（评论）

（评论）
(comments)

原始链接: https://news.ycombinator.com/item?id=41614663

该文章表明，视频游戏可以利用认知偏差来操纵玩家，导致他们感受到暂时的快乐（“喜欢”途径），而不是真正欣赏游戏而获得的享受。重复这个过程，直到玩家不由自主地参与游戏。这种行为类似于吸毒成瘾，可能不会强烈地表现出喜欢（提供参考）。技术方面，游戏采用原生开发工具包（NDK），依赖Java执行。本机代码必须编译为共享对象，并使用 JavaScript 本机接口 (JNI) 与 Java 交互。为了使游戏与常规 Android 系统兼容，必须编写本机方法或使用预定义 NDK 活动中的现有方法。但是，非游戏功能可能需要额外的 JNI 工作。此外，Java 和 NDK 之间的通信可能具有挑战性，因此建议处理 Android 进程间通信 (IPC)，而不是直接处理 JNI。与所讨论的游戏相比，分析的应用程序仅包含由黄色球、绿色矩形组成的简单设计，并且缺少自定义字体、纹理背景、跟踪分数和用户界面增强等元素。总体而言，与引用的游戏相比，分析的应用程序显得不太复杂。

Really wish the app store had a "only apps under 10MB" filter.

The fastest, least ad-filled and micropayment filled apps are usually the small ones. By downloading a 3 megabyte thermometer app you'll be much happier than a 150 megabyte thermometer app.

I remember there was a publisher in Play Store who had very small apps like single digit kb flashlight, sudoku, calender, etc. I can't find them now. Those apps were really small all within <200kb

This is something that really bothered me - I had an app that was small and worked fine on the latest Android OS, yet they took the app and account down because we hadn’t uploaded a new version in a year. Appeals didn’t help

3P app developers are also complicit. Often they deliberately cut off support for old OS's and old devices, because it's "too hard" to support them or whatever. Everyone seems to be working together to keep us on the hamster wheel.

Granted, it is hard. It's a whole extra version to QA on. If it works fine, fine, but if there are consistent negative user reviews on a version with

We don't support old iOS versions at all. We can't source new devices on old iOS versions so we can't reliably develop or test on them.

There may not have been any. Individual app-store reviewers can block you any time they feel like it, the guy checking your appeal is the same, and none of them have any real pressure to behave unless you have money and corporate power behind you.

I'm no fan of Google, but it's slightly more complicated than that, there's a lot of security and privacy stuff that can't be enforced if your app was build 6 years ago and still slopping around.

Does that really matter for a local-only 5KB app that only talks with my phone‘s flashlight, or reads sensor data? Now, maybe for the 500MB adware-filled “flashlight” app that connects to 100s of servers and demands access to everything my device can do, but that would be banned on any competent app store anyway.

I don't know if this is still the case, but at one point the permission needed to access the flashlight also gave access to the camera. And there aren't restrictions on network connections from apps. (I'd love to have app network access restricted by permissions, but that would be a large change.)

And in any case, Android has had built-in flashlight support for a while now, for any phone that has a camera with a flash. Is the "turn the screen bright white" style still useful with modern Android?

I use the minesweeper, sudoku and solitaire apps from dustland design (search pub:Dustland Design) they're very minimalistic and clean.

There's also currency / unit converter and calendar by Sam Ruston which are in the same vein very good and clean.

I can't even get angry at that, because it seems like just another popular business pattern:

* build a nice product

* become popular and gain trust with customers

* sell the company to a scammer

* profit!

> I’d do that in a heartbeat.

> My customers can follow me to my next project.

If you are willing to sell your customers to an ad firm, why should they trust your next project?

Weird that their GitHub says "without ads", but the apps in the play store say contains ads. It looks like they're doing ads/paid model in the app store, are they ad-free from F-Droid?

For reference the first app I got on Apple App store (ignoring the ad result) for "thermometer" is > 100 MB. Looking at the first ~dozen only 1 comes in under the <10 MB category. The two biggest offenders of huge app sizes are shipping cross platform runtimes (the kind that tend to throw in the kitchen sink, not the kind that act as a thin layer) and tracking/analytics bloat.

That reminds me of one reason I got out of mobile app development, totally forgot about until now

Often times the hiring managers wanted to see something more akin to a portfolio, like an art project, for apps that many times didn’t exist anymore or have a production server up anymore

But the more arbitrary metric was trying to be sure that I worked on anything “big”

And the 8-12 megabyte package sizes - which I spent a lot of time optimizing with many competence inspiring techniques - would signal that the app or service or userbase wasn't big. Which had nothing to do with anything, could have hundreds of millions of downloads and users

In that space there is a huuuge incentive for bloatware

I continue to be puzzled by how much smaller apps are on Android, ex. Took me 9 tries, including ads, to find a thermometer app over 7 MB. I've worked on both platforms for years and yet don't really know why. Only guess is Android has a much richer tradition of vector art over bitmaps, and Swift libraries had to be compiled in for years until ABI stability enabled using dynamic linking to OS ones

Because a thermometer is software and software is imperfect. Perhaps it made some assumptions that causes phones that were released after the app was released to drain the battery very quickly. Or it has a calculation error where over time it accumulates a significant difference between the measurement data and the data that is rendered on screen. Or perhaps it's using an API that we all thought was safe, but turns out it's not. Or it needs to use an API to get temperature data (thermometer can have different meanings) and the API no longer exists.

Even something as silly as an app that does nothing can run into these issues. The APIs and other interfaces used to run applications are imperfect. Sometimes doing nothing about it is a choice, sometimes the vendor doesn't deem that acceptable and then it is no longer a choice. Either way, the application will have to adapt or degrade (to the point where it degrades out of existence).

A calculator app doesn't need that many megabytes of code and assets to be a calculator app. So if an app is way bigger than it should be, it usually means one of two things (usually!):

1. The app was not very optimised, perhaps created by a novice, containing a lot of things it doesn't need.

2. The app used to be really small, but a lot of extra code was added to serve you ads, profile you for better targeting or do sneaky stuff you didn't ask for.

If a trip to the baker took 172 days, there would be over 171 used days to justify; if it took 172 engineers to change a lightbulb, it would have to be a very special lightbulb or explanations should be in order. Besides uses of concern of the extra resources spent, it simply just makes no sense.

i would say 'whatever is in the extra 171.97 megabytes'

i wrote a calculator app including its own implementation of decimal floating point and it's still only 20 kilobytes

People arrived to this shore after having experienced that what they considered unbelievable and untenable is actually believed and held by some - who may not even seem to be particularly an uncommon tail of an emerging population.

And this is why a good '/S' keeps you safe from misunderstanding.

I have always felt there is something fascinating going on behind Flappy Birds' infamous difficulty curve that warrants deeper study.

On the one hand, there is no actual progression or ramping of difficulty in the game itself. The difficulty level remains the same whether your current score is 0 or 10 or 100. But every new highscore represents a new summit that the player has to scale. The first and maybe the most frustrating summit to scale is scoring a single point. To get your score into the double digits, the player has to have basic mastery of the core mechanics - including the precise physics, and timings- and learn how to handle a certain number of scenarios. The obstacles on the path to triple-digit territory and beyond seem almost self-imposed. The fear and tension as you approach your own highscore is the biggest impediment to breaking your highscore. Once you break that highscore - the hand tremors magically disappear the next time you approach it, only for it to re-appear as you near your new highscore.

All this, when the basic concept of the gameplay is deceptively simple. Like i said, there are many layers to unpack for someone who is willing to look into it.

Look at almost ANY arcade game of the 70s/80s. Most of them weren't designed with an end in mind. You play as long as you can for the accolades that come with posting your initials on the scoreboard.

Asteroids is a quintessential example - relatively flat difficulty curve once you've mastered the game - it really comes down to a test of the player's endurance. Scott Safran set a record game that lasted a grueling 60 hours.

https://en.wikipedia.org/wiki/Scott_Safran

EDIT: Anyone who has EVER tried for a high score (whether a personal best, or a world record) is familiar with the natural nervousness that increases in direct proportion to how close you are to breaking it. That's not a Flappy Bird thing, that's a literal every game thing. Go watch a live stream of a speed runner that's got a heart rate monitor attached to the feed for example.

Many early 80s 8-bit console games worked this way. I'm thinking of games like Transbot for the Sega Master System. There are a few different levels, with various enemy configurations and scenery, but it has no ending and goes on forever without any real changes.

This was the case for me with Flappy Bird and also that old Temple Run game circa 2011-ish.

It's almost more of a game of focus or how much you will _think_, because once you're distracted a bit and forget those physics or timing just one time, you're probably done.

"It is surmised that conditioning is enforced via several cogni- tive biases that trick a player into expecting euphoria (liking- pathway), when instead frustration is yielded – with condi- tioning being iterated to a point that the player is motivated to interact with the game on a foremost instinctual level. We posit that these stimulations of the wanting-pathway may lead to players interacting with the game not only with- out actually liking it, but also without knowing why they are interacting with the game. Indeed, this calls for drawing another parallel between drug addiction, and play behaviour in which liking may be barely exhibited (cf. [16, 38, 40])."

I did something similar in Nim and published it in 2020 (less pretty graphics however). The difference is that I went deeper and actually wrote an assembler for the Dalvik bytecode and .apk files:

https://www.youtube.com/watch?v=wr9X5NCwPlI&list=PLxLdEZg8DR...

The code in the repo has unfortunately bitrotten. I am sometimes thinking to try and resurrect it Some Day™... from time to time I think of some simple app I could write if it was a bit more polished.

Less than 4k loc.

   457 android_native_app_glue.c
   360 audio.c
   802 game.c
   201 init.c
    93 main.c
    39 mouse.c
    38 shaders.c
   229 texture.c
  1377 upng.c
    27 utils.c
  3623 total

A student of mine had an assignment to write a game using SFML, they wrote a FlappyBird clone and it was like a few hundred lines of code. It's not a very complex program to write. To be honest, I think 4k is too much :)

game.c is 800 odd lines. There are some optimizations you could do here and there (e.g load digit sprites in an array to avoid the switch case 1/2/3... stuff).

The bulk of the 3000 is fluff that you need because this is C on Android, not SFML.

It compiles to 37kb of 32bit code and to 48kb of 64bit code.

  /lib/arm64-v8a/libflappybird.so    48kb
  /lib/armeabi-v7a/libflappybird.so  37kb
  assets:     29kb
  icon:        3kb
  signature:  12kb

Plus the manifest (2kb) and resources.arsc (0.5kb)

It was also made to work on exactly one hardware specification, with no operating system to speak of. This flappy bird clone works on an immeasurable number of devices, with varying hardware AND software configurations!

Well these days cosmocc -mtiny is more like 120kb, now that our binaries support ARM platforms too (like Android!) and all the code I've needed to add to make sure Cosmo works reliably for a longer tail of edge cases. But that's saying a lot, since unlike these APKs cosmo binaries don't need a gigabyte JDK to run.

Also even the "pure" C one depends on Java, because Activities exist only on the Java side of Android.

In practice to be usable on a standard Android system, native code must always be compiled to a shared object, with JNI entry points to be called from Java userspace.

The only option is to write such native methods ourselves, or use one of the two predefined Activities for NDK that already expect specific functions to be present on the shared library.

Additionally, the zero Java part only works, if what NDK exposes as stable API is enough, and from Google's point of view, that is only for games, or faster compute, everything else requires JNI fun.

As tip, it is easier to deal with Android IPC for Java <-> NDK communication, than going through JNI boilerplate.

Couldn't one simply make the boilerplate once, as a library, that takes the pertinent bits as arguments? In which case if your app is C anyways it would make sense to just keep it simple with that.

That came to mind for me too, but unfortunately the name has already been taken by a few clones; the most notable being a trivial reskin that uses a floppy disk icon instead of a bird.

Could this technique, using rawdrawandroid to write C applications for Android, also use raylib (and other C frameworks)?

And maybe could this developing system be used through Termux, to have a C development environment on Android for Android?

It is refreshing, and nice to see programs/games/apps that are "crafted" vs just slapped together out of existing, bloated third party components.

I think it can be optimized quite a lot by not using a stock PNG decoder library, because all images are quite simple and can be generated from non-pixelated smaller sprites (many images are pre-scaled by 2x, which can be done during the postprocessing) or from a simple algorithmic code.

Weirdly, I think the challenge would be more difficult going to Android than adding graphics while keeping the size down.

It would not at all surprise me to see a near perfect Flappy Bird under 4k (graphics and all) as a PC .com

I'd be curious to see what the minimum size of a simple C program would be. Say something that displayed a pixel that bounced up and down as you tapped.

Do you still accept more optimizations? :-) I believe there are tons of mechanical substitutions that can be made there, for example `i%17?r+=z:r+='|\n|'+z` should simplify into `r+=i%17?z:'|\n|'+z`.

Great work! Good to see what only it takes to run on Android! On the other hand it also shows how much comes "for free" or made easier by using the provided sdks. For example volume control doesn't work while running this. Also resuming the game after switching away. Maybe that's relatively easy to save and restore state, though.

I'm not sure if this is a joke or not, but there's no actual graphics in this besides a yellow ball and green rectangles, while the OP game has actual textured pipes, a textured floor, a background scene. The score counter looks to use a basic system or browser font, while the OP game has a custom font. The OP game has also high score tracking, sharing, a proper main screen instead of just dropping you into gameplay.

Most Android apps are huge because they bundle tons of assets just to accommodate the “initial experience of the user”. Also, using bloat libraries and frameworks (any shipped by Google), increase the apk size.

Nowadays Google offers a solution for this problem called app bundling. It’s especially good if you build a mono app that behaves differently in certain regions. Instead of delivering a raw apk, you deliver a region specific app bundle.

I'm unaware of any apps that behave totally differently in different regions.

Sure - there are sometimes a few disabled features in one region or another, but is that really worth shipping a totally different binary for?

Even language packs can be tiny even for 200+ languages if they're pure text.

It's only when you get language/region specific artwork that there's a problem.

I made a sub-100k Android app once (I am now banned from the Play Store, and I should be lucky they didn't delete my Gmail account too) and every time I opened the IDE (Android Studio at the time) it would automatically add a Google "support library" to the project that Google obviously wanted to force me to use. If I forgot to remove it and built the app, it would be closer to 10MB. So that was the minimum size of almost every Android app at the time.

On a technicality, yes, but I think most developers who use it do so because of the large collection of React-style user interface components which are built into it. Now that Jetpack Compose exists (which allows for traditional Java-based widgets to be used in a 'reactive' way), personally I don't see any reason to use Flutter. That 'drawing directly' aspect puts Flutter apps several years behind in terms of performance, accessibility and reliability, since pretty much everything in the Android API needs to be reimplemented in Dart.

Interesting to see what a Windows-based project looks like. I haven't used Windows for ages. Seeing the .bat files and vsproj files gave me nostalgic feelings.

i realised recently, there is a correlation between the file size of a game and how likely i am to enjoy it.

the smaller the file size the more likely i am to enjoy it. and the opposite is true.

i think part of it is time investment. having less time. i dont see much value in 60gb of 4k graphics textures.

pac man on the atari or snes is maybe less than 100kb, while modern pac man could be easily 10gb or more. same for tetris or any game with the same gameplay that hasn't changed much.

（评论） (comments)

（评论）
(comments)