Names should be as short as possible while still being clear
benhoyt.com
benhoyt.com
compare acquire and take_ownership
If you put 10 software engineers in a room, you'll get 10 different answers to any question. Every company you work for will be filled with endless "smart guys", ready to spring in to action and nitpick you to death about arbitrary things.
It's exhausting.
We need more languages that come with rigid rules for things like style, so we can stop wasting time and get on with solving problems.
Maybe a more accurate phrasing would be, "Can't wait to retire and do literally anything else with my time, including writing programs with which I answer to no one about coding style.", Hah.
You probably know this but for discussion's sake: in many modern forges, your pull request can't be merged unless the comments have been resolved/closed. So you basically cannot ignore these "feedback" items; you have to engage the comment whether that means actioning it or refuting it.
I currently work with the worst pedant I've encountered in my career to date, and I can empathise with wanting to leave the industry.
I have one of these too, one of my PRs received a comment on the third round of review about correcting a minor typo in a comment that was made years ago by someone else.
It is still just a tool and does require the team to buy-in and utilize it, but we've found it helpful.
Regardless, I think that's just a function of your team and your dynamics that you've built up.
On the current team I'm on we for example don't require this (at least on GitHub its configurable), there is an understanding of nits, how to declare them and what that means, as well as an understanding of using request changes vs approve and so on. We've built this up by having discussions about what the purpose of reviews are (eg making sure your code is understandable by someone other than you), having a common understanding that perfect should never get in the way of good, etc.
We regularly have retrospectives where, if we were having issues with really bad pedantry, I would hope it could be brought up.
It's.. quite different where I am now. Change requests for untouched/irrelevant code, tone problems, etc. I'd be inclined to try and 'raise the tide' but I'm just trying to leave instead for many other reasons.
So the "actioning" part becomes their work.
Attitude is, if I can argue with someone and have it in a fun productive way - even if we don’t tackle most important thing we should work on at the moment - it still might be valuable.
If I have to deal with smartass that nitpicks on variable names, he wastes my energy and any will to do further work for the day or if it is more heated and goes over some “Astro architecture” might even burn my willpower to do any work till the end of the week.
Consider for example the bikeshedding society did on whether to distance 1.5m or 2m during COVID, and how precisely and how long to wash their hands. We had a ton of those. When people feel out of control they want to own a decision that makes them feel like they're back in control.
Programming is complex at scale. But bikeshedding about how many characters our lines, how many lines our functions, and how to name things is always there for us to argue about.
Oh and whether a class is truly single responsibility, that's an evergreen.
"The term was coined as a metaphor to illuminate Parkinson's Law of Triviality. Parkinson observed that a committee whose job is to approve plans for a nuclear power plant may spend the majority of its time on relatively unimportant but easy-to-grasp issues, such as what materials to use for the staff bikeshed, while neglecting the design of the power plant itself, which is far more important but also far more difficult to criticize constructively."
Source: https://en.wiktionary.org/wiki/bikeshedding
A form of "one-man-bikeshedding" is when you keep postponing important work, by constantly checking email, cleaning the house, or doing other unimportant busywork. You are not ignorant about what is important. But you procrastinate as it's much more difficult and painful to face it.
The same dynamic is present also in groups, where groups naturally tend towards endlessly debating nonsense, as it's easy-to-do busywork.
Again the point that I'm trying to make is that someone making a big deal out of standards is generally not a result of other things being "difficult to figure out, discuss, or acknowledge" but instead because some people genuinely care a good bit about sticking to standards.
Yep, it gets tedious. It really grates on me when reviewers are willing to die on a hill over some minutia while the codebase is in tatters. Arguing over extremely minor duplication while there exists two, active, physical copies of an application in separate repos (a real example) is the most pennywise poundfoolish waste of time and it drives me crazy.
So I did that! And at the same time cleaned up the code, fixed bugs, and adding required new functionality, and all while merging in constant change from the main codebase.
After reducing a 1M line codebase to 280k lines, when almost done, after that 8 months, I was asked "how much longer".
My response of "a month for testing" was apparently too long for the new manager, who turfed all those fixes, including the new functionality.
And I wasn't even behind schedule, I was ahead.
Once abandoned, everyone just copy and pasted my changes... in triplicate, into the old codebase.
Which took 3 months, instead of 1.
Bah. A year of my life wasted.
To be fair, it's possible bureaucratic process got in the way. If their commit / deployment process didn't allow for any changes to hit the production branch until it was "finished" then there wasn't really an opportunity for them to increment.
That seems likely considering "the new manager, who turfed all those fixes, including the new functionality" suggests other organizational problems. If one month is too long for the new manager, the new manager's goal seems to be to "do things" rather than to "solve problems".
Others are discussing incremental changes, that doesn't work well when underlying frameworks have significant breaking changes.
No everyone, I didn't choose the framework. No, I couldn't change it. No, I didn't choose the path.
While I know people are trying to help, suggestions starting with "you should have" aren't very helpful here.
But they might be helpful to others.
There are ways to do these large refactorings that allow for incremental changes, never checking in a commit that breaks the build and where tests are passing at each stage. Your experience doesn't sound great and I'm sorry that happened to you, but I just wanted to add this to say that it's not necessarily personal.
The responses to far, to my subjective reading, imply that the OP did something stupid and that's why they were in that situation. Again, it's just a personal interpretation, but I don't think it's a far-fetched one, and I don't think we have the context to make that assumption.
I agree that it can be useful to highlight the possible pitfalls to readers, but it's my opinion that it could be done in a more empathetic and tactful way that highlights our collective lack of context while still providing general advice to readers. I also think it would tend to lead to a more nuanced and useful conversation for the same readers.
Not excusing or defending it, just trying to serve up a reason why it happens.
Entirely this. I have a preferred style when I'm writing my own code. When I'm writing code for someone else, though, I will roll with whatever the house style is[1].
I just want consistency and predictability. Once I know the naming scheme, I should be able to take a really good guess as to what any given function or variable name is by knowing what they do, and vice versa.
[1] I consider it a red flag if a company doesn't have an established house style. It matters not what the style consists of, only that it exists and is mostly adhered to.
The source code has two purposes - to be processed (compiled/interpreted) by software, and to be read by humans. While it is relatively clear what is considered good code for the first purpose, it is harder to say what it is for the second one.
An essay could be written in an elegant style or be full of clumsy wordings and hard to understand sentences, despite having the same meaning. The same is true for source code.
Naming matters more for interface types and APIs than for local variables.
"acquire" and "take_ownership" might be equally good names for an operation, but you should pick one and stick to it throughout the codebase. And it would make sense to poll other people about what they think is a good name. Especially if you don't really care one way or another. Other people may have valid perspectives or they may just be better at expressing meaning through names.
You can't really nail everything down in the official style guide for a language without it becoming baroque and oppressive. If you want to avoid bikeshedding, have a process in place for choosing naming conventions. Name someone as the arbiter. If consensus cannot be reached within a reasonable amount of time, the arbiter decides. It might help to keep a record of naming choices so people understand the reasoning behind why certain practices are chosen.
Most programming is a social activity in that multiple people will have to deal with the code. Not caring about something as essential as naming, and wanting to avoid effort because you don't care, can be seen as a kind of anti-social behavior.
On well functioning teams I've worked on, naming tends to not be a big issue because over time a team develops a common vocabulary. When in doubt, ask someone else on the team how they interpret a name.
I have actually been in the throes of battle with co-workers over relaxing our variable naming standards because they are too damn prescriptive, which (in my opinion) reduces code clarity, while also setting up conflicts between those who think code should be as clear as possible in the file versus those who think that we should follow the standard even though we know it to be flawed and produce overly-verbose code.
My core argument is very much "who cares as long as it's clear" but having a specific piece of ammo where I can say "here are several arguments specifically against our current style" may convince some of these folks to loosen up a little bit. The bike-shedding is indeed ridiculous; going back and forth in Swarm for days over names for code that has fundamental algorithmic flaws. Awful stuff.
Again with the hypothetical dichotomies!
take_ownership -> acquire (fine at first)
give_ownership -> unacquire (not even a word)
split_ownership -> halfacquire (complete nonsense)
then there needs to be a way to split ownership other than an equal 2-way split and you're fucked.
I haven't seen this exact situation, but I've seen stuff like this. And sometimes, I was the programmer.
Names should be as short as possible while still being meaningful AND keeping in mind future work requirements.
On a tangent, I'm always frustrated when as a developer, I don't learn much about the actual business. There's always a situation down the road where if you had known what the company was planning on doing in the next quarter, or couple years, you could have saved everyone some work by doing things differently.
This reminds me of when some code was lightly edited to remove references to "master" and "slave", only to end up with a function "enminion".
Ever dealt with a rotating code review that passes through multiple hands? It's funny watching the principal architect and technical fellow ask me to refactor my code again to remove the changes that the other one asked me to make!
I spent a month in one such code review.
No wonder management hates listening to engineers. If we have a good point to make, they ignore it because of the last 3 engineers who wasted time with stylistic, pedantic arguments.
I agree software need to be basic error free, compilable, buildable and deployable before naming and structuring aesthetics becomes important.
> Who cares!? Those are both great names for a function...
Now the author has written many good quality working software systems, it is natural that they want to talk about next level things.
The actual names that came out of these exercises are generally pretty good, and often a shorter name would win over a longer one.
I have to come up with many names for things in code every day. I can't imagine the slowdown having to workshop every name would incur...
An example would be a set of enum names that I need to share with my teammates, where it might be a lot of hassle to replace them all if we don't decide on proper names now, but any set of names would work.
This is where I'm at now.
Decades ago, I was an "only spaces, no tabs" guy. Still am to an extent. But what made me more open was realizing it wasn't spaces or tabs specifically that mattered, only that code is readable, and consistently formatted code is more readable. Seeing things like `gofmt`, which mix tabs and spaces, and comes with the language? Your debate about spaces versus tabs is over, now we can get back to tackling real problems, and your code will be rejected at CI if it isn't run through the standard formatter. Same thing with `black` for Python.
Above all else, be consistent. I can understand hate for "standard" language libraries that have functions with wildly different naming schemes or input and output arguments that aren't consistent - those should DIAF (looking at you, PHP "standard" library). But as to the names themselves? Doesn't particularly matter much. Unless you're the Bruce Tognazzini of API design, you probably don't have anything of substance to back up why it should be "make_" versus "create_".
Aren't you still doing the same thing though? You have your arbitrary rules and they're absolutely truths... until you change your mind about them. Except before the rule was "only space no tabs" and now it's "be consistent"?
But yes, "A foolish consistency is the hobgoblin of little minds." The question is, what is "foolish"? I'm arguing against things like `strpos` and `str_rot13`. I'm arguing against having similar functions where argument order is inconsistent. These are the sorts of things that slow developers down and force them to check the documentation every time just to make sure they aren't introducing bugs by default.
I'm sure you also felt in good company about tabs or not tabs or whatever it was you liked.
Then explain how consistency is a bad thing.
Nothing against this.
> Then explain how consistency is a bad thing.
This is a really boring conversation, because if I give a good example you'll just define consistency for that example not to apply. For example, picking the token case of naming functions in such and such way, it occurs to me that I don't really care if you break rules are long it's done in a way that doesn't affect the interface. If something can be considered an implementation detail, I don't care that it doesn't follow the same rules, as long as the interface doesn't allow you to see it. But then you'll just say something like "oh but 'consistency' doesn't mean a rule applies under any and all circumstances, there's always caveats" bla bla bla.
It's a boring conversation, my point is that you had a dogma, and now you have a different dogma. You think you're evolving, I think you're changing but not evolving in a bigger way.
Tabs vs spaces is a great example, since tabs are just better ( :) ). The attempts to control how others view the code are frankly bizarre. Why not go all the way and control the typeface, font size, and syntax highlighting by sharing code via screenshots?
Let everyone choose their preferred indentation size, it's not only polite, but also an accessibility concern.
That said, in this particular case, I think Russ could have chosen a better example. Neither "acquire" nor "take_ownership" are terrible. But when you're a language designer (like Russ), the names you choose shape people's code and their thinking for the next 10, 20, even 50 years. When you're a library designer (many of us are), it might only be a few years, but still.
I think as code reviewers we should take this as seriously as the context demands: for internal names and local variables, it's not too important. For public API names, very important. For language changes, extremely important.
Some good quotes from Bob Nystrom, author of "Crafting Interpreters":
"Names are the structure we impose on the formless sea of bits that is computing." http://journal.stuffwithstuff.com/2016/06/16/long-names-are-...
"For me, thinking about names is inseparable from the process of design. To name something is to define it." http://journal.stuffwithstuff.com/2009/06/05/naming-things-i...
I understand that it's important to think critically about the philosophical North stars of your work. This is precisely what it means to be an "expert".
I appreciate the work you're doing adding your voice to the discussion.
That being said, it doesn't change the fact that, for me, it's tiring working in a profession where nothing is "real". There is never going to be an end to these kinds of discussions, because everything is abstract, relative, and ever-changing.
Unlike many other professions grounded in the limitations of physical reality: carpentry, civil engineering, hair cutting, etc, Software Engineering is a house of cards of artificial abstractions. The tools and architectural patterns are mostly free from physical constraint, so they can change at will, and they do.
What is "good" is highly dependent on context, and in our case that context is fluid, changing along every dimension all the time, making it very hard to establish a shared understanding of any given problem.
This translates into a work culture where there is always room for debate about what to do, since there is almost never an objective best next step. And debate we do.
Interesting comment, thanks. It certainly puts your original comment in perspective, and I respect that position. I suppose this is related to the position that "computer science is more art than science".
Serious devs care.
I was once at a software conference where Bjarne Stroustrup was givnig some presentations, and in the after-session questions, I thought I'd ask about something that had been bugging me about software design in general — I'd found myself spending what I thought might be excessive time to designing naming systems for the projects I worked on, and I wondered what he thought about spending time on naming conventions.
With no hesitation, he said "Not at all, naming is very important", and very useful to get right up front, even for solo projects, as good clear naming helps speed the entire building and maintenance.
So, yes, both "acquire" and "take_ownership" are relatively synonymous, but one will clearly be better and more consistent with the other in the context of your naming system. And this will matter to at least some serious devs.
(But YMMV; if you have tried it a couple times and found it doesn't help you a bit, then you probably shouldn't spend more time on it)
I have the complete opposite opinion. Most software engineers are obsessed with style, and I don't think accommodating it is the right solution. I think the right idea is just not enforcing any stylistic rules at all (while obviously keeping the readability and maintainability-related rules) and just let everyone write code how they want, while forcing them to tolerate other people's coding, naming, brace placement, etc. styles.
For instance. I've never been able to understand why something like `systemctl` could be so clumsily named. If you have to do sysadmin work, sometimes you will be using the command a lot. If we ignore the fact that I see more and more people typing out entire commands without the use of completion, `systemctl` lacks a short, unique prefix and ends up being quite awkward to type.
On my system `sys` is the prefix for 26 commands, and `system` is the prefix for 24 commands. So I end up having to type 7 keystrokes to type systemctl on my system with completion. Sure, I could alias it, but then I'd have to replicate it everywhere, plus I'd be forming habits that just cause more frustration later.
Another way to fail is to choose a prefix that has bad typing flow. For instance I wrote a program that has the prefix "aq". Try to type it. Awkward, isn't it?
(If you don't think any of this matters, perhaps you shouldn't be the one to pick names for programs that are run on the command line).
But the most productivity-sapping mistake is names that are too vague.
The idea that "names should be as short as possible while still being clear" is a truism I don't think anyone disagrees with. But it's a balance that nobody ever perfectly gets right.
So if you're going to err, it's better to err on the side of clarity rather than shortness. E.g. lines like:
fetcher = repolib.IndexFetcher(args.arch)
counts = collections.Counter()
for file in repo.files:
repo_fetcher = repolib.IndexFetcher(args.arch)
file_counts = collections.Counter()
for repo_file in repo.files:
index = RepositoryIndex(args.architecture)
statistics = RepositoryStatistics()
for repository in index.retrieve_repositories():
for repository_file in repository.files:
statistics.increment_file_count(repository_file.name)
It can be made even less verbose but at this point there's no real need.
Naming things should be short as possible while still being clear, unless making it longer helps something else.
Name things short as possible while still clear, unless longer helps something else.
Name short as possible while clear, unless longer is better.
Name short and clear, unless longer better.
If longer not better, name short & clear.
if_longer_not_better_name_short_clear
ifLongerNotBetterNameShortClear
What was “clear” two years ago may be unclear today because we have four times as many functions. Via the pigeonhole principle, there are no short names that are sufficiently clear, and as X goes to infinity even the medium ones become problematic.
Use a modern IDE with autocomplete. Try to make the beginning of your names as unique as possible. Few people will have to type in the suffixes.
Used to be we would ask other developers what book they would buy all of their coworkers. Answers have changed over time, anything from Brooks to Gamma. But for my money, I would buy them a thesaurus.
If you use the same word in two places in the code, they should be sharing code and data structures. If they don’t, you should use a synonym. Never recycle names, and be sure you aren’t recycling. Many people are wrong here, claiming two things are the same but with no code shared, and in many cases I’ve seen, using two different definitions of the same word.
By default, developers assume that a requested change is mandatory. So as a reviewer, it's up to you to explain the difference between a change you're suggesting because it's more concise to use a list comprehension instead of a loop vs a change you're requiring because the change list introduces a race condition that will cause customers to lose money.
If you do that, you'll make everyone you work with happier
The coder in that case knows they can ignore that without consequence if their preference or perspective is different. The reviewer also knows that and doesn't feel bad if that advice is ignored.
It's been helpful for me because I'm a picky reviewer and feel compelled to mention things that my years of experience tell me are "wrong." But I've also come to realize it's often not wrong, just not as good for potentially non-obvious reasons. But my devs will never learn the potentially "better" way if I don't point it out. So "nit" gives me a way to communicate and teach without the pressure of the dev needing to make an immediate change.
So I’d say amount of time invested in naming won’t payback.
Q: Why can't programmers tell Christmas for Halloween?
A: Because OCT 31 = DEC 25.
I feel like the author missed the obvious; a name should reduce cognitive load as much as possible. Length is one factor that is relevant to this goal, but not the only factor.
When a name is too long, this is usually a design issue, not a stylistic issue, i.e. writing Swiss Army functions.
Also, takeOwnership is not inferior to acquire, it's just about a wash.
- In Mathematica, if you want to plot a vector field, you use VectorPlot[]
- In MATLAB, you use... quiver(). You know, because it holds a lot of arrows. Short and clear, eh?
Very unclear to anyone who is new to that functionality, though.
As contrasted with "VectorPlot": was that "VectorPlot", "VectorGraph", "TensorPlot", or "PlotVector"?
"quiver" is short and sweet and slightly unusual but also conceptually related in a sense to the function.
I'm not saying that "quiver" is the better name, though I'm offering some arguments as to why it might be.
"Shaka, when the walls fell."
Sure it's poetic, but WTF do you mean?
Now we see that was a bad idea - longer names just allow for more wasted words and are actually less descriptive. Some kind of social or algorithmic convention makes them quicker to parse and understand.
I suppose some compromise will be proposed one day. Maybe a prefix to classify the method name, then some short description. Or even an IDE convention to pop up a short description for method names to allow clarification in-situ.
The author's real problem is not about an excess of characters, it's that names are often long because they include redundant or irrelevant information. Still an incredibly minor issue, but there's some legitimacy to that. Better advice would be to keep names informative.
fetcher = repolib.IndexFetcher(args.arch)
counts = collections.Counter()
for repo in fetcher.get_repos():
for file in repo.files:
counts[file.name] += 1
for repo, count in counts.most_common(10):
print(repo, count)
His basic premise is correct: Names should be as short as possible while still being clear. Unfortunately he didn't follow his own premise. Err on the side of too long but clear, because maintenance is 90-99% of the project's life cycle and reading code is at least 5x harder than writing it. You should name variables/methods as if you won't be touching it for 6 months and you want to be able to have a decent understanding of what it does within 5-10 seconds of reading it.
>That last sentence is poetry.
It’s a borrowing from Strunk & White, “make every word tell.”
Rust's flavor of variable shadowing is a great way to support developers to keep names short.
I agree 100% with this. Overly long names are very painful to work with.
But naming things is very hard. I'd be happier if people just adopted a consistent naming scheme that is predictable. I'd put up with the crazy long names in exchange for that.
After that I mostly agree with the author that short and clear are the next two priorities.
They’ve made it so that naming cannot be discussed. If you see a horrendous name and suggest an improvement, there’s no overt response, but later you hear from your friend that the poor namer is accusing you of wanting to die on a hill.
Dude. It was just a suggestion. And it was labeled as a nit. You don’t have to take it, even though we’d be better off if you did.
So I’ve all but abandoned ever suggesting improved names, unfortunately. Being a good team player is more important. I name my stuff well, but what others do is up to them.
Whether something is too short depends on your context. requests.get might be ambiguous for someone who has never seen the code base before, but will quickly get obviously with a little exposure, so does the code base have dedicated maintainers?
The skill of good naming isn't distributed evenly and OP's names are pretty good, so I'd be happy for them to come along to my code base and rename things as long as it was within other constraints (not consuming too much time, doesn't break api, etc).
I think the point includes reducing unnecessarily complex terminology as well. Do we have to “execute” a function or can we just simply “run” it?
Who said naming was difficult? Obviously all you need to do is make names as short as possible, but no shorter!
Simple!
A bit extreme, but the point is that all of this is just a waste of time. A WASTE OF TIME. Name it close to what you think it is and move on with your life.
This isn't wrong.
Wrong is obsessing over something so stupid.