I don't want to come off as a linux-elitist but it's simply amazing how much easier my job is on linux. A good example recently was setting up some libraries for a C++ program I was writing to show to my class. Most of them were using Windows and visual studio, took about 15 minutes to download all the headers and libs, and show them how to configure a VS solution to link them. Not too big a deal but on linux, it only took about 30 seconds to pacman and gcc -l the lib. Little things like that keep me interested in linux as a dev tool.

When i started ocaml. Now most languages use
import lib;
But not ocaml. No, it uses
open lib;
So i had the genius idea of writing my own lib and submitting my program as

open Bob;
let () =
show "vegana";;

So I found this consulting job a while ago thinking that some extra cash while studying would be nice to have.

I meet with the guy, a researcher trying to start a business up, good for him I think, maybe we'll hit it off, continue working, why not? Except he has no clue how to write working code, all he ever did was writing matlab scripts he says, thats why he hired me he says.
Okay, fine, you do your job I do mine.

He hands me the contract, its about comparing two libraries, finding out which one is better suited for his job, cool, plots and graphs everywhere.

Except this is an unpaid job. YOU WHAT?! It's a test job. FINE. At least it'll look good on my resume.

We talk about the paid part where I'm supposed to scale the two libraries, looks good, as expected from an ML engineering perspective. It comes to payment. The dude has no idea how taxes work, says he has a set amount to pay and not a penny more. I explain with examples how taxes are paid, how you get reimbursed for them and so on. Won't budge. Screws me over.

Opens the door for other jobs I think, he'll learn next time I think and take the job.

Fast forward a month, 90% of the job done, he adds a third thing to compare. Gives a github link to a repo with 2 authors, last commit a year ago. There are links to a 404, claiming compiled jars. Fuck.

Not my first rodeo, git clone that shit, make compile, the works. The thing uses libs that ain't in no repo, that would be too easy. Run, error, find lib, remake all the things, rinse repeat.

The scripts they got have hardcoded paths and filenames for 2 year old binaries, remake that shit.

It works, at least I get a prompt now. Try the example files they got, no luck, some missing unlinked binary somewhere, but not a name mentioned. Cross reference the shit outta the libs mentioned on readme, find the missing shit, down it.

Available versions are too new, THE MOLDING NUTCRACKER uses some bug in an old version of the lib.

I give up. Fuck this. This ain't worth the money OR time. Wanker...

Fuck jQuery. The only reason I see anyone using it legitemately is because of backwards compatibility. Almost every jq method is either native js or native css. The problem is, some devs become practically dependent on a library. By then, they are no longer js devs. They are jQuery devs. When you find yourself going to the docs of a lib before native methods 9 times out of 10 you've gone past the turning point. When you find yourself including jQuery instinctively, you're gone. StackOverflow is a great example of this:

Question - 1 up
Pure JS answer - 0 ups
jQuery answer (same length) - 2 ups and accepted

Come on man. It's 2018! We shouldn't be writing jQuery anymore. Native methods ftw!

I'm finally writing unit tests consistently thanks to a simple file organization decision.

I'm not doing pure TDD, but at least I'm writing the tests immediately after writing a module, and I make sure they run ok.

What I'm doing is Instead of putting the test files in a "tests" dir at the root of the project, I have the tests right next to the source code.

So if I have a dog.x file, I also have a dog.test.x file next to it.

I'm not inventing gunpowder here. I've seen several people do this.
But it's something that is not generally made a default or advised to do.
Like I said; test frameworks in general go with the classic "tests" dir.

But for me this is day and night in whether I write the tests or not.

Which makes sense. Imagine the classic scenario of the "tests" dir, and you just created a file deep into a hierarchy, let's say src/lib/console/windows/dog.x

This means that if you want to write tests for that, you need to make sure the hierarchy tests/lib/console/windows/dog.test.x exists

If the test file already exists, but you want to access both files, you need to traverse deep for each.

Also, it's actually harder to keep track which files have unit tests and which do not.

Meanwhile, if the test files are next to the source, all these problems disappear.

That doesn't mean there are no other challenges with testing, like testing untestable things, like system calls or http requests, but there are ways to deal with that.

Sorry, need to vent.

In my current project I'm using two main libraries [slack client and k8s client], both official. And they both suck!

Okay, okay, their code doesn't really suck [apart from k8s severely violating Liskov's principle!]. The sucky part is not really their fault. It's the commonly used 3rd-party library that's fucked up.

Okhttp3

yeah yeah, here come all the booos. Let them all out.

1. In websockets it hard-caps frame size to 16mb w/o an ability to change it. So.. Forget about unchunked file transfers there... What's even worse - they close the websocket if the frame size exceeds that limit. Yep, instead of failing to send it kills the conn.

2. In websockets they are writing data completely async. Without any control handles.. No clue when the write starts, completes or fails. No callbacks, no promises, no nothing other feedback
3. In http requests they are splitting my request into multiple buffers. This fucks up the slack cluent, as I cannot post messages over 4050 chars in size . Thanks to the okhttp these long texts get split into multiple messages. Which effectively fucks up formatting [bold, italic, codeblocks, links,...], as the formatted blocks get torn apart. [didn't investigate this deeper: it's friday evening and it's kotlin, not java, so I saved myself from the trouble of parsing yet unknown syntax]

yes, okhttp is probably a good library for the most of it. Yes, people like it, but hell, these corner cases and weird design decisions drive me mad!

And it's not like I could swap it with anynother lib.. I don't depend on it -- other libs I need do!

Writing simple terminal input/output lang (Hello, what's your name, hi) in D.
Compiling in D using standard lib: 6.1M
Compiling in D with inline assembly, no standard lib or libc, using syscalls: 1048 bytes!
I'm still freaking out a little.

What I'm doing now, writing a JS library for a simple kitchen timer (like, something that can be wound up, is ticking, can be paused, etc). Here's a list of neat stuff I've learned:

Polyfilling as a lib author (I decided against it).
Packaging the lib (using Rollup, ES6 modules are totes cool).
Using flow to add static typing in strategic places (started appreciating types in JS since reading up on functional programming).
Modelling state and transitions using an explicit state machine. (Fucking finally. There's usually an implicit state machine somewhere, only spread out all over the app...)
Using mostly side-effect free methods, being very explicit about when and why things are mutated).
Test-first/TDD (ish) using Jest and the awesome Wallabyjs.
Freeing up mental capacity by letting Prettier format my code for me (it was hard to let go but totally worth it).
Started using git.

Did all work on Ubuntu after pretty much a lifetime of Windows (initially to separate work from gaming) and finally swapped MS Visual Studio for Atom.

When it's finished I'm going to publish it on GitHub, which will also be a first for me. Might try out some CI platform while I'm at it.

tl;dr: wrote some js, felt good

Context: New to typescript. Writing a thing, doing it for work, good opportunity to stretch my dev legs. Using a propriety lib, alternatives not an option.

Rant begin:
SOOOO, who the fuck thought THIS was a good idea:
1. Lib has minified react in dev (because closed source) meaning no downstream errors AND the entire premise of the lib is that a widget is a react component, so I'm writing typescript react the entire time without downstream errors
2. SHIT docs. By that, I mean there's an API reference page that's so sparse there's literally a set of CRUCIAL interfaces that only say the word 'Interface' on them. That's it. that's what i get. It's an interface. NO FUCKING SHIT SHERLOCK, what the fuck is it though? What's its purpose? Is it an interface for a dog? A dog that has a 'shit' property? or a cat? or a cat eating dog shit? Nobody fucking knows - the docs sure as fuck don't care.
3. No syntax highlighting - editors, IDEs (i've tried a few) can't even find the lib inside this environment, so Code and everything else thinks I'm importing shit that doesn't even exist - so no error prediction, code completion based on syntax of the library, none of that.
4. There are some EXTREMELY basic samples - these samples exclusively use React classes - no function components, no hooks, nada - just classes and even perfect replicas of the sample code display erratic behavior like errors about missing props, so that's mostly FUCKING USELESS
5. And this... this is where the straw breaks the fucking camel's back... there's no... there's no hot reloading... Do you know what that (in conjunction with the previous 4 fuckups) means?

When I write anything or I fuck up (which of course I'm doing every time I write half a line because how the fuck?) I have to restart the client and server EVERY FUCKING TIME and manually test to see if the error (THAT ONLY GETS REPORTED IN THE LOCAL UI) is gone or different.
Then, once I see the error, it isn't an error: it's the minified React error-decoder link and guess what? It isn't really clickable a link OR copyable, meaning that every FUCKING time I get a new error, I have to MANUALLY TYPE A FUCKING 50 CHAR URL TO FIND OUT A GENERIC REACT ERROR MESSAGE WITHOUT A LINE NUMBER OR ANY FUCKING CONTEXT. I HAVE TO DO THIS CONSTANTLY TO SEE IF ANYTHING I'M DOING EVEN WORKS.

6. There's no github to complain to the maintainers or search for issues because it's NOT FUCKING OPEN SOURCE so there is literally nothing to be fucking done about it.

This is due in a week and a half, found out about it last Friday. How's your day going?

PS: good to be back after a long respite from dev ranting.

Is writing hand-optimized SIMD code even still worth it? Thinking about writing my own little math library for my game engine but I've tried writing a hand-optimized `dot(normalize(b - a), foo) >= bar` and somehow it's actually slower than writing the same thing using a math lib which is implemented exclusively with scalar math and auto vectorized by llvm

LLVM... I kneel

So I thought to myself.
Hey I'll go ahead and use python, it will make this easier than using c++.

So I start looking at python.
And I start looking at specific common functions that c/c++ and .net all offer.

Like writing a fucking png image.
And I start seeing 3rd party libs that are at version 0.2

And so I say, this is supposedly the language data people love. which would include searching gis data too right ?

Everybody touts this level for ai and machine learning and all this other bullshit but I can't even create a fucking image ? And every document points to this same lib where it comes to creating this image ? at version 0.2 ?? 20 years or more after PNG was created ?

So I look up geotiff, and see 0.4........ so..... what is this language good for again ? I can parse json in javascript and do the other things I want...

Oh scatterplot generation ? What is it being displayed in jpeg ? Maybe the jpeg implementation is good. because you know i just use scatterplots constantly. yup. most of the data I require to analyze uses scatterplots. not risk.

fun.

oh and look django.... who the fuck uses django ?

and omg it makes me format my text or the run bombs.....

jesus. rpg much ?

I'm just... I'm not seeing...
WHY ?????????

and then I have zimmermans voice buzzing in my head about just using goddamn .net

Writing a library is so much nicer than writing an app. You can concentrate on the core thing you want to achieve and need to handle just a few files.

And you can still test it live even if it's a node module: just symlink it into the node_modules folder of your app.

Opinions

Hello, I’m considering building a web framework.

My ideal features would be:
Customizable authentication system(considering using a jwt lib)
Embedded DB(bolt db)
ORM( writing my own)
REST api to DB (via code generator)
Code generator(generation of models and views via cli)
GUI to db(some admin dashboard)
CORS(web service right?)

Why?
Ease of development
Fast prototyping of small-medium web services.
Fun.

My question is, do i have to many things on my platter? Should i narrow it down into less featured framework? What feature should I focus on? How should i benchmark it? Should i write tests for absolutely everything or just for exported methods? What should i take into consideration when developing ORM API, Auth API...

The language is Go

Thank you for your input

Anyone knows a C# lib for writing IPTC Metadata to images which has a fucking dokumentation?