0:00 1. Eliminate if/else branching; early termination 2:16 2. Ambiguous definitions: ‘is’ function prefix to denote boolean return 4:05 3. Self documenting code: avoid belaboured comments 5:37 4. Consistent formatting. Use Eslint + Prettier to automate code style 7:14 5. DRY business logic. Look for opportunities to refactor. Make sure to test! 8:37 6. Fail or exit functions fast. Related to 1. 9:28 7. Avoid magic values - declare and use CONSTANTS instead. 10:41 8. Avoid violating single responsibility. Prefer to use pure functions (no side effects) 11:57 9. Overly clever code (code golf). Leads to impenetrable single liners (have to rewrite in order to debug) 14:00 10. Premature optimisations.
The isPrime example, you use Math.sqrt in the condition, this will in many languages, including js, cause the sqrt to be calculated on each pass of the loop so you should break that out to a variable and the comment can then be added to the creation of the variable, this will make it even more clear that the comment is for why we use Math.Sqrt and improve performance and reduce line complexity since the loop conditional will be a simple i
This video is an excellent example of what uncle Bob mentioned in his book about being pragmatic with your own clean practices. All your tips are in the book, and the tips you don’t judge it to be good you just don’t have to use. People flame uncle Bob with the “functions shall be short” tip, but they forget the pragmatism mentioned by Bob in the introduction of the book.
You made me realise that my JavaScript code was complete doodoo and made me redo it all again. This time I actually understood what I was trying to do. Thank you.
Just a small addition on the example of DRY; what's covered here is to replace three functions that do the same thing with one function that can do anything. If you want to enforce the passed-in value is always one of a specific set of strings, it should still be fine to use these other three functions. They can cover an abstraction, enforcing the passed "action" to be a specific value. Such as "return logAction("logged in")" in the first example function.
Yep. DRY is dangerous advice because many times you don't repeat yourself but rhyme. Code should be allowed to rhyme, if it isn't then you've created an abstraction layer that will need logic. In worst case in the abstraction layer. And the more rhyming code you try to fit in the more logic you need, thus the abstraction starts growing and growing and growing. Don't repeat yourself. But don't be afraid to rhyme.
Thank you Connor!!! I am forever dealing with "code smells" because people are not taking the time to do clean code. And I always end up cleaning up THEIR mess when I have to do the daily merge of everyone's code. 😭
Try and make it their problem, that is, try to get the teammates to use the same principles as you want to. That way you are less burdened when doing merges and they get their stuff merged with more ease and probably learn from it. Where possible / feasible, use linting and code guidelines to automate the guidance.
I have been coding since 1969 and constantly learning. With ober 13 Lang under my best (1st love is assembly on a mainframe) I always agonize over "is it readable?" If not them clean it up! I write with the intention that someone else may have to maintain this. I also never take code personally, as I have seen many coders do. Instead I will show my code to others and ask them "is this easy to understand?" If the answer is no, I need to fix it. On each function, I place "function notes block" which are a block of comments, right before the function definition. It describes why the function is used, all the variables used, along with global variables, and in the case of assembly, what registers are touched or referenced. I used to run the code through a custom print program to output all these "function note blocks" to be used along any documentation I have. It shows the functions sorted alphabetically, the module they are in along with the line number where they are defined. Comes in very handy when something goes wrong
#1 Mostly agreed, but I think the last part when there are two ways of ending function, one with return and the other one with no return is more confusing than just one else statement #4 There was a semicolon missing in one line ;)
Thanks for video! Principle #6: I think you have forgotten to mention about direct returning the value instead of placing it to const and then returning it Principle #7 I think better naming for the second variable would be TAXABLE_TRANSACTION_MULTIPLIER
Great clear video 👍. But to be honest, the majority of the examples are covered by using a good ESLint configuration. Add some Prettier on top for the spacing / formatting.
Over commenting. I’m in university and I have to comment everything line for line and have a massive comment block on the top explaining how it works overall. Why is this taught if that many comments is actually a bad thing?
Actually learning how to comment is a skill that is difficult to teach. Your prof’s approach is likely because he is just a tad lazy, but is also not totally terrible. As long as your comments aren’t just rewording the code, you’re doing ok. Explain the thinking behind it, why a more obvious approach isn’t valid, etc. Have fun!!
When you are writing a comment, think about the why not that what. The code already does and should be written in a manner that what is doing is clear. There are occasions that a “what” type of comment is needed. For example, the algorithm is complicated and long, or obscure, or using some hidden feature that it’s not so easy to understand by just looking at the code. Normally, a comment like is more useful when the piece of code is doing something that it’s not common for the environment/industry.
example 1 in Principle 9 is actually rather subtle O(n^2) because the reduce is O(n) and the spreading again is O(n), so its two nested loops. Another reason why the second approach is plain better ;) The second is only O(n), becasue its two consecutive loops.
I'm not sure if it would be another principle worth considering, but over the last year I've stopped dealing with the traditional try/catch error handling construct and instead return results and if my code throws an exception it is an "unhandled" exception to terminate the app (I'll have a hook into unhandled exceptions so I can log it before crash). You can think of it like Zig's "panic" feature. I just find the traditional try/catch something that makes code execution more difficult to reason about. I also use error codes so if the caller of a function received an error it's easier to check the error code for certain kinds of errors that can be handled or possibly panic. If I'm using a library that subscribes to the try/catch paradigm then I'll wrap it around my own thing that hides that construct (and only expose the things I use from the library). Probably not perfect and, I'm sure, the methodology would conjure up some discussion and debate but it has made my experience more fun and the code I write a bit more iron clad.
Nice list. For #1, I actually prefer writing the last conditional with an else statement and no return statement. If I have fresh eyes on the code, my first question would be "why does the child version have a return"
13:14 the reason why the bitwise expression is confusing in that example is because it evaluates to a number, 1 or 0, instead of true or false. Js deals with truthy/falsy for conditional expressions so it still works, but it should really evaluate to a boolean if being used in a conditional, e.g. n & 1 !== 0 Great tips otherwise 🔥
The side-effect with "never nesting" is you have a default behavior if a combination of checks unexpectedly work (either though a bug or unanticipated inputs). Could be good, could be bad, depending.
There's a principle I've been talking about that might fall under "single use" but I call it "avoiding code omniscience". When we develop things in a shared project, we don't want our code to impact current code or potential future code in a way that we didn't anticipate. But we also don't want our code to increase the complexity of the project such that people working on it require knowledge of your code - i.e. requiring code omniscience. Like if we have a validation process that prevents certain items being added to a cart in an ecommerce system, unless it's hooked in directly to the add to cart system, every controller that implements add to cart will need to be updated to add this - you've increased the complexity. Or similarly, if the code we write adds a flag to the item being added, and if it's there, the validation occurs. Someone rocking up years later would have to be aware of the flag if they want to make a new system that creates gift boxes (bundles of items) or something. Documentation and tests can solve this problem but....
this video slaps. I'm a huge believer that optimization should come only went a bottleneck presents itself, and in all other instances making code that is human readable is paramount. To assume how the code will be ran on the hardware, especially in javascript of all things, and attempt to pre-optimize is a fools errand. What is vastly better is easy to digest at a glance human readable code, where you can intuit how it should be used and how it can use extended and maintained. Great stuff!
Regarding Optimization. In more than 20 years writing Production Java code for one of the largest international banks, I have had to optimize my code twice. The first time a change to the XSLT library implementation made processing take 50x longer. Simply adding a line of code to set an undocumented but supported flag returned to normal. The second time, Dodd Frank caused volume to increase 10x overnight. Because I had written the worker code to be reentrant all I had to do was change config parameter # of workers from 1 to 16 to increase throughput 10-fold (and add an extra CPU core and run code analysis and deadlock detection overnight). Bottom line: Just write Clean Code and - as Conner says in the video - let the compiler worry about the insignificant local optimizations that are such tempting distractions; that's its job, not yours.
2:25 If it is just the comparison of two numeric values, then why we are sending the object to this utility function? Namely, that function does not need to know how the "length" of a "password" object is retrieved. Rather the place that is taking care of the business logic already know what password object is and how its length can be retrieved, so we can directly send the value instead of the object. With doing so we can convert this utility functions into a library which would not be depend on any object (e.g. StringUtils in Java Apache Commons). Plus, minimum password constant value can be retrieved from another constant util class's function instead of directly referencing to the variable from a function, so that the checkPasswordLength does not need to know also where actually that constant is located. My point is that, there should be minimum change in the rest when we change something in somewhere.
when it comes to quotes (in languages that accept both), I use single quotes generally, unless the string contains a single quote, in which case I use double quotes (I prefer that to escaping).
Nice content, good job. The first tip is a double edged sword. Too many control statements can also be confusing. You have to check the entire method to understand what all it can return. Sometimes the if-else blocks which populated a variable and return at the end is easier to understand.
Couldn't agree more with all of these, though Python has instilled a habit of using 'SINGLE_QUOTES' for constants and identifiers, and "double quotes" for misc./ "natural" text in me
Now a days, I write the code and get my code reviewed by AI to see if there can be any potential improvements or refactoring we can do in terms of clean code and it helps me to write clean code. Anyways these tips are really golden! A lot of times I make these mistakes and get review comments from my Tech lead about these mistakes
AI is very very bad at clean and efficient code! I used CODY for code smells and it doesn't even see context it says: "check that pointer" and I always go: "check line xyz for your remark number x" And it sees it's covered. It comes up with idiotic ideas of naming variables really long like n "number" for readability. For fuck sake, it's the only parameter in the function, the type says its an u64 what else do you want, Cody? I am not going to type number everytime :D So my advise is, be very critical because AI is really really bad at coding!
Suggestion: It's a very good useful video, but instead of “Principle #n” for the names of the timestamps, it would be more useful if they were meaningful names. I like to read the description and jump to specific topics. @sidouglas did something similar. It can be more succinct. For instance, instead of “Eliminate if/else branching; early termination”, it can be “Guard clauses” or “Return early”.
Nice video, thanks! Straight to the point! In principle 9, I think that the example was not very good regarding the 2 loops VS 1. Reduce can be a little hard to get used to, but once you do, it's really easy to read and understand what it does. I used to chain array functions as you did here, but mainly because I never took the time to work with reduce. Now it's just easy, and if you have big arrays, loop twice VS once is really not a good idea.. I agree thaw trying to be clever while programming does not make sense, but reduce is just a basic array operator 🤷🏻♂️
Not sure I agree with 9, i'll take that reduce instead because if we apply this rule, you'll see people chaining filter map sort and such which is rather messy imho
I just watched first part of the video, u removed if else and added more ifs, the pronlem with your code style is that when it will come to priority of your ifs, because just swap two of them and your code does not work as expected, it mean if i have 10 ifs i should keep in mind not to change the order od ifs in future when i add or remove a condition, good video and thanks for sharing but who made if else knew better than us if not they just made if without else.
🎯 Key points for quick navigation: 00:00 *🌐 Avoid unnecessary nesting and branching to improve code readability and maintainability.* 02:16 *📝 Clarify ambiguous code by using descriptive names and function names that reflect their behavior.* 04:09 *🔍 Use comments sparingly and effectively to explain complex logic rather than reiterating what the code does.* 05:32 *🛠️ Maintain consistent code formatting for readability and reduce potential errors.* 07:25 *🔄 Apply the DRY principle to eliminate code duplication and enhance maintainability.* 08:50 *⚡ Fail fast and early in functions to minimize unnecessary processing and enhance code efficiency.* 09:47 *🎩 Replace magic numbers or values with named constants to improve code readability and maintainability.* 11:27 *🔧 Ensure functions adhere to the Single Responsibility Principle to enhance clarity and maintainability.* 12:08 *🔄 Prefer clear and readable code over overly clever solutions for better maintainability and understanding.* 14:08 *⏳ Avoid premature optimization; optimize code when necessary based on actual performance bottlenecks.* Made with HARPA AI
"easy to understand" vs. "more performant" while I agree, creating code that is complex for the sake of convolution is bad. I am not in agreement with easy to understand code should be used over performant code. Just comment your complex functions or lines stating what it does. In applications that are running 100s or 1000s of actions these things don't really matter. When you're running millions of actions asynchronously sometimes milliseconds matter.
Bad Performance is often a result of poor data structure usage, or api calls, etc. VS something like “how should I sort this array”. In these cases I easy to understand is preferred imo.
1b) using exptions for flow control is not a good idea. Additionally showFullVersion and showChildrenVersion are alternatives. With new style it looks like showChildrenVersion is a kind of error handling. 5) you introduce a new method and increase complexity. Methods will be used. So someone will use it to log something else or even abuses it. And then other logic is coupled. You couple business logic of three methods. It they are independent than its a bad idea. It you want to make sure that they behave equally -only then - its a good idea. 8) such things can sometime be a good soliution. Especially if setting the var area later is costly. for instance if your are in a database transaction than this can be good. Or is you process an image and updating it later would mean to run through millions of pixels again. All other points are ok.
@@maelstrom57 using Exceptions for flow control is an anti pattern. (btw. cuddled else is never a clean code, especially with intendation of 2. And for the latter: showChildrenVersion is on the same abstraction level like showFullVersion. As I said: if there are barely any children than you construction is kind of ok. That is what i mean as kind of error handling. That a child wants to see this site is an error so to day. But if there are equally amount of children and adults than both methods are on the same level. They are alternatives. And a ternary operator (or a switch if more than two) would be a better solution than.
Does practicing all these things before seeing this video already makes me a senior dev? I don't even know what stage i am.. Maybe junior, intermediate or senior.. but whose checking
5:37 I use a combination. If I have variables at the beginning or end of the string, I don't use ${var} but var + or + var because that's two less character [4 chars VS. 2 chars]. If it's variables inside a string, it's more effective to use ${var} because it's less characters than using " + var + " [4 chars VS. 5 chars]. 8:37 I use if(variable?.trim?.()) [ I know it's not bulletproof because an object with a trim() method might come in, but in many cases it's legitimate to use my way, especially if I've tested somewhere before that it's not an object. ]
Hey Conner) thanks for the useful tips. But something is not clear for me in example with reduce and filter. If I'm not mistaken distructuring in reduce will cause the time complexity Am I right?
It can depend, but generally speaking no. It's a pretty common practice to add constants to the top of the file that sort of act as a configuration setup. That said, I don't think it would be the worst thing for this to be inside the function either.
For the first principle, inverting if..else only works in general with if..else throw or if..else return) but in most cases it will make everything worse. Here are some tips to keep track of what clause brackets corrsepond to -corresponding bracket have the same color on vscode you can clearly see it on the video. Maybe with another color theme they would be more visible -there is a keyboard shortcut to go from a closing/opening bracket to the corresponding opening/closing bracket -when your mouse is over a bracket the corresponding one will be highlighted .
I'm sure you can find a bunch of cases where this isn't optimal. But generally speaking, the important part of what he is demonstrating is that getting much of the validation out of the way right at the beginning of the function makes the rest of the code much more readable. The return statements on the bottom are ambiguous, but those are not the main point of the principle
