Introduction

What ?

Brain is like a draft notebook. A place to store all my notes.

Why ?

While browsering dev.to, I found this article :

Takes Notes On Everything

Then I decide to give this concept a shot.

Licence

You can do whatever the fuck you want with this.

        DO WHAT THE FUCK YOU WANT TO PUBLIC LICENSE
                Version 2, December 2004

Copyright 2017 papey

Everyone is permitted to copy and distribute verbatim or modified
copies of this license document, and changing it is allowed as long
as the name is changed.

        DO WHAT THE FUCK YOU WANT TO PUBLIC LICENSE
TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION

 0. You just DO WHAT THE FUCK YOU WANT TO.

Dev

Golang

http package stuff

HTTP Status Code

Do not use number directly.

Don't do this

w.WriteHeader(404)

Do this

w.WriteHeader(http.StatusNotFound)

Error message

http.Error(w, http.StatusText(http.StatusUnauthorized), http.StatusUnauthorized)

Flusher

A Flusher can be used to allow an HTTP handler to flush buffered data to the client.

package main

import (
    "fmt"
    "io"
    "net/http"
    "time"
)

func main() {
    http.HandleFunc("/", func(w http.ResponseWriter, r *http.Request) {
        ended := make(chan bool)
        f, flushable := w.(http.Flusher)
        if flushable {
            go func() {
                ticker := time.NewTicker(1 * time.Second)
                for {
                    select {
                    case <-ticker.C:
                        fmt.Println("debug")
                        io.WriteString(w, "#")
                        f.Flush()
                    case <-ended:
                        return
                    case <-r.Context().Done():
                        return
                    }
                }
            }()
        }
        time.Sleep(5 * time.Second)
        ended <- true
        w.Write([]byte("ok"))
    })

    http.ListenAndServe(":8080", nil)
}

Working with packages

In go, every project, is, at some part, a package.

A package is simply a directory containing go source file.

Organising projects

In your project, you may need multiple, isolated packages. Here is the most common project layout in found and I used in all my projects :

layout -- top project directory
├── build -- all things related to ci
│   └── ci
│       └── ci.yml
├── cmd -- main package, entrypoing
│   └── layout.go
├── Dockerfile -- container build instructions
├── init -- if usefull systemd service file
│   └── layout.service
├── internal -- private packages, ie not to be used in another project
│   └── private
│       └── private.go
├── LICENSE -- license file
├── pkg -- public packages, ie packages like libs, to be importer in another project
│   └── public
│       └── public.go
└── README.md -- readme

Using go mod

Init project

go mod init github.com/username/project

Add an external dep to your project

Import the required package

import "github.com/example/lib"

The next time you will run a go command it will fetch the package, eg :

go build

In repository packages

Of course, you can also use isolated package you write inside your repository (usefull if you need to split things up)

Let's say, for example, I want to add a new public package foo in my project,

mkdir -p pkg/foo && echo "package foo" > pkg/foo/foo.go

After adding stuff in your foo.go file, import it using

import "github.com/username/project/foo"

Testing tips, golang edition

Documentation

Files

Test file should be place in the package directory and should be name using the following convention :

repo : github.com/go/pkg
package : github.com/go/pkg/example
package file : example/pkg.go
test file : exemple/pkg_test.go

Run

go test github.com/go/pkg/package

fmt.Println is not working

Gniagniagnia, use

t.Log()

t.Logf()

also,

go test github.com/go/pkg/package -v

How to fail

Mark test as failed (next tests executed)

t.Fail()

Mark test as failed AND exit

t.FailNow()

Print and mark test as failed

t.Error()

t.Errorf()

Print, mark test as failed AND exit

t.Fatal()

t.Fatalf()

I don't want my tests to be messy (kudos @athoune)

Assert

Shit, i want to disable my oauth test on my CI

if os.Getenv("DRONE") == "true" {
	t.Skip("Skipping test in CI environment")
}

Short mode

This test is too long ? Skip it !

if testing.Short() {
    t.Skip("Skipping test in short mode")
}

go test github.com/go/pkg/package --short

Cleanup test cache

go clean -testcache

Elixir

Lamba or Anonymous Functions

Create an anonymous function and bind it to a variable

Simple one

iex> func = fn -> IO.puts("Hello") end
#Function<21.126501267/0 in :erl_eval.expr/5>
iex> func.()
Hello
:ok

One with arguments

iex> func = fn t -> IO.puts(t) end
#Function<7.126501267/1 in :erl_eval.expr/5>
iex> func.("Hello")
Hello
:ok

Another solution is the & operator used as syntastic sugar

iex> func = &(&1 + &2)
&:erlang.+/2
iex> func.(2, 2)
4

GenServer

How to create a scheduled job (kudos href)

Code

defmodule Jobs do
  # Jobs module is based on GenServer
  use GenServer

  # Init with `init` as initial value, then continue
  def init(init) do
    # call to handle_continue
    {:ok, init, {:continue, :work}}
  end

  # Exec job on continue, then reschedule
  def handle_continue(:work, state) do
    {:noreply, work_then_reschedule(state)}
  end

  # Handle info and pass it to continue
  def handle_info(:work, state) do
    {:noreply, state, {:continue, :work}}
  end

  # Get timer from config.exs
  def get_timer_config() do
    {:ok, timer} = Application.fetch_env(:app, :timer)
    timer
  end

  # Do the important stuff
  defp work_then_reschedule(state) do
    # Modify state
    state = state + 1

    IO.puts(state)
    IO.puts("Work, then reschedule !")

    # Reschedule, later
    Process.send_after(self(), :work, get_timer_config() * 1000)

    # Return updated state
    state

  end
end

Usage

iex> {:ok, pid} = GenServer.start_link(Jobs, 1)
{:ok, #PID<0.251.0>}

Dates

When working with dates in Elixir, nerver use "<" or ">", because sometimes, it will not work the way you want

iex(16)> first = Timex.parse!("Tue, 29 Oct 2019 16:00:00 +0000", "%a, %d %b %Y %H:%M:%S %z", :strftime)
#DateTime<2019-10-29 16:00:00+00:00 GMT Etc/GMT+0>
iex(17)> second = Timex.parse!("Wed, 06 Nov 2019 23:00:00 +0000", "%a, %d %b %Y %H:%M:%S %z", :strftime)
#DateTime<2019-11-06 23:00:00+00:00 GMT Etc/GMT+0>
iex(18)> first < second
false

Timex

For example, with Timex comes the compare function.

Timex.compare(first, second)
-1

And return value should be considered as follow :

-1 : first is before second
0 : first is equal to second
1 : second is before first

Rust

serde

Serde is rust create used to serialize and deserialize stuff.

For example, this can be used to deserialize toml into a dedicated struct.

Example, with toml

Here is how serde after version 1.0 (included) should be used with serialize/deserialize mechanisms.

Be careful, there is breaking changes before version 1.0 and you can find confusing docs.

`main.rs` file

// uses
use serde::Deserialize;

#[derive(Deserialize)]
struct Config {
    name: String,
    url: String,
}

fn main() {
    let content = r#"name = "example"
url = "https://example.com""#;

    let conf: Config = toml::from_str(content).unwrap();

    println!("name: {}, url: {}", conf.name, conf.url)

}

`Cargo.toml` file

[package]
name = "safiste"
version = "0.1.0"
authors = ["El Famoso Safiste"]
edition = "2018"

# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html

[dependencies]
serde = { version = "1.0.94", features = ["derive"] }
toml = "0.5.1"

Output

name: example, url: https://example.com

Testing

`Cargo test` and `println!`

To see stdout when using cargo run (kudos CobaltVelvet)

cargo test -- --nocapture

Run all the tests, even if one of them fail

cargo test -- --no-fail-fast

Ruby

Playing with arrays

Check if all array elements are equals

arr = [0, 0, 0, 0, 0, 0, 0]
puts true if arr.uniq.length == 1

Array includes a sub array

(subarray & array) == array

Compute values occurences

[1, 1, 2, 3, 4, 5, 5, 5, 5].tally
=> {1=>2, 2=>1, 3=>1, 4=>1, 5=>4}

Reduce vs Each With Object

Both methods serve the same purpose : from an Enumerator create a single value out of it

Reduce

Reduce is prefered when you need to produce a simple value because it reduce over the returned value

[1, 2, 3].reduce(:+)
> 3

We can illustrate this simply with this snippet :

[1, 2, 3].reduce do |acc, v|
  acc += v
  0
end
> 0

Each With Object

Each with object is prefered when you reduce on a hash or some sort of complex object beacause is use the accumulator value and not the returned one

[1, 2, 3].each_with_object({ sum: 0 }) do |v, acc|
  acc[:sum] += v
  0
end
> {sum: 6}

Fun fact, reduce takes |acc, v| while each_with_object take |v, acc|

Interfaces in TypeScript

Handbook

Overview

Interface in TypeScripts are used to create a blueprint or an abstraction of a class.

An interface can contains properties (mandatory, optional or readonly)

interface Example {
  mandatory: string;
  optional?: string;
  readonly ro: string;
}

And methods

interface Example {
  mandatory: string;
  optional?: string;

  concat(sep: string): string;
}

To use the interface, a class should implement it,

class Impl implements Example {
  mandatory: string;
  optional?: string;

  constructor(m: string, o: string) {
    this.mandatory = m;
    this.optional = o;
  }

  contact(sep: string): string {
    return `${this.mandatory}${sep}${this.optional}`;
  }
}

The Promise case

What about a promise based method ? Do not forget what async/await does. It wraps all the things in Promise<T>.

In order to add promise to your interface, just specify a Promise as return type.

interface Example {
  mandatory: string;
  optional?: string;

  promize(arg: string): Promise<string>;
}

Decorators in TypeScript

Decorators are an experimental feature of TypeScript (and also a JavaScript stage 2 feature, meaning they will be soon included in standard JS) allowing you to inject specific behaviors to classes, properties, methods, accessors or parameters.

This features allow some kind of meta-programming and dependency injection, called at runtime.

This is mainly used in libs to add specific behaviors to your own code.

For example, TypeORM, an ORM lib, use this feature to give a nice way for users to annotate their models, the dedicated char to use decorator is @ :

@Entity()
export class Person {

  @PrimaryGeneratedColumn()
  id: number;

  @Column()
  lastname: string;

}

Example

This features needs to be explicitly set as enabled in your tsconfig.json :

    "experimentalDecorators": true

Decorators are just functions, for example, here is a property decorator

class Decorator {

    // Call to the emoji decorator with a string passed as argument
    @emoji("🦍")
    name: string = "";

    constructor(name: string) {
        this.name = name;
    }

}

// The actual decorator code, this is, in fact a decorator factory
// It's a high order function returning the actual decorator
// It's a common and nice way to have access to a larger scope to
// play with the args passed as params (emojo, here)
function emoji(emojo: string) {
    // Return the actual decorator
    return function (target: any, key: string) {

        // get the actual value
        let val = target[key];

        // customize getter
        const getter = () => {
            return val;
        }

        // and setter, to add some nice emojos
        const setter = (next: string) => {
            val = ${emojo} ${next} ${emojo};
        }

        // Apply thoose changes to the actual object property
        Object.defineProperty(target, key, {
            get: getter,
            set: setter,
            enumerable: true,
            configurable: true
        });
    }
}

const example = new Decorator("hello");

console.log(example.name);

Even if this example is quite useless, it gives an overview of the possibilities of this feature.

More ressources can be found in the Handbook

js

Presentation of the main ES6 features

Spread Operator

It allows an iterable to expand items in it.

array = [1, 2];
appended = [3,4];

all = [...aray, ...appended]
console.log(all)

This can be usefull in some situations like this one :

array = [1, 2, 3, 4];
console.log(Math.min(...array));

instead of the old syntax

array = [1, 2, 3, 4];
console.log(Math.min.apply(null, array));

Rest Operator

Represent infinite number or args in an array

const sum = (...numbers) => {
    return numbers.reduce((previous, current) => {
        return previous + current;
    });
};

Arrow Function

Function can now be written like this

const f = (a, b) => a + b;
console.log(f(1,2));

instead of

function f(a, b) {
    return a + b;
}
console.log(f(1,2));

Default parameters

With arrow syntax !

const f = (a = 10) => console.log(a);
f();
f(1);

Destructuring (arrays and objects)

With arrays

const array = [1, 2];
const [a, b] = array;
console.log(a, b)

or objects

const person = { name: "Jesus", age: "33" };
const { name, age } = person;
console.log(name, age);

Web

front

Babel and babel-preset-stage-2

If you want use the "Spead Operator" eg ... in javascript, you need babel-preset-stage-2.

yarn add -D babel-preset-stage-2

Then enable it your babelrc :

{
    "presets": [
        ["es2015", { "modules": false }],
        ["stage-2"]
    ]
}

JVM

Java

Defaults

Java uses int (64 bits) as default type for integer numbers

Java uses double as default type for floating point number (double precision float precision)

Regex

Non capturing group

8 cats
9 dogs
10 cows

Match and capture ONLY the number of cats and dogs using ?:, for non capturing groups

(\d+) (?:cats|dogs)

Tools

Git

Fetch a remote branch

Create a local branch that tracks a remote branch

git checkout --track origin/develop

If you want to change the name of the local branch (NOT A GOOD IDEA)

git checkout --track -b gniagnia origin/develop

Amend without editing commit message

git commit --amend --no-edit

Copy current HEAD sha sum to clipboard

git rev-parse HEAD | xclip

Conditional Git configuration

Since git 2.13, there is an option to include configuration (or not) based on a condition

Here is an example use case

# Specific configurations for work and personnal repos
[includeIf "gitdir:~/code/work/"]
    path = .gitconfig.work
[includeIf "gitdir:~/code/oss/"]
    path = .gitconfig.oss

If include, settings in any configuration found file provided by the path parameter will by append to or will overide current configuration.

Switch to previous branch

git checkout -

SVN

Commit only selected files

svn commit -m "Message" README src/example/script.pl

Ops

k8s

kubectl

kubectl is the main command line tool for k8s. See reference

Raw commands

Sometimes it's handy to just run a container without all the yaml mess.

Running a container

kubectl run --image=IMAGE name

kubectl run nginx --image=nginx

Create a deployement

kubectl create deployment --image=nginx nginx

Gotcha, deployement do not have a --replicas option, you need to use kubctl scale after deployement creation

Generating a manifests, blazzing fast way

kubectl run app --image=app --dry-run=client -o yaml

kubectl create deployment --image=redis redis --dry-run=client -o yaml

Of course, if you need to write it into a file

kubectl create deployment --image=redis redis --dry-run=client -o yaml > deploy-manifest.yml

Switching namespaces contexts

kubectl config set-config $(kubectl config current-context) --namespace=target

Cluster Upgrade

Versions

vMajor.Minor.Patch semantic versioning

K8S compatibility is as follow : current and n-2 version are supported.

Always upgrade minor to minor releases.

kubeadm

Check latest/upgradable version

kubeadm upgrade plan

Master node upgrade

Drain and mark it at Unschedulable

kubectl drain master

kubectl get nodes to verify that master node is unschedulable

kubeadm update

apt install kubeadm=X.Y.Z-00

kubeadm version to ensure version number

Node update

kubeadm upgrade apply vX.Y.Z

kubectl version to validate

Update kubelet

apt install kubelet=X.Y.Z-00

kubectl get nodes to validate

Mark master node as schedulable

kubectl uncordon master

Worker nodes upgrade

Repeat the same actions done on master for each worker node

🎉

macOS

iTerm

Sync sessions

Toogle the feature

⌘ + ⇧ + i

Will toggle sync and input will be sent to all sessions

GNU/Linux

Debian

Upgrade to latest stable

Enure backups !

Ensure everything is up to date

apt update
apt upgrade
apt dist-upgrade

Check package db consistency

dpkg -C

Check hold packages

Hold packages will not be upgraded, ensure there is no deps shit

apt-mark showhold

If there is anything, check with dpkg --audit

Update source list

Check

sed s/$OLDSTABLE/$STABLE/g /etc/apt/sources.list
# for exemple
sed s/stretch/buster/g /etc/apt/sources.list

Is it's ok, go

sed -i s/stretch/buster/g /etc/apt/sources.list

Do not forget stuff in source.list.d

find /etc/apt/sources.list.d/ -type f -exec sed -i s/stretch/buster/g {} \;

Simulate

apt update
apt list --upgradable

Go

apt update
apt full-upgrade

Clean

apt autoclean

Enjoy !

Shell Tips

Find stuff eating all SWAP (kudos gardouille)

find /proc -maxdepth 2 -type f -iname status -exec grep -HR VmSwap {} \; | grep -v " 0 kB"

Tools

Curl

Add form-file in a POST request

curl -F "field-name=@/my/path" http://example.com

POST is implicit because of -F.

Ansible

List all tags

ansible-playbook -i host.targets -v site.yml --list-tags

Start at a specific task (life savior)

ansible-playbook -i host.targets -v site.yml --start-at-task "The audacious task"

List hosts

ansible-playbook -i host.targets -v site.yml --list-hosts

Limit hosts

ansible-playbook -i host.targets -v site.yml --limit hostname

Jinja2, templates & carriage return

To tell Jinja2 to not mess with carriage return in templates add

#jinja2: trim_blocks:False
---

at the top of the template file

Import vs include (kudos @href)

If you want to exec a sets of tasks when a condition is true, use import_tasks.

- name: Include init tasks
  import_tasks: init.yml
  tags:
    - init
  when: proof.stat.exists == False

ìnclude_tasks will add all tasks to play run, even if the when condition is false.

Exec task(s) if a specific service is found

- name: Is Docker running ?
  service_facts:

- name: Push Telegraf docker input config if needed
  template:
    src: inputs/input.docker.conf.j2
    dest: /etc/telegraf/telegraf.d/input.docker.conf
  notify: reload telegraf
  when: "'docker' in services"

Docker

Inspect

Get container ip

docker container inspect --format='{{range .NetworkSettings.Networks }}{{ .IPAddress }}{{ end }}' upbeat_babbage

MongoDB

Non interactive conn from CLI (kudos lukhas)

Docs says

mongo --username user --password 'passwd'

Reality says

mongo --username user --password='passwd'

OpenSSL

Checking

CSR

openssl req -text -noout -verify -in request.csr

Key

openssl rsa -in privkey.key -check

Cert

openssl x509 -in certficate.crt -text -noout

Tmux

El famoso terminal multiplexer (sorry not sorry but screen is a piece of shit).

Sync panels

Sync panels can be quite usefull to do the same tasks on a different set of servers.

`tmux.conf` file

[...]
unbind S
bind S set-windows-option synchronize-panes
[...]

How to use it

Typical use case can be updating let's say 3 Debian servers.

start tmux
split in 3, use ssh to connect to the targets¹.
use bind prefix + S to activate sync²
apt update && apt upgrade
And voilà !

a dedicated session can be used to automate stuff, more on that later...

by default, Ctrl + b

Misc

Archlinux Urgent Chroot

In case of failure, here is the savior :

loadkeys fr
cryptsetup luksOpen /dev/nvme0n1p5 luks
opts=defaults,noatime,nodiratime,ssd,compress=zstd
mount -o $opts,subvol=@ /dev/mapper/vg-arch /mnt
mount -o $opts,subvol=@/home /dev/mapper/vg-arch /mnt/home
mount -o $opts,subvol=@/var /dev/mapper/vg-arch /mnt/var
mount -o $opts,subvol=@/var/log /dev/mapper/vg-arch /mnt/var/log
mount -o $opts,subvol=@/var/cache/pacman/pkg /dev/mapper/vg-arch /mnt/var/cache/pacman/pkg
mount /dev/sdX1 /mnt/boot
arch-chroot /mnt