You misunderstood him. The person who admires simplicity because he understands the difficulty of the problem is the genius. I would love to see how Jia compares.
You should check out the average "grown" LabVIEW program, that's running an experiment that went through the hands of 3 generations of PhD grad students and their respective undergraduates. THAT is the true form of spaghetti code.
I don't really understand the point being made here... Vanishingly few docs have useful graphics, but they're important. What are we even looking at in this video, an OOP command flags parser? For OOP, a graph like this is actually useful.
@@sullivan3503 not about the literal graph, but more about how documentation, much like this graph here, is often a tangled incomprehensible mess, defeating the very purpose of its own existence
Someone spent 5 seconds configuring a plugin to autogenerate dependency graphs. More time was spent criticizing it here. It's still readable enough and it'd take more time handcrafting pixel-perfect graphs than you'd save because of better readibility. Plus, you'd risk people forgetting to updating the graph to reflect reality. Moreover, a lot of "dependencies" here are just basic utility types like "string", so we can't really conclude from this graph that this solution is overly complex (as some other comments suggest).
@@AlfaToTheOmega Nobody reads or cares about this graph and it's completely useless, no one gets anything out of it. But someone though it would be fine to autogenerate everything and dump it on a page without caring what it's outputting. I guarantee you 0 people have gotten any value from it. In addition, the graph software is poorly written for the various reasons blow points out so it's just stupid bullshit on more stupid bullshit.
@@CashsCoffeeguy is referring to the clean code book and the philosophy of breaking everything down into tiny methods that each do almost nothing and obscure what is going on in service of pretty looking object oriented code
When you write a single line of goto: REEEEEEEEEEEEEEEEE That's Horrible Programming, cause it breaks the flow When people write things like that: Everything is fine, that's expert software engineer right there!
@@chudchadanstud This statement contains too much dogma. There are definitely a few situations where it can make sense to use goto. And luminaries like Kernighan also confirm this.
@@starc0w There has never been a situation when it makes sense to use goto. Use a switch case, for/while loop, recursion functions etc. You have no reason to use them. They don't exist in Jia too. JB hates them too.
f you take this as him complaining about diagram generation, I get his point. If his argument is about the actual header dependencies, I mean who actually cares. It's meaningless. I bet it's mostly due to using lots of forward declarations to save on rebuild times. Remember, LLVM is a huge project that takes forever to rebuild.
I feel like the point of forward declarations is to cut down on includes, not tricking the compiler into accepting mutual and circular dependencies. It's a trick that's only available because .h files exist and need some help to not be hogs.
@ndy-dp1bh I honestly don't get people like you, or the guy in the video. I mean---He is literally the irony of his own video. He is complaining about mostly meaningless shit without providing any context. I worked on massive projects like these and eventually, if you run it through a dependency graph generator, you will end up with something as ugly as this. I don't know how about YOU, but from the video itself I can't even make out the details of the dependencies and without any in depth analysis it is very naive to claim that it is a "mess" that can be outright simplified. As far as there are no cycles and the dependencies are inverted---It is not a problem. Even if you are doing redundant includes, the compiler will optimize that away.
I'm actually not sure if he's complaining about the people who created the content of the graph or those that implemented how it's rendered... or is it both?
THIS! He is shooting the messenger. He must mistakenly think that the llvm project wrote the graph visualizer. Doxygen and graphviz are great. Jon does not get it.
I recently got into programming as a hobby, and when I make little apps for my own use, I don't really care how messy the code is. If it works, then I'm happy, and after a while I get the urge to improve it. I enjoy the process, but ultimately it would be a huge pain for anyone other than me to deal with it. I wonder if what's happening in the code world is that people refuse to care about the bigger picture, and just put their nose down on the specific task they want to accomplish without regard for how it will work for others
Not to that degree but it does. You will find people that submit pull requests with a few changes and each change mixes in something that they forgot to do earlier in the same PR. If you tell them to combine and simplify the PR they will look at you funny. They don't know what purpose the git history has in understanding the reason for a change and write crappy commit messages for thousand line diffs. They don't care about 'craft' and we all know a little how it feels. Mind you, I'm using the same approach to write crappy changes as I go along. But when I'm done, I clean up my mess even if it's for my own eyes.
Code should be periodically refactored like you do with your small projects. But with large projects where you have multiple programmers with varying degrees of design pattern knowledge and style, it becomes very difficult to do that. And to go through a codebase that large to refactor will take way too long. Time that no company will allow you to take. They'd rather have you work on projects that other clients are waiting for.
@@cristianstoica4544 yeah I always clean up. I cannot stand leaving code in a mess, for my own sanity. And I try to be as detailed as possible in my commit comments to explain the reason for why something was implemented or refactored in a specific way. I try to leave comments that I would have liked others to have left for me.
Usually when I come around such representations I remember my early code architecture curses from university: If you have crossing lines (which where actually not allowed in my UML course) it's under-abstraction. If you have graphs with many nodes down the line, it's over-abstraction. Here we have both :D
When I see complex and obscuring documentation like this I just do things my own - documentation should help, not to confuse. Though, if it was “my boss wants this to work” I’d be fucked. There’s a lot of software out there which is hyped like crazy with awesome demos, but when you want to do something specific to your environment you just get stuck with no help. In old good days it was just pure math and computers. Rock solid. You learn the math and just code it into computer. These days it’s about hype, loads of confusing abracadabra terminology and “duck you!” minded documentation.
There is nothing obscuring about it. In fact, people are accusing it of the opposite, of showing an include dependency graph - giving more details they don't feel they need or is useless. If you go read the auto-generated reference documentation then don't be surprised you find reference documentation. Which almost always implies that only parts of it will be useful because usually you only to the reference looking for specifics. Admittedly this is not a pretty (auto-generated) graph, but for any C++ developer it obvious what it is (since it says next to it). If one is perusing the *reference documentation* and *not* looking for an include graph then one wouldn't bother spending time on it. Blow spent more time laughing on this than there most likely ever was spent CPU time on generating the page. I don't know the context from this clip. But if the idea was to learn how to use this Command Line API, then one does not go the reference documentation, but rather documentation actually meant to teach usage, like llvm.org/docs/CommandLine.html#quick-start-guide .. (okay, probably not command line, but codegen flags, but then reference doc is generally the last and not first place to go).
@@dumbfailurekms That “new school” of programming started like 10 years ago or something. Most of it is just extreme hype over some basic concept or idea. And obviously documentation is not there when you want to do something complex beyond glorified demos. I figured out just to do things my own way with barebones programming language - it’s faster and stress-free. Shitty libraries and documentation is not my problem.
@@thewhitefalcon8539 There is so, so many shitty production code out there just because dev’s were not able to figure out how to use certain platform ans just “patched” things together and released just to get this shit off their shoulders. This is modern day programming, nobody cares about quality, the code is barely operational.
I don't get what he finds so funny about this. This is a typical GraphViz diagram generated automatically from header dependencies. This has nothing to do with "actual programming". And nobody in the comment section acknowledged that. How dumb can Blow fans be?
I'm no expert but the funny part is that the graph is a tangled mess. Which seems to be a problem with the coding style or the way languages are designed
I might be wrong, but isn't this (autogenerated) diagram just an #include hierarchy of a C++ header file? The file includes five other header files, also listed in the top left corner, among others vector and string. These typically depend on some more headers themselves (often times the same ones), so you get a rather dense dependency graph. Doesn't necessarily mean the underlying code is bad. Otoh, this kind of representation isn't particularly helpful either. And yes, #includes are a rather crude and outdated mechanism, basically recursively copying pieces of text into your compilation units. This is programming like 40 years ago, and brings some serious problems. That's why modern C++ offers modules as a replacement.
yeah, and the url is clearly visible in the video. im not sure why more people didnt check. its a stupid graph but it doesnt speak at all of the code quality. maybe people are just looking for a reason to get mad
@@MightyAlex200 How is it even a stupid graph? Because the indegree of some of the nodes is a bit high? Many graphs of such complexity that they are actually useful is bound to have an indegree like that!
It's like having a separate header file for each declaration or define (which, btw, should be accessed only via microservice or, in a worst case scenario, singleton).
This is LLVM, it’s known for being complex. I believe the complexity is stemming from supporting so many OSs that everything has been abstracted to a degree that most of us should never see. I wonder if this is following clean code principles, that also could be why it looks like this.
The abstraction is the problem. Everyone is so pre-occupied for being "general". It doesn't have to be general, literally just write a fucking separate targeted implementation rather than making abstract functions that work on every platform. That's what leads to shit like this. I work with the unreal engine code base and it's the same shit because they try to make everything super abstract and general and no one (not even the devs by the comments in the code) understands wtf it's doing.
@@astrixx If abstraction is your issue, then you should consider writing your code in Assembly, which is very platform dependent. All programming languages are written as abstractions and you can’t really remove that if you want to port them to different platforms, the problem will always be present at some point - even if it is just a bunch of functions, someone will turn those into an abstraction layer for their project, but now there isn’t any team that is tasked to maintaining the integrity of that code and the likelihood that many individual abstractions will fail is higher. So, it’s better to take the lesser of two evils, or maybe do the Python approach, where the low level code is available, but it’s recommended to use the higher level abstraction in most projects - you can look at the the packages in the os library as one of the best examples, this could be taken all the way down in a library for writing programming languages. Just remember that every language is a trade off between simplicity and deep control of the hardware, so most are extremely abstract and the complexity in undoing those abstractions will be messy.
@@astrixx I completely agree. The desire to abstract everything and utilise generics as much as possible leads to a code that's both unreadable and unmaintainable
AI code will look much, much worse. AI will, for instance, constantly invent new communication protocols. Instead of standardizing, it will constantly obfuscate.
@@lepidoptera9337 in time AI will code quality might trend downwards, since it will train using public repositories like github and in time more and more AI generated code will be in those repos, so it will feed itself the garbage code it produces.
It's not just "look at the documentation" in this case, it's reserve a whole week, prepare various jugs of coffee, get a notepad and a pen, and plow through this nonsense we've left you
puzzle game developer rants about actually useful and complex software and there are 150 meatriders in the comments who learned how to install linux 2 years ago parroting him like they're L8 engineers from AWS
Yup, the Dunning-Kruger here is astonishing, especially from commenters, but JoBlow too. I doubt most commenters here have done anything more complex than a CSV parser, or a basic web server.
It's funny how you specifically call him "puzzle game developer" to disregard his 30 years of programming experience, including some noticeable achievements (which are hard to underestimate, like making a programming language, and making multiple critically and commercaly acclaimed games (one of which is 3D 1st person open world), from ground up, without any game engine) You definetly have some similar amount of experience to talk like that, right?
No thats the whole point of it. They use unintelligent artifical intelligence to save time. That the result is just worth nothing is a sympthom of modern software development.
Include (or import) dependencies are IMO one of the strongest arguments against using text files as source code. A file is an artificial grouping of symbol definitions and creates lots of unwanted dependency sprawl.
Bad documentation on what? The programming language he hasnt done yet and which is in closed beta? Yea right, he should definetly focus on docs right now, very important
They never heard about Miller rule that humans can focus on 7-10 things at time. For IT/CS people it implies that a single diagram should never have more than 7-10 components or the client or your team or whoever that you explain system to may have problems of understanding it. Always split huge diagram into small ones.
Yes, and some of it is bad, just like some text-based programming is bad. This graph technically contains a lot of information, but none of it use usable for a human.
When writing code you connect a bunch of characters together. It isn't all that different. The good thing about node based is that syntax errors doesn't exist, the bad thing is that nodes and wires take up a lot of space. In the future we will have a middle ground, a code editor where you will be able to write code without the ability to express either syntax or even type errors.
Visual programming is a thing but it is normally a very bad idea because it leads to confusing spaghetti like this. The only reason to use visual programming is if you don't understand text. It's generally much quicker to type than it is to draw the same thing (and you can fit more on the page when it's typed). This video isn't about visual programming though, just a diagram of the codebase.
This is why I hate many highly advertised/popular software ... libraries, frameworks, you name it. Many big name doodads have garbage documentation, auto generated mess plus a few "cookbook examples", and when you cannot fucking get productive trying to follow that shit and ask around, you get told you are stupid. Why cannot we have proper documentation that contains actual prose explaining the semantics, and not only list APIs with comments? Maybe too much to ask goddamn.
I think the point is that *he* would not choose to generate it. But sure, the complexity of the code here might not be the real problem, it's how this "bad attempt at a ball of threads" is not helping anyone. (Except Jon to have a good laugh.)
Is not the fact that "complex code makes complex graphs" that is true, is why make a graph system, that ends making graphs for complex code that are unreadable. And then, have the courage to call that docs! IMO is better to just not show any graph at all, IMO no help is better than bad help.
What the hell even is this? Like a graph of what types are used in what functions and what functions are called by other functions or something? I don't even know what the hell the arrows are trying to illustrate. Dependency tree or something? Someone explain, please...
i was about to say the same thing. if i was to graph all the signal connections between functions for a program im currently working on.... this is probably what it would look like :x
Its a class dependency graph, which is, to be blunt, a very stupid thing to print out. Usually when you are designing a OOP class system each class depends on dozens of others because you are trying to encapsulate complex things inside it. So you end up with this ridiculous amount of arrows. It might be useful to do it centered on 1 class, kind of like a "find all references", but even the utility of that drawn as a graph is debatable. Drawing it N:N is idiotic. What dependency graphs should really be used for is for module or layer dependencies, where at that level you don't expect one module to depend on too many other ones. Or you might want to quickly identify circular dependencies you might need to deal with. This is pretty much a case of the wrong tool for the job, and Jonathan should know that and explain instead of just laughing at it.
junior game programmer here, and I do this. matter of fact honestly mine are worse. Can someone kindly explain what i should do instead? I just, when I write state machines I want to know the complex ins and outs of which states transition to which states, so i am not just going in blind and messing up the extremely important order of state change logic.
You shouldn't do anything else instead, this video is just about ego and pretending to be better than others. These are the types of graphs you get when you follow clean code principles. Some who pretend to be better than other say you have to follow clean code guidelines and other as we can see here say you shouldn't. At the end the only result that really counts is if it works and performs, that's all your future boss will be interrested in, he couldn't care less about the rest.
@@dawidkotlinski Nobody, don't try to be smart. Try learning to read the entire text since you clearly did not understand it and chose just one word out of the entire text and responded to that one word. That is exactly the same as randomly shouting pancake. It makes no sense.
@@TheGalantir couldn't agree more. 95% of "tech influencers" are all about that. one of the most complex c++ projects available online, Unreal Engine, is way worse than LLVM in terms of headers and code complexity in general. yet I don't see it failing because of it
For state machines you have to write the states interpreter and them just create tables of rules. This is how they have been written since 1980. There is nothing to mess, not even code to write.
I solve problems like this with my homemade software that creates zoomable, rotatable colour-coded 3D connection models. It has arrow merging, too, if the 3 dimensions are not otherwise sufficient to facilitate legible geometries.
it's just a graph, auto-generated for the reader's convenience. if you don't like this way of displaying the information then just don't look at it lol? there's lists displaying the same information a bit further down
What, exactly, can you see in this mess? Absolutely nothing. It violates everything we know about how the brain processes information AND everything we know about how computers should be processing information.
John's reaction on the graph reminds me of this discussion I had with my roommates the other day. I'm in electrical engineering, one is in industrial engineering, one is in chemical engineering, and the fourth is in biological engineering. One of them made a joke that if we combined our collective brainpower, we might be able to solve a three-body diagram. I legit went into an almost 10min long fit of laughter.
It sounds as if you could write a lot of good jokes with that premise. Have the four of you ever gone fishing? Or rock-climbing? Or tried to assemble IKEA furniture?
I totally get the general sentiment that this spaghetti stuff is unreadable, but his laughing isn't illuminating or productive. Worse, it remains entirely possible (however improbable) that there is a lot of wisdom in the diagram that just hasn't been explained properly through the documentation. This reminds me of a TED talk where Eric Berlow shows how you can drastically simplify a complex food chain in an ecosystem by focusing on "nodes of influence" -- that is, by both grouping similar vectors and then prioritizing the ones which have the most impact. My favorite line is at the end: *"[What we found is] the more you step back and embrace complexity, the better chance you have of finding simple answers -- and it's often different than the simple answer you started with. So, for any problem, the more you can zoom out and embrace complexity, the better chance you have of zooming in on the simple details that matter most."* ruclips.net/video/UB2iYzKeej8/видео.htmlsi=kG_Vu6CsEG2sthv0
When Jonathan started laughing manically, I thought of a crossover with Terry from TempleOS for some reason. I'd think of Terry's voice asking "is that programming *slur word*" and "they glow in the dark". There is something about being the outcast criticising the whole software industry on how they do stuff and you doing things in your own way while laughing hysterically on how things have become.
no wonder I cannot learn programming, someone make it so complicated? I need to learn how to be natural programmer. Maybe Mr Jonathan Blow can help us become natural programmer and make everything simple.
Because I am an idiot, I’m not sure what I’m looking at: is that a chart of every function call or something? Is that something that can be generated for any language or is that something special to whatever language he’s looking at? Could someone please explain to my dumb ass what I’m looking at 😂
It's funny to read the lofty academic debates down here when it's clear most people (myself included) are unsure what exactly Jon is criticising here. The automatic diagram generation? The complexity? The imports? That some dude actually thinks programming is doodling a line with a marker around some boxes? Is this complexity even relevant to anyone? Who knows, but I'll just interpret it however I like it the most and plaster my rant here
Love it or hate it this is how young people are being exposed to programming (i hate it). Programming used to be done by punch cards. It's one positive trait is that it is a decent way to explore concepts with minimal syntax with immediate visual cues. Whether or not this type of programming has merit or not is irrelevant; Young people are using these tools.
That's not programming, it's cable management!
Good one. xd
Lmao
i have seen cable management. capable management can be beautiful.
this is not cable management.
this is lack of cable management.
Cable mismanagement @@tanko.reactions176
it looks like my cable management anyway
"An idiot admires complexity, a genius admires simplicity." -Terry A. Davis
You misunderstood him. The person who admires simplicity because he understands the difficulty of the problem is the genius.
I would love to see how Jia compares.
A realist admires salary
the most powerfull sentence of our time
@@kuklama0706 real big brain here
"so grug say again and say often: complexity very, very bad"
I've never seen the phrase "spaghetti code" represented so literally before!
You should check out the average "grown" LabVIEW program, that's running an experiment that went through the hands of 3 generations of PhD grad students and their respective undergraduates. THAT is the true form of spaghetti code.
open source dev: read the docs
the docs:
i dont like it, but its true...
The docs:
My experience trying to setup neovim
I don't really understand the point being made here... Vanishingly few docs have useful graphics, but they're important. What are we even looking at in this video, an OOP command flags parser? For OOP, a graph like this is actually useful.
@@sullivan3503 not about the literal graph, but more about how documentation, much like this graph here, is often a tangled incomprehensible mess, defeating the very purpose of its own existence
Someone somewhere is very proud of that graph
Only if they have seen it. Which is unlikely because idiots never view their creation as a whole.
I mean idk if I could do the paths that well (to go around other objects and stuff) but I'm somewhat of a noob
Someone spent 5 seconds configuring a plugin to autogenerate dependency graphs. More time was spent criticizing it here.
It's still readable enough and it'd take more time handcrafting pixel-perfect graphs than you'd save because of better readibility. Plus, you'd risk people forgetting to updating the graph to reflect reality.
Moreover, a lot of "dependencies" here are just basic utility types like "string", so we can't really conclude from this graph that this solution is overly complex (as some other comments suggest).
@@AlfaToTheOmega Nobody reads or cares about this graph and it's completely useless, no one gets anything out of it. But someone though it would be fine to autogenerate everything and dump it on a page without caring what it's outputting. I guarantee you 0 people have gotten any value from it. In addition, the graph software is poorly written for the various reasons blow points out so it's just stupid bullshit on more stupid bullshit.
Imagine how much time they must've spent getting rid of recursive references
Yeah, it's weird they even managed to compile it.
They did not
what a decade of "clean code" gurus does to ya
can you elaborate on this? just trying to learn
@@CashsCoffeeguy is referring to the clean code book and the philosophy of breaking everything down into tiny methods that each do almost nothing and obscure what is going on in service of pretty looking object oriented code
I guess reading clang llvm-ir output is way, way easier than figuring out any of this.
@ctxz9580I did the same too
im never complaining about template errors ever again
When you write a single line of goto: REEEEEEEEEEEEEEEEE That's Horrible Programming, cause it breaks the flow
When people write things like that: Everything is fine, that's expert software engineer right there!
exactly lol
I've seen colleagues act disgusted when they see a medium-sized if-else-if chain... Feels like I'm working with professional cargo cultists.
Don't ever use goto, like ever. They are unscoped loops.
@@chudchadanstud
This statement contains too much dogma.
There are definitely a few situations where it can make sense to use goto.
And luminaries like Kernighan also confirm this.
@@starc0w There has never been a situation when it makes sense to use goto. Use a switch case, for/while loop, recursion functions etc.
You have no reason to use them. They don't exist in Jia too. JB hates them too.
f you take this as him complaining about diagram generation, I get his point. If his argument is about the actual header dependencies, I mean who actually cares. It's meaningless. I bet it's mostly due to using lots of forward declarations to save on rebuild times. Remember, LLVM is a huge project that takes forever to rebuild.
I feel like the point of forward declarations is to cut down on includes, not tricking the compiler into accepting mutual and circular dependencies. It's a trick that's only available because .h files exist and need some help to not be hogs.
yeah, sanest take in here
Yeah, feels like he conflates mediocrity of the visualization tool with the codebase itself.
@ndy-dp1bh I honestly don't get people like you, or the guy in the video.
I mean---He is literally the irony of his own video.
He is complaining about mostly meaningless shit without providing any context.
I worked on massive projects like these and eventually, if you run it through a dependency graph generator, you will end up with something as ugly as this. I don't know how about YOU, but from the video itself I can't even make out the details of the dependencies and without any in depth analysis it is very naive to claim that it is a "mess" that can be outright simplified.
As far as there are no cycles and the dependencies are inverted---It is not a problem.
Even if you are doing redundant includes, the compiler will optimize that away.
Am I the only one that thinks the blow fan channels are copying each other
they definitely are
Yeah, I saw this same moment two times already. I even copied my comment from one of those videos here.
Hey now, no consthpiracy theories allowed! ☝💢
@BufordTJustice42069 Wordcel
I don't mind if they're copying each other. The alternative would be trying hard not to cover the same material in a race to the bottom.
rare jblow laughing moment
He is literally always laughing in every video
I just had some pizza, but this made me want to make spaghetti 😋
I'm actually not sure if he's complaining about the people who created the content of the graph or those that implemented how it's rendered... or is it both?
THIS! He is shooting the messenger. He must mistakenly think that the llvm project wrote the graph visualizer. Doxygen and graphviz are great. Jon does not get it.
I recently got into programming as a hobby, and when I make little apps for my own use, I don't really care how messy the code is. If it works, then I'm happy, and after a while I get the urge to improve it. I enjoy the process, but ultimately it would be a huge pain for anyone other than me to deal with it.
I wonder if what's happening in the code world is that people refuse to care about the bigger picture, and just put their nose down on the specific task they want to accomplish without regard for how it will work for others
Thats not my experience, but i guess youre free to speculate on things you know nothing about.
Not to that degree but it does. You will find people that submit pull requests with a few changes and each change mixes in something that they forgot to do earlier in the same PR. If you tell them to combine and simplify the PR they will look at you funny. They don't know what purpose the git history has in understanding the reason for a change and write crappy commit messages for thousand line diffs.
They don't care about 'craft' and we all know a little how it feels. Mind you, I'm using the same approach to write crappy changes as I go along. But when I'm done, I clean up my mess even if it's for my own eyes.
Code should be periodically refactored like you do with your small projects. But with large projects where you have multiple programmers with varying degrees of design pattern knowledge and style, it becomes very difficult to do that. And to go through a codebase that large to refactor will take way too long. Time that no company will allow you to take. They'd rather have you work on projects that other clients are waiting for.
@@cristianstoica4544 yeah I always clean up. I cannot stand leaving code in a mess, for my own sanity. And I try to be as detailed as possible in my commit comments to explain the reason for why something was implemented or refactored in a specific way. I try to leave comments that I would have liked others to have left for me.
Software developers are paid a lot of money to not do this
Oh Jesus they built a NEURAL NETWOKR!
Usually when I come around such representations I remember my early code architecture curses from university:
If you have crossing lines (which where actually not allowed in my UML course) it's under-abstraction.
If you have graphs with many nodes down the line, it's over-abstraction.
Here we have both :D
That's horrible cable management.
Elon Musk: "Good part is no part"
Programmers: ...
When I see complex and obscuring documentation like this I just do things my own - documentation should help, not to confuse. Though, if it was “my boss wants this to work” I’d be fucked. There’s a lot of software out there which is hyped like crazy with awesome demos, but when you want to do something specific to your environment you just get stuck with no help. In old good days it was just pure math and computers. Rock solid. You learn the math and just code it into computer. These days it’s about hype, loads of confusing abracadabra terminology and “duck you!” minded documentation.
relax buddy you were what? 4 yrs old during the old days
There is nothing obscuring about it. In fact, people are accusing it of the opposite, of showing an include dependency graph - giving more details they don't feel they need or is useless. If you go read the auto-generated reference documentation then don't be surprised you find reference documentation. Which almost always implies that only parts of it will be useful because usually you only to the reference looking for specifics.
Admittedly this is not a pretty (auto-generated) graph, but for any C++ developer it obvious what it is (since it says next to it). If one is perusing the *reference documentation* and *not* looking for an include graph then one wouldn't bother spending time on it. Blow spent more time laughing on this than there most likely ever was spent CPU time on generating the page.
I don't know the context from this clip. But if the idea was to learn how to use this Command Line API, then one does not go the reference documentation, but rather documentation actually meant to teach usage, like llvm.org/docs/CommandLine.html#quick-start-guide .. (okay, probably not command line, but codegen flags, but then reference doc is generally the last and not first place to go).
This documentation was produced by people whose job is to make documentation look impressive.
@@dumbfailurekms That “new school” of programming started like 10 years ago or something. Most of it is just extreme hype over some basic concept or idea. And obviously documentation is not there when you want to do something complex beyond glorified demos. I figured out just to do things my own way with barebones programming language - it’s faster and stress-free. Shitty libraries and documentation is not my problem.
@@thewhitefalcon8539 There is so, so many shitty production code out there just because dev’s were not able to figure out how to use certain platform ans just “patched” things together and released just to get this shit off their shoulders. This is modern day programming, nobody cares about quality, the code is barely operational.
I don't get what he finds so funny about this. This is a typical GraphViz diagram generated automatically from header dependencies.
This has nothing to do with "actual programming".
And nobody in the comment section acknowledged that. How dumb can Blow fans be?
I'm no expert but the funny part is that the graph is a tangled mess. Which seems to be a problem with the coding style or the way languages are designed
I wonder if he realises the llvm guys didn’t write doxygen or graphviz
Not relevant
What makes you think he doesn't?
Now this is REAL speghetti code 😂
"a picture is worth a thousand words" well a messy diagram is worth a thousand spaghetti sauce
smallest javascript dependency graph
"An idiot admires complexity, a genius admires -simplicity- a good laugh." -Terry A. Davis
I might be wrong, but isn't this (autogenerated) diagram just an #include hierarchy of a C++ header file? The file includes five other header files, also listed in the top left corner, among others vector and string. These typically depend on some more headers themselves (often times the same ones), so you get a rather dense dependency graph. Doesn't necessarily mean the underlying code is bad. Otoh, this kind of representation isn't particularly helpful either. And yes, #includes are a rather crude and outdated mechanism, basically recursively copying pieces of text into your compilation units. This is programming like 40 years ago, and brings some serious problems. That's why modern C++ offers modules as a replacement.
yeah, and the url is clearly visible in the video. im not sure why more people didnt check. its a stupid graph but it doesnt speak at all of the code quality. maybe people are just looking for a reason to get mad
@@MightyAlex200 How is it even a stupid graph? Because the indegree of some of the nodes is a bit high? Many graphs of such complexity that they are actually useful is bound to have an indegree like that!
100% agree with you.
This video is an irony of the exact thing he is complaining about. Holy hell.
@@TootNZ Fully agree with you.
I check the URL but the graph does not automatically show. How do I get that graph to appear?
It's like having a separate header file for each declaration or define (which, btw, should be accessed only via microservice or, in a worst case scenario, singleton).
microservices and singletons, found the OOP corporate slave
It's the Lorentz Attractor!
nowadays internet is for looking "how not to program"
This one really got him 💀
This is LLVM, it’s known for being complex. I believe the complexity is stemming from supporting so many OSs that everything has been abstracted to a degree that most of us should never see. I wonder if this is following clean code principles, that also could be why it looks like this.
ya its not a bs platform game u can shit out on godot these dayz lel.
people dont recommend clean code principles anymore, most of the advice is terrible
The abstraction is the problem. Everyone is so pre-occupied for being "general". It doesn't have to be general, literally just write a fucking separate targeted implementation rather than making abstract functions that work on every platform. That's what leads to shit like this. I work with the unreal engine code base and it's the same shit because they try to make everything super abstract and general and no one (not even the devs by the comments in the code) understands wtf it's doing.
@@astrixx If abstraction is your issue, then you should consider writing your code in Assembly, which is very platform dependent. All programming languages are written as abstractions and you can’t really remove that if you want to port them to different platforms, the problem will always be present at some point - even if it is just a bunch of functions, someone will turn those into an abstraction layer for their project, but now there isn’t any team that is tasked to maintaining the integrity of that code and the likelihood that many individual abstractions will fail is higher. So, it’s better to take the lesser of two evils, or maybe do the Python approach, where the low level code is available, but it’s recommended to use the higher level abstraction in most projects - you can look at the the packages in the os library as one of the best examples, this could be taken all the way down in a library for writing programming languages.
Just remember that every language is a trade off between simplicity and deep control of the hardware, so most are extremely abstract and the complexity in undoing those abstractions will be messy.
@@astrixx I completely agree. The desire to abstract everything and utilise generics as much as possible leads to a code that's both unreadable and unmaintainable
his laugh 😂😂😂😂
Exactly hahah, I couldn’t help but join in. Legit laugh.
I don't understand what's wrong. I'm in biology and I work with graphs like this every day
it's drawing time
"AI will replace us!"
What AI code will look if it were unchained without little to no programmers:
AI code will look much, much worse. AI will, for instance, constantly invent new communication protocols. Instead of standardizing, it will constantly obfuscate.
@@lepidoptera9337 Not to mention it'll constantly do stupid shit no one asked for because it's an effing prediction machine not an intelligence
@@lepidoptera9337 in time AI will code quality might trend downwards, since it will train using public repositories like github and in time more and more AI generated code will be in those repos, so it will feed itself the garbage code it produces.
It's not just "look at the documentation" in this case, it's reserve a whole week, prepare various jugs of coffee, get a notepad and a pen, and plow through this nonsense we've left you
lol did they delete it after this? I don't see it on the page
I think so, i dont see it too. You can see it in the wayback machine though
I actually had to use Wayback Machine in order to find it. I wanted to use it for the thumbnail. It's those blue lines in the thumbnail.
puzzle game developer rants about actually useful and complex software and there are 150 meatriders in the comments who learned how to install linux 2 years ago parroting him like they're L8 engineers from AWS
Yup, the Dunning-Kruger here is astonishing, especially from commenters, but JoBlow too. I doubt most commenters here have done anything more complex than a CSV parser, or a basic web server.
It's funny how you specifically call him "puzzle game developer" to disregard his 30 years of programming experience, including some noticeable achievements (which are hard to underestimate, like making a programming language, and making multiple critically and commercaly acclaimed games (one of which is 3D 1st person open world), from ground up, without any game engine)
You definetly have some similar amount of experience to talk like that, right?
@@nerdError0XF Yes.
I see Windows, I leave.
Literal spaghetti code
The LLVM team has been real busy! 🤣😂🤣😂
No thats the whole point of it. They use unintelligent artifical intelligence to save time. That the result is just worth nothing is a sympthom of modern software development.
LLVM use Ai? where did you hear that@@llothar68
I guess Jon Blow doesn't like spaghetti.
I can't tell whether you are making fun of the guys that wrote the code the graphic was generated from or the guys that wrote the graphic generation.
well why choose
It looks like the milky way.
where is the meatballs and the tomato sauce?
The tofudreg of information technology
Include (or import) dependencies are IMO one of the strongest arguments against using text files as source code. A file is an artificial grouping of symbol definitions and creates lots of unwanted dependency sprawl.
I had to read the UML spec and I must justify that
poor spaghetti didn't deserve to be roasted that bad
that's what dogmatic belief looks like
Bad documentation has got to be something you get judged for at the pearly gates
Bad documentation on what? The programming language he hasnt done yet and which is in closed beta? Yea right, he should definetly focus on docs right now, very important
Knees weak, arms are heavy...
Maybe these "elitists" don't want you to understand their systems. If you did, then they wouldn't be special anymore.
programming is like gaming, but programming
Graphviz hell. BTW. This is not that bad. Have you seen diagrams generated by Yosys?
It almost looks like they were attempting to represent code as a wiring diagram for a neural net, and forgot to color code the wires.
I don't know what's going on here, who Blow is, or why this is in my recommendations, but man I do know I could go for some spaghetti right about now.
Spegetti code goes brrr. 😂
Last time I made something that looks like that my cheap biro was refusing to work.
Keep this channel active bro pls !
True. Merging arrows sounds like serious programming. And not that shit we see here.
"it would still be unreadable, but it would be less unreadable"
They never heard about Miller rule that humans can focus on 7-10 things at time. For IT/CS people it implies that a single diagram should never have more than 7-10 components or the client or your team or whoever that you explain system to may have problems of understanding it. Always split huge diagram into small ones.
The definition of spaghetti code.
I calculated only about 60 objects. Not so complicated compared to our projects. We have like 7 stages inheritance chains.
Isn't visual programming a thing? instead of writing code you connect a bunch of nodes together? kind of like Scartch, Unity, or Unreal Engine?
Yes, and some of it is bad, just like some text-based programming is bad. This graph technically contains a lot of information, but none of it use usable for a human.
When writing code you connect a bunch of characters together. It isn't all that different. The good thing about node based is that syntax errors doesn't exist, the bad thing is that nodes and wires take up a lot of space.
In the future we will have a middle ground, a code editor where you will be able to write code without the ability to express either syntax or even type errors.
Visual programming is a thing but it is normally a very bad idea because it leads to confusing spaghetti like this. The only reason to use visual programming is if you don't understand text. It's generally much quicker to type than it is to draw the same thing (and you can fit more on the page when it's typed). This video isn't about visual programming though, just a diagram of the codebase.
This is why I hate many highly advertised/popular software ... libraries, frameworks, you name it. Many big name doodads have garbage documentation, auto generated mess plus a few "cookbook examples", and when you cannot fucking get productive trying to follow that shit and ask around, you get told you are stupid. Why cannot we have proper documentation that contains actual prose explaining the semantics, and not only list APIs with comments? Maybe too much to ask goddamn.
"I dont understand the diagram heh heh heh heh heh heh Hilarity"
Do we think a graph like this wouldn't be generated for this man's code?
Facts. Any useful software project is likely to have a dependency graph of this complexity.
I think the point is that *he* would not choose to generate it.
But sure, the complexity of the code here might not be the real problem, it's how this "bad attempt at a ball of threads" is not helping anyone. (Except Jon to have a good laugh.)
Is not the fact that "complex code makes complex graphs" that is true, is why make a graph system, that ends making graphs for complex code that are unreadable. And then, have the courage to call that docs! IMO is better to just not show any graph at all, IMO no help is better than bad help.
this fucking game developer lmao.
What the hell even is this? Like a graph of what types are used in what functions and what functions are called by other functions or something? I don't even know what the hell the arrows are trying to illustrate. Dependency tree or something? Someone explain, please...
i was about to say the same thing. if i was to graph all the signal connections between functions for a program im currently working on.... this is probably what it would look like :x
Its a class dependency graph, which is, to be blunt, a very stupid thing to print out. Usually when you are designing a OOP class system each class depends on dozens of others because you are trying to encapsulate complex things inside it. So you end up with this ridiculous amount of arrows.
It might be useful to do it centered on 1 class, kind of like a "find all references", but even the utility of that drawn as a graph is debatable. Drawing it N:N is idiotic.
What dependency graphs should really be used for is for module or layer dependencies, where at that level you don't expect one module to depend on too many other ones. Or you might want to quickly identify circular dependencies you might need to deal with.
This is pretty much a case of the wrong tool for the job, and Jonathan should know that and explain instead of just laughing at it.
junior game programmer here, and I do this. matter of fact honestly mine are worse. Can someone kindly explain what i should do instead? I just, when I write state machines I want to know the complex ins and outs of which states transition to which states, so i am not just going in blind and messing up the extremely important order of state change logic.
You shouldn't do anything else instead, this video is just about ego and pretending to be better than others.
These are the types of graphs you get when you follow clean code principles.
Some who pretend to be better than other say you have to follow clean code guidelines and other as we can see here say you shouldn't.
At the end the only result that really counts is if it works and performs, that's all your future boss will be interrested in, he couldn't care less about the rest.
Why model a state machine with classes
@@dawidkotlinski Nobody, don't try to be smart. Try learning to read the entire text since you clearly did not understand it and chose just one word out of the entire text and responded to that one word. That is exactly the same as randomly shouting pancake. It makes no sense.
@@TheGalantir couldn't agree more. 95% of "tech influencers" are all about that. one of the most complex c++ projects available online, Unreal Engine, is way worse than LLVM in terms of headers and code complexity in general. yet I don't see it failing because of it
For state machines you have to write the states interpreter and them just create tables of rules. This is how they have been written since 1980. There is nothing to mess, not even code to write.
Good lord, so much elitism.
High Level Programming taken too far
I solve problems like this with my homemade software that creates zoomable, rotatable colour-coded 3D connection models. It has arrow merging, too, if the 3 dimensions are not otherwise sufficient to facilitate legible geometries.
I think it's sort of cute than Jonathan named his PC "Warrior".
it's just a graph, auto-generated for the reader's convenience. if you don't like this way of displaying the information then just don't look at it lol? there's lists displaying the same information a bit further down
What, exactly, can you see in this mess? Absolutely nothing. It violates everything we know about how the brain processes information AND everything we know about how computers should be processing information.
John's reaction on the graph reminds me of this discussion I had with my roommates the other day. I'm in electrical engineering, one is in industrial engineering, one is in chemical engineering, and the fourth is in biological engineering. One of them made a joke that if we combined our collective brainpower, we might be able to solve a three-body diagram. I legit went into an almost 10min long fit of laughter.
It sounds as if you could write a lot of good jokes with that premise. Have the four of you ever gone fishing? Or rock-climbing? Or tried to assemble IKEA furniture?
Please tell us what happens when you four enter a bar
Basic spaghetti programming all over again 😵😵
You are sayin this as if you wrote at least single line of code in your life.
Who's THEY.
WHAT'S THE CONTEXT!
They - computer science professors and experts
This is what happens when you give graph viz dot any kind of complexity to render. Useless for visualising the graph, ironically
Average Unreal Engine Blueprint experience
They’re not making French Fries, Jonathan …
Makes me hungry for spaghetti
"It would still be unreadable, but it would be less unreadable"
Interesting
I totally get the general sentiment that this spaghetti stuff is unreadable, but his laughing isn't illuminating or productive. Worse, it remains entirely possible (however improbable) that there is a lot of wisdom in the diagram that just hasn't been explained properly through the documentation. This reminds me of a TED talk where Eric Berlow shows how you can drastically simplify a complex food chain in an ecosystem by focusing on "nodes of influence" -- that is, by both grouping similar vectors and then prioritizing the ones which have the most impact. My favorite line is at the end: *"[What we found is] the more you step back and embrace complexity, the better chance you have of finding simple answers -- and it's often different than the simple answer you started with. So, for any problem, the more you can zoom out and embrace complexity, the better chance you have of zooming in on the simple details that matter most."*
ruclips.net/video/UB2iYzKeej8/видео.htmlsi=kG_Vu6CsEG2sthv0
Spaghetti in a Nutshell! 😂😂😂
I appreciate the 40 seconds laugh
When Jonathan started laughing manically, I thought of a crossover with Terry from TempleOS for some reason. I'd think of Terry's voice asking "is that programming *slur word*" and "they glow in the dark". There is something about being the outcast criticising the whole software industry on how they do stuff and you doing things in your own way while laughing hysterically on how things have become.
no wonder I cannot learn programming, someone make it so complicated? I need to learn how to be natural programmer. Maybe Mr Jonathan Blow can help us become natural programmer and make everything simple.
I had to explain to my dog that it’s water is ok because he said dog water at 3
LSP has better documentation
The best Jon's video EVER 😂
Looks like the documentation graph for SAP.... not even joking.
no way! it's LLVM?! 😂😂😂
Because I am an idiot, I’m not sure what I’m looking at: is that a chart of every function call or something? Is that something that can be generated for any language or is that something special to whatever language he’s looking at? Could someone please explain to my dumb ass what I’m looking at 😂
2:15 "You do not understand the whatever whatever factorialization of the subclass..."
eminem been real quiet since this one dropped
It's funny to read the lofty academic debates down here when it's clear most people (myself included) are unsure what exactly Jon is criticising here. The automatic diagram generation? The complexity? The imports? That some dude actually thinks programming is doodling a line with a marker around some boxes? Is this complexity even relevant to anyone? Who knows, but I'll just interpret it however I like it the most and plaster my rant here
Love it or hate it this is how young people are being exposed to programming (i hate it). Programming used to be done by punch cards. It's one positive trait is that it is a decent way to explore concepts with minimal syntax with immediate visual cues. Whether or not this type of programming has merit or not is irrelevant; Young people are using these tools.