages that people use to manage their package mangement. I do like the immediacy of an interpreter and the freedom to iterate quickly with a scripting language.

There are other languages that I need to use in specific product which I don't know how to write from scratch. I edit existing code in those languages but they aren't worth including here.

Side note: up to this point I felt like I was echoing Mirek's post as I'd intended. That changed when I started trying to write a couple of sentences about how I use the shell ... and found that I have a lot to say. This shouldn't be a surprise because I have written about the shell many times on this blog and I've lost count of how many colleagues I've introduced to its power and charms over the years. Yet still, surprise it was, and I couldn't trim my thoughts down, so I just went with it.

Shell and Linux utilities: I do a lot of exploratory testing of back-end services which means I am often inspecting HTTP traffic, payloads, log files, and telemetry data. There are a handful of shell commands and utilities (mostly Bash, but also Zsh) that I use frequently to help me find a way through, and the beauty of them is that they can be plumbed together in natural ways to answer the specific question I have at any given moment.  

I can compose those answers at the command line, interactively and iteratively but, once I have something that I think I want to re-use I can drop it into a script and suddenly it's a tool in its own right. There's no reimplementation, the script is simply the command or sequence of commands I already worked out.

Once it's in a script I can parameterise or loop or whatever to extend its scope, if I need to. The script is also documentation of what I did that can be attached to test notes or bug reports, or shared with colleagues, or serve as a starting point for another round of work with slightly different questions.

Occasionally I'll create something with general applicability and add it to my personal repo or a project repo. This can sometimes come about by accident, but it's more likely that I'll find I've been asking the same kind of question repeatedly and don't want to solve it from first principles any more.

An example of that from recent work would be a couple of scripts to dump a Mongo database and to diff two dumps. The first script (dump_db) will reach to a nominated remote environment, or a local Mongo, and export a database as a set of BSON files. It then converts these to JSON and uses jq to pretty print them to new files. The second script (diff_db) compares pretty-printed files across the two dumps and produces a list of the differences. 

This is extremely valuable when we change something about how an application persists its data, or are concerned that we might have side-effected that unintentionally. I can run one version of the application and dump Mongo, then do the same actions on another, and dump Mongo again. The diff gives us some idea whether the kind of changes we expect (or no change) occurred.

If I wasn't in a largely back-end world, would I still use command-line tools like these? For sure. I use shell for all sorts of tasks, such as searching codebases or applying the same configuration change across multiple repositories, or port-forwarding to some remote environment so that I can test against an application deployed there. 

If I wasn't in a Linux environment would I still use command-line tools like these? I think I probably would. On Windows I might go for Powershell or WSL to provide the interactivity and scope for experimentation with fast feedback that I generally want, at least initially. 

If I was starting from scratch in testing would I still use command-line tools like these?  I really hope so. Much of the way I work comes from seeing how others work, being curious about how I could work more efficiently, and just trying things. I'd like to think that I'd be open to learning what are, I agree, bizarre, unfamiliar, and sometimes unfathomable commands, options, and escaping policies in order to get the benefits.

So where does Bash fit for me? It fits in the places where I need it and where I don't know a better alternative. I'll continue to use it in those places until it doesn't.

--00-- 

I showed Mirek this post before I published it and asked for his comments, which I've edited down here:

Fully agreed about Groovy, and with a sentiment that working with Jenkins ecosystem is painful once things start to go off the rails.

Also fully agreed about the mess that is management of Python package management. Unfortunately that's one of the things I decided to understand deeper and I emerged with strong opinions. If I were starting today, one of the better tools is uv.  It's actually harder for me as I'm thinking about all the cruft I have accrued over the years, but ...

The properties of the shell that you mention are undeniably strong side of shell. Composability of commands and how much you can do with very basic tools. That you can write a command and then just put it in a file to turn into a script, and extend it from there. That it's easy to share these scripts, or store them in personal library.

I do use shell all the time, but usually I would choose specific tools. For JSON I would use fx for interactive usage, or I would write quick Python script to parse it and extract whatever I need (or use Python shell to do that). Honestly, I prefer that to learning arcane jq syntax.

I guess I didn't give shell justice in my blog post, but that post was really mostly about Rust. At the same time, I did create crazy things with shell in the past, and honestly, I would rather discourage that - just jump to proper programming language sooner. 

It's a tangent thought, but this conversation reminds me of when I was taking my first steps in programming. I saw all these robust, clean programs and I strove to write similar code. It took me some time and seeing work of some people to realize that sometimes it's fine to cut corners.

--00--

This is great. The tools we choose are dependent on multiple factors such as the tools we know, and know how to use well enough, which tools we think will fit the problem in front of us, whether we've recognised the problem correctly, whether we have time to find out, whether we have the appetite to look for a better tool, and so on. (I went deep into this in Take Your Pick.)

I come to my context with my background, Mirek to his with his, and you to yours with yours. So it's no surprise that we make different choices about which tool to start with, and why, and when to change to something else.

Staying alert to the rabbit hole, being open to try another way, and actively looking for new approaches that could be valuable feels like the obvious thing to do, to me. I believe I do that: between writing the first section of this post and writing this postscript, I've updated the Mongo dumper and differ to use a new tool, jd, that gives more helpful output than the earlier version, and removes some friction that had become annoying. 

If I needed to do more for that task I'd probably switch to Python because I feel like the quick value has been obtained from bash and I understand the task and data well enough to write a specialised tool. I just don't think it's worth investing in that right now.

Which brings me back to the question I posed earlier: where does Bash fit for me? And I think I have the same answer: it fits in the places where I need it and where I don't know a better alternative. I'll continue to use it in those places until it doesn't.
Image: Wikimedia