 oblique strategies for programmers: hardcode everything do repeat yourself you probably don’t need that optimise early, it’s fun try a new language feature maybe you can do this entirely with tags and filters try turning your program inside out restate the problem the naïve algorithm is probably fast enough it is easier to ask forgiveness than permission it is easier to correct a wrong program than to write a correct program create a new programming language for this 66 comments
  @arnelson i am saying these strategies are good and we should do them. reject arbitrary notions of programming morality @arnelson we might have a different notion of what makes a “good” program- but these notions are irrelevant to the process of creating a program in the first place. it is easier to correct a wrong program than to create a program perfectly from blank canvas  @marjolica @zens @arnelson bubble sort is great if you only have like 20 items, or the list is nearly sorted already 😃 @marjolica @zens @arnelson also if it's not fast enough it'll show up on the profiler and you'll fix it @JoshJers @zens @arnelson back in 1974 I inherited a large Fortran program (about 2000 punch cards) that ran on IBM370 mainframes. The purpose was to calculate the UK power stations fuel use. It used to take 2 minutes of CPU time to run the entire program. It included a bubble sort to put the list of the power stations into 'merit order'. There were about 3000 items in the list. When I replaced bubble sort by quick sort I reduced the total run time to 2 seconds with big savings to our budget. @marjolica @zens @arnelson Yeah, 3000 is definitely the wrong number of elements for bubble sort to be a good choice, *especially* in 1974 😁 @marjolica @JoshJers @arnelson well, the rule of thumb is “probably” not “definitely”. sometimes we work ourselves up about scaling problems that just never happen @zens @marjolica @arnelson exactly! The ultimate point is "write it the most straightforward way you can to start, then fix performance problems when they become apparent" @JoshJers @marjolica @arnelson and also, once again, these are strategies for when you get stuck, not general rules of thumb @zens good idea, might just need a bunch of xkcd panels on note cards, Randall Munroe is leaving money on the table  oblique strategies for programmers 2: don’t worry about the tests, failure is subjective delete the framework, see what doesn’t break consider goat farming quack quack you’re probably using the wrong version of a dependency check it in anyway, it’s someone else’s problem now try and refactor it, then lose all your unsaved work in a kernel panic, it really gives you perspective. play zelda for a while, the solution will come to you upon collecting your 68th restless cricket  @zens get stuck, argue with randoms about random stuff until the answer magic morphs into your neurons. getting some people who think this is a joke. No! I am entirely serious! let me explain. “Oblique Strategies” is a card deck by Brian Eno designed for getting him and his friends unstuck creatively, when they reach a creative block. Programmers get stuck too, and when we do, often it is because we have some unbreakable and unstated rule we’re holding in our heads uncriticially. We exclude a heap of solutions as “invalid” without trying them because we believe the solutions are not good enough according to some arbitrary standard. All the rules of programming are made up. mostly by white dudes looking to sell a seminar. If you can solve the problem and follow all of the rules you wanna follow? cool, good for you. this isn’t for you. but if you’re getting stuck because you’re hung up on an algorithm being suboptimal, even though it’s barely used? maybe let it go  @zens the number of times I’ve reviewed a pull request and approved with comments such as “this isn’t ideal but it’s got the job done and we need this out pronto” is pretty high. Sometimes good enough code is what’s required even when we endeavour for perfect. @zens THANK YOU for this thread there is so much Cargo Cult nonsense around programming a regex is only one problem spaghettify your code mix your concerns nest another loop don’t collaborate be imperative; the computer is just here to do your bidding oblique strategies for programmers 3: replace a hard problem with a much harder problem what would forth do? ask an expert in something other than programming rename all your variables after your children rewrite it from scratch in godot can you use the UI while you’re high? count your blessings, then curse twice that amount hilbert curves are neat. See if you can use a hilbert curve somehow. A Star is the wrong solution for this. GOTO considered helpful  @zens What if you didn't? Are you sure? How do you know? Defer the decision [e.g. last responsible moment]. Get rid of one thing. Take a shortcut. Relax an assumption. Code like nobody's watching.    Do it all with regular expressions Write it in Arm Cortex M0+ assembly - a Raspberry Pi Pico is only four bucks Test the implementation Metaprogram against your better judgement Change your editor's font Read the stdlib source Debug at a lower level Debug at an even lower level Ask the wrong question as publicly as you can get away with Re-code the same project over and over Read a related book Read an unrelated book @sakiamu @codefolio if I were to make a free pdf card deck template available would you mind if I included these?  @zens write code generators for things you can type in minutes, and code generators to make code generators when you get stuck, bored or both. > It's easier to ask forgiveness than permission. Lol. I don't know in which context you meant this, but it's excellent in many contexts: Imagine closing a prgram an then seeing a popup: "We're sorry, we didn't save your work." <Close> <Ok> @zens Oblique Programming Strategies sounds like a book that needs to be written. Each chapter covers a strategy with both imperative programming and functional programming examples. @SteeleDynamics and also logical, constraints, quantum, mechanical, and analog examples @SteeleDynamics let’s see if you can solve the problem using only beads and matchboxes Example 4.3: Suppose you need to augment a program such that the interface is additive. (i.e., Allowing for the easy addition of additional operations which lookup the correct procedure to call based on the types of arguments supplied. This is usually called Argument Dependent Lookup or ADL.) Rewrite the program to use an additive interface and create new procedures that get called using ADL. Answer 4.3: Question significant choices you've made in your life. @zens have done 100% of these... finally cracked "create a new programming language for this" this week @zens BTW, another technique I've personally found useful (with the same general vibe as this thread): write it in the stupidest / least clever way possible; only once that solution works do you optimize/clean up @zens is there a web page or style guide with these on somewhere so I can use it to nudge my team into less over-engineering? |
Gregory's Server
@zens Tag yourself, I’m Optimize Early, It’s Fun.