2017-03-11

3 things that are needed to make a successful product

In my 10 years in the software industry. I have create a number of products and have worked on a lot of projects. Looking back at the products/projects that have been successful, one this stands out. There are 3 critical pieces for a software product / startup.

0. A value proposition

This is a dead giveaway. So, I haven’t even counted this. Without a value proposition you don’t have anything. Your product must provide value to your customers.

1. Domain knowledge

You need to have someone on your team with the knowledge of the domain. Ideally you would have come up with a product idea because of a good understanding of the pain points. And of the things that can provide value. This is also fairly easy to understand.

2. Marketing Strength

You also need to have someone who can market your product. An awesome product without marketing is a dead product. You need to either build your marketing expertise or get someone who is good at it.

Marketing is one of the things that is often overlooked. People think that if the product is good people will buy it. This is completely false. You need a lot of hustle to market your product.

3. Technical expertise

You obviously need someone who can build a usable product which provides value. But this is the last on the list.

Many people come to me with ideas for startups. I always tell them about these 3 things. The next time you want to build a startup, think about these skills. Without one of these you are dead in the water.

2017-02-13

Getting started with Elm

I had a good time presenting a talk about “Getting Started with Elm” at the awesome nghyderabad The audience was very interactive and the food was great. Shout out to Fission Labs for the awesome venue!

Here are a few useful links which should help you learn Elm

elm-lang home page: http://elm-lang.org/
docs: http://elm-lang.org/docs
examples: http://elm-lang.org/examples
A handy online book: https://guide.elm-lang.org/
Online editor: http://elm-lang.org/try
Elm packages website: http://package.elm-lang.org/

Code for the app that we built: https://github.com/minhajuddin/getting-started-with-elm/blob/master/Counter.elm

2017-02-09

How to fix Ecto duplicate name migrations error

If you run into the following error while running your Ecto migrations

ReleaseTasks.migrate
** (Ecto.MigrationError) migrations can't be executed, migration name foo_bar is duplicated
   (ecto) lib/ecto/migrator.ex:259: Ecto.Migrator.ensure_no_duplication/1
   (ecto) lib/ecto/migrator.ex:235: Ecto.Migrator.migrate/4

You can fix it by running 1 migration at a time

mix ecto.migrate --step 1

This happens when you are trying to run two migrations with the same name (regardless of the timestamps). By restricting it to run 1 migration at a time you won’t run into this issue.

Ideally you should not have 2 migrations with the same name :)

2017-02-02

Lightweight xml utility to pluck elements

jq is an awesome utility for parsing and transforming json via the command line. I wanted something similar for xml. The following is a short ruby script which does a tiny tiny (did I say tiny?) bit of what jq does for xml. Hope you find it useful.

#!/usr/bin/env ruby
# Author: Khaja Minhajuddin

require 'nokogiri'
require 'parallel'
require 'etc'

if ARGV.count < 2
  puts <<-EOS
  Usage:
    xml_pluck xpath file1.xml file2.xml

  e.g.
    xml_pluck "//children/name/text()" <(echo '<?xml version="1.0"?><children><name>Zainab</name><name>Mujju</name></children>')
    # prints
    Zainab
    Mujju
  EOS
  exit -1
end

xpath = ARGV.shift
Parallel.each(ARGV, in_processes: Etc.nprocessors) do |file|
  doc = Nokogiri::XML(File.read(file))
  puts doc.xpath(xpath)
end

2017-01-24

Key transformation after decoding json in Elixir

In a previous blog post we saw how to do case insensitive retrieval from maps. A better solution for this if there are many key lookups is to transform the input by lower casing all the keys just after decoding. The solution from the blog post would iterate over each {key, value} pair till it found the desired key. However a proper map lookup doesn’t iterate over the keys but uses a hashing algorithm to get to the key’s location in constant time regardless of the size of the map.

Anyway, Here is the solution to transform each key for input JSON. Hope you find it useful :)

defmodule KeyTransformer do
  def lower_case_keys(input) do
    transform_keys(input, &String.downcase/1)
  end

  def transform_keys(input, tx_key_fun) when is_list(input) do
    Enum.map(input, fn el -> transform_keys(el, tx_key_fun) end)
  end

  def transform_keys(input, tx_key_fun) when is_map(input) do
    Enum.reduce(input, %{}, fn {k, v}, acc ->
      Map.put(acc, tx_key_fun.(k), transform_keys(v, tx_key_fun))
    end)
  end

  def transform_keys(value, _tx_key_fun), do: value
end

ExUnit.start

defmodule KeyTransformerTest do
  use ExUnit.Case
  import KeyTransformer

  test "simple map" do
    assert lower_case_keys(%{"NAME" => "Khaja"}) == %{"name" => "Khaja"}
    assert lower_case_keys(%{"NAME" => "Khaja", "Age" => 3}) == %{"name" => "Khaja", "age" => 3}
  end

  test "nested map" do
    assert lower_case_keys(%{"Mujju" => %{"NAME" => "Khaja"}}) == %{"mujju" => %{"name" => "Khaja"}}
  end

  test "deeply nested map" do
    assert lower_case_keys(%{"Children" => %{"Mujju" => %{"NAME" => "Khaja"}}}) == %{"children" => %{"mujju" => %{"name" => "Khaja"}}}
  end

  test "list of maps" do
    assert lower_case_keys([%{"NAME" => "Zainu"}]) == [%{"name" => "Zainu"}]

    assert lower_case_keys([%{
            "NAME" => "Khaja Muzaffaruddin",
            "agE" => 2,
          }, %{}]) == [%{"age" => 2, "name" => "Khaja Muzaffaruddin"}, %{}]
  end

  test "nested list of maps" do
    assert lower_case_keys(%{
                             "JUlian" => [%{"Movie" => "Madagascar"}]
                           }) == %{"julian" => [%{"movie" => "Madagascar"}]}
  end

  test "deeply nested list of maps" do
    assert lower_case_keys(%{"MovieGenres" => [%{
                             "JUlian" => [%{"Movie" => "Madagascar"}]
                           }, %{"Ho" => 33}], "OK then" => "little story"}) == %{
    "moviegenres" => [%{"julian" => [%{"movie" => "Madagascar"}]}, %{"ho" => 33}], "ok then" => "little story"
  }
  end
end

2017-01-19

How to learn vim properly

Vim is the editor of my choice, I love it a lot. I try to find vim bindings everywhere I can, A few apps which have good vim bindings

Chrome with vimium

The terminal with a proper ~/.inputrc. My ~/.inputrc below

# ~/.inputrc
#vim key bindings
set editing-mode vi
set keymap vi
# do not bell on tab-completion
set bell-style bell

set expand-tilde off
set input-meta off
set convert-meta on
set output-meta off
set horizontal-scroll-mode off
set history-preserve-point on
set mark-directories on
set mark-symlinked-directories on
set match-hidden-files off

# completion settings
set page-completions off
set completion-query-items 2000
set completion-ignore-case off
set show-all-if-ambiguous on
set show-all-if-unmodified on
set completion-prefix-display-length 10
set print-completions-horizontally off

C-n: history-search-forward
C-p: history-search-backward

#new stuff
"\C-a": history-search-forward

Once you set this up, many repls will respect these bindings. For instance irb, pry respect these. As a matter of fact any good terminal app which use the readline library will respect this.
Tmux is another software that has vim bindings

So, whenever I work with someone people always seem to be impressed that vim can do so much so simply. This is really the power of vim, vim was built for text editing and it is the best for this job. However, learning it can be quite painful and many people will abandon learning it in a few days.

There is a very popular learning curve graph about vim

Source

The part about vim is partially true, in that once it clicks everything falls into place.

Notepad is an editor which is very easy to use, but if you compare it to programming languages it has the capability of a calculator. You put your cursor in a place type stuff and that is all. Vim lets you speak to it, in an intelligent way Anyway, I am rambling at this point.

The reason I am writing this blog post in the middle of the night is because many people ask me “How should I setup vim?”, I’d love to have it look/work like yours. And many times I point them to my vimrc. However, if you are planning on learning vim, don’t go there. Start with the following ~/.vimrc


set nocompatible

" plugins
call plug#begin('~/.vim/plugged')
Plug 'tpope/vim-sensible'
Plug 'kien/ctrlp.vim'
Plug 'matchit.zip'
runtime macros/matchit.vim
call plug#end()

" Ctrlp.vim
let g:ctrlp_map = '<c-p>'
let g:ctrlp_cmd = 'CtrlP'
let g:ctrlp_working_path_mode = 'ra'
let g:ctrlp_custom_ignore = {
  \ 'dir':  '\v[\/]\.(git|hg|svn)$',
  \ 'file': '\v\.(exe|so|dll)$',
  \ 'link': 'some_bad_symbolic_links',
  \ }

That is all, no more no less.

To finish the installation, you need to do 2 things:

Run curl -fLo ~/.vim/autoload/plug.vim --create-dirs https://raw.githubusercontent.com/junegunn/vim-plug/master/plug.vim
Run vim +PlugInstall from your terminal

A few simple tips on how to learn vim properly:

Finish vimtutor on your terminal 3 to 4 times. Read everything 3 to 4 times and actually practice it.
Learn about vim movements, commands and modes
Open your vim editor at the root of the project and have just one instance open, don’t open more than one instance per project. This is very very important. I can’t stress this enough. To open another file from your project, hit Ctrl+P
Start with a simple vimrc, The one I pasted above is a good start.
Learn about buffers / windows and tabs in vim and how to navigate them.
Add 1 extension that you think might help every month. And put a few sticky notes with its shortcuts/mappings on your monitor.
Use http://vimawesome.com/ to find useful plugins.

Most important of all: Don’t use any plugin other than sensible and CtrlP for the first month

Once you learn to speak the language of vim, using other editors will make you feel dumb.

2017-01-18

A simpler way to generate an incrementing version for elixir apps

Mix has the notion of versions built into it. If you open up a mix file you’ll see a line like below:

# mix.exs
defmodule Webmonitor.Mixfile do
  use Mix.Project

  def project do
    [app: :webmonitor,
     version: "0.1.0",
# ...

If you are using Git, there is a simple way to automatically generate a meaningful semantic version. All you need to do is:

Tag a commit with a version tag, like below:

1	git tag --annotate v1.0 --message 'First production version, Yay!'

Put a helper function which can use this info with git describe to generate a version


defp app_version do
  # get git version
  {description, 0} = System.cmd("git", ~w[describe]) # => returns something like: v1.0-231-g1c7ef8b
  _git_version = String.strip(description)
                  |> String.split("-")
                  |> Enum.take(2)
                  |> Enum.join(".")
                  |> String.replace_leading("v", "")
end

Use the return value from this function as the version

# mix.exs
defmodule Webmonitor.Mixfile do
  use Mix.Project

  def project do
    [app: :webmonitor,
     version: app_version(),
# ...

The way this works is simple. From the man pages of git-describe

NAME git-describe - Describe a commit using the most recent tag reachable from it

DESCRIPTION The command finds the most recent tag that is reachable from a commit. If the tag points to the commit, then only the tag is shown. Otherwise, it suffixes the tag name with the number of additional commits on top of the tagged object and the abbreviated object name of the most recent commit.

So, if you have a tag v1.0 like above, and if you have 10 commits on top of it, git-describe will print v1.0-100-g1c7ef8b where v1.0 is the latest git tag reachable from the current commit, 100 is the number of commits since then and g1c7ef8b is the short commit hash of the current commit. We can easily transform this to 1.0.100 using the above helper function. Now, you have a nice way of automatically managing versions. The patch version is bumped whenever a commit is made, the major and minor version can be changed by creating a new tag, e.g. v1.2

This is very useful when you are using distillery for building your releases

2017-01-11

Case insensitive key retrieval from maps in Elixir

I ran into an issue with inconsistent naming of keys in one of my provider’s json. This is really bad data quality, the data that is being sent should have consistent key names. Either uppper, lower, or capitalized, but consistent. Anyway, this provider was sending data will all kinds of mixed case keys.

Here is some elixir code that I wrote to get keys using a case insensitive match. There is an issue on the Poison decoder project which should render this useless, however till that is fixed you can use the code below:

defmodule CaseInsensitiveGetIn do
  def ci_get_in(nil, _), do: nil

  def ci_get_in({_k, val}, []), do: val
  def ci_get_in({_k, val}, key), do: ci_get_in val, key

  def ci_get_in(map, [key|rest]) do
    current_level_map = Enum.find(map, &key_lookup(&1, key))
    ci_get_in current_level_map, rest
  end

  def key_lookup({k, _v}, key) when is_binary(k) do
    String.downcase(k) == String.downcase(key)
  end
end

ExUnit.start

defmodule CaseInsensitiveGetInTest do
  use ExUnit.Case
  import CaseInsensitiveGetIn

  test "gets an exact key" do
    assert ci_get_in(%{"name" => "Mujju"}, ~w(name)) == "Mujju"
  end

  test "gets capitalized key in map" do
    assert ci_get_in(%{"Name" => "Mujju"}, ~w(name)) == "Mujju"
  end

  test "gets capitalized input key in map" do
    assert ci_get_in(%{"Name" => "Mujju"}, ~w(Name)) == "Mujju"
  end

  test "gets mixed input key in map" do
    assert ci_get_in(%{"NaME" => "Mujju"}, ~w(nAme)) == "Mujju"
  end

  test "gets an exact deep key" do
    assert ci_get_in(%{"name" => "Mujju", "sister" => %{"name" => "Zainu"}}, ~w(sister name)) == "Zainu"
  end

  test "gets an mixed case deep map key" do
    assert ci_get_in(%{"name" => "Mujju", "sisTER" => %{"naME" => "Zainu"}}, ~w(sister name)) == "Zainu"
  end

  test "gets an mixed case deep key" do
    assert ci_get_in(%{"name" => "Mujju", "sisTER" => %{"naME" => "Zainu"}}, ~w(sIStER NAme)) == "Zainu"
  end

  test "gets a very deep key" do
    map = %{
      "aB" => %{
        "BC" => 7,
        "c" => %{"DD" => :foo, "Cassandra" => :awesome, "MOO" => %{"name" => "Mujju"}}
      }}
    assert ci_get_in(map, ~w(ab bc)) == 7
    assert ci_get_in(map, ~w(ab c dd)) == :foo
    assert ci_get_in(map, ~w(ab c moo name)) == "Mujju"

    assert ci_get_in(map, ~w(ab Bc)) == 7
    assert ci_get_in(map, ~w(ab C dD)) == :foo
    assert ci_get_in(map, ~w(ab C mOo nAMe)) == "Mujju"
  end
end

2017-01-05

Script to analyze the structure of an xml document

While working with XML data, you often don’t find the WSDL files and may end up manually working through the document to understand its structure. At my current project I ran into a few hundred XML files and had to analyze them to understand the data available. Here is a script I created which prints all the possible nodes in the input files

#!/usr/bin/env ruby
# Author: Khaja Minhajuddin <minhajuddin.k@gmail.com>

require 'nokogiri'

class XmlAnalyze
  def initialize(filepaths)
    @filepaths = filepaths
    @node_paths = {}
  end

  def analyze
    @filepaths.each { |filepath| analyze_file(filepath) }
    @node_paths.keys.sort
  end

  private
  def analyze_file(filepath)
    @doc = File.open(filepath) { |f| Nokogiri::XML(f) }
    analyze_node(@doc.children.first)
  end

  def analyze_node(node)
    return if node.is_a? Nokogiri::XML::Text
    add_path node.path

    node.attributes.keys.each do |attr|
      add_path("#{node.path}:#{attr}")
    end

    node.children.each do |child|
      analyze_node(child)
    end

  end

  def add_path(path)
    path = path.gsub(/\[\d+\]/, '')
    @node_paths[path] = true
  end
end

if ARGV.empty?
  puts 'Usage: ./analyze_xml.rb file1.xml file2.xml ....'
  exit(-1)
end

puts XmlAnalyze.new(ARGV).analyze

It outputs the following for the xml below

<?xml version="1.0" encoding="UTF-8"?>
<root>
  <person>
    <name type="full">Khaja</name>
    <age>31</age>
  </person>
  <person>
    <name type="full">Khaja</name>
    <dob>Jan</dob>
  </person>
</root>

/root
/root/person
/root/person/age
/root/person/dob
/root/person/name
/root/person/name:type

Hope you find it useful!

2016-12-23

Bash completion script for mix

Bash completion is very handy for cli tools. You can set it up very easily for mix using the following script.

#!/bin/bash
# `sudo vim /etc/bash_completion.d/mix.sh` and put this inside of it
# mix bash completion script

complete_mix_command() {
  [ -f mix.exs ] || exit 0
  mix help --search "$2"| cut -f1 -d'#' | cut -f2 -d' '
  return $?
}

complete -C complete_mix_command -o default mix

2016-12-22

How to show your blog content in your Rails application

I recently released LiveForm which is a service which gives you form endpoints (I’d love to have you check it out :) ) I wanted to show my blog’s content on the home page, It is pretty straightforward with the rich ruby ecosystem.

First you need a way to get the data from your blog. The LiveForm blog has an atom feed at http://blog.liveformhq.com/atom.xml . I initially used RestClient to get the data from the feed.
Once we have the feed, we need to parse it to extract the right content. Some quick googling led me to the awesome feedjira gem, (I am not gonna comment about the awesome name:))
feedjira actually has a simple method to parse the feed from a URL Feedjira::Feed.fetch_and_parse(url)
Once I got the entries, I just had to format them properly. However, there was an issue with summaries of blog posts having malformed html. This was due to naively slicing the blog post content at 200 characters by hexo (the blog engine I use), Nokogiri has a simple way of working around this. However, I went one step further and removed all html markup from the summary so that it doesn’t mess with the web application’s markup: Nokogiri::HTML(entry.summary).css("body").text
Finally, I didn’t want to fetch and parse my feed for every user that visited my website. So, I used fragment caching to render the feed once every day.

Here is all the relevant code:

The class that fetches and parses the feed

class LiveformBlog
  URL = "http://blog.liveformhq.com/atom.xml"
  def entries
    Rails.logger.info "Fetching feed...................."
    feed = Feedjira::Feed.fetch_and_parse(URL)
    feed.entries.take(5).map {|x| parse_entry(x)}
  end

  private
  def parse_entry(entry)
    OpenStruct.new(
      title: entry.title,
      summary: fix_summary(entry),
      url: entry.url,
      published: entry.published,
    )
  end

  def fix_summary(entry)
    doc = Nokogiri::HTML(entry.summary)
    doc.css('body').text
  end
end

The view that caches and renders the feed

<%= cache Date.today.to_s do %>
  <div class='blog-posts'>
    <h2 class='section-heading'>From our Blog</h2>
    <% LiveformBlog.new.entries.each do |entry| %>
      <div class=blog-post>
        <h4><a href='<%= entry.url %>'><%= entry.title %></a></h4>
        <p class='blog-post__published'><%= short_time entry.published %></p>
        <div><%= entry.summary %>...</div>
      </div>
    <% end %>
  </div>
<% end %>

Screenshot of the current page

2016-11-13

How to deploy a simple phoenix app on a single server using distillery

If you find issues or can improve this guide, please create a pull request at:

2. Setup the server

We’ll be running our server under the user called slugex. So, we first need to create that user.

## commands to be executed on our server
APP_USER=slugex

# create prent dir for our home
sudo mkdir -p /opt/www
# create the user
sudo useradd --home "/opt/www/$APP_USER" --create-home --shell /bin/bash $APP_USER
# create the postgresql role for our user
sudo -u postgres createuser --echo --no-createrole --no-superuser --createdb $APP_USER

3. Install the git-deploy rubygem on our local computer

We’ll be using the git-deploy rubygem to do deploys. This allows deploys similar to Heroku. You just need to push to your production git repository to start a deployment.

## commands to be executed on our local computer
# install the gem
# you need ruby installed on your computer for this
gem install git-deploy

4. Setup distillery in our phoenix app (on local computer)

We’ll be using distillery to manage our releases.

Add the distillery dependency to our mix.exs

1
2
3

defp deps do
  [{:distillery, "~> 0.10"}]
end

Init the distillery config

# get dependencies
mix deps.get
# init distillery
mix release.init

Change rel/config.ex to look like below

...
environment :prod do
  set include_erts: false
  set include_src: false
  # cookie info ...
end
...

5. Setup git deploy (local computer)

Let us setup the remote and the deploy hooks

## commands to be executed on our local computer

# setup the git remote pointing to our prod server
git remote add prod slugex@slugex.com:/opt/www/slugex

# init
git deploy setup -r "prod"
# create the deploy files
git deploy init
# push to production
git push prod master

TODO: release this as a book

6. Setup postgresql access

## commands to be executed on the server as the slugex user

# create the database
createdb slugex_prod

# set the password for the slugex user
psql slugex_prod
> slugex_prod=> \password slugex
> Enter new password: enter the password
> Enter it again: repeat the password

6. Setup the prod.secret.exs

Copy the config/prod.secret.exs file from your local computer to /opt/www/slugex/config/prod.secret.exs

1 2	## on local computer from our phoenix app directory scp config/prod.secret.exs slugex@slugex.com:config/

create a new secret on your local computer using mix phoenix.gen.secret and paste it in the server’s config/prod.secret.exs secret

It should look something like below:

# on the server
# /opt/www/slugex/config/prod.secret.exs
use Mix.Config

config :simple, Simple.Endpoint,
  secret_key_base: "RgeM4Dt8kl3yyf47K1DXWr8mgANzOL9TNOOiCknZM8LLDeSdS1ia5Vc2HkmKhy68"
  http: [port: 4010],
  server: true, # <=== this is very important
  root: "/opt/www/slugex",
  url: [host: "slugex.com", port: 443],
  cache_static_manifest: "priv/static/manifest.json"

# Do not print debug messages in production
config :logger, level: :info

# Configure your database
config :simple, Simple.Repo,
  adapter: Ecto.Adapters.Postgres,
  username: "slugex",
  password: "another brick in the wall",
  database: "slugex_prod",
  pool_size: 20

6. Tweak the deploy scripts

7. One time setup on the server

1
2
3

## commands to be executed on server as slugex
MIX_ENV=prod mix do compile, ecto.create
MIX_ENV=prod ./deploy/after_push

Logger

Exception notifications

Setup systemd

6. One time setup on server (on server as slugex user)

## commands to be run on the server as the slugex user
cd /opt/www/slugex

# create the secrets config
echo 'use Mix.Config' > config/prod.secrets.exs
# add your configuration to this file

# update hex
export MIX_ENV=prod
mix local.hex --force
mix deps.get
mix ecto.create

6. Nginx configuration

7. Letsencrypt setup and configuration

9. TODO: Configuration using conform

10. TODO: database backups to S3

10. TODO: uptime monitoring of websites using uptime monitor

10. TODO: email via SES

10. TODO: db seeds

10. TODO: nginx caching basics, static assets large expirations

10. TODO: remote console for debugging

sudo letsencrypt certonly –webroot -w /opt/www/webmonitor/public/ -d webmonitorhq.com –webroot -w /opt/www/webmonitor/public/ -d www.webmonitorhq.com

11. Check SSL certificate: https://www.sslshopper.com/ssl-checker.html

Common mistakes/errors

SSH errors

Improvements

Automate all of these using a hex package?
Remove dependencies on git-deploy if possible
Hot upgrades

2016-11-01

How to extract bits from a binary in elixir

Erlang and by extension Elixir have powerful pattern matching constructs, which allow you to easily extract bits from a binary. Here is an example which takes a binary and returns their bits

defmodule Bits do
  # this is the public api which allows you to pass any binary representation
  def extract(str) when is_binary(str) do
    extract(str, [])
  end

  # this function does the heavy lifting by matching the input binary to
  # a single bit and sends the rest of the bits recursively back to itself
  defp extract(<<b :: size(1), bits :: bitstring>>, acc) when is_bitstring(bits) do
    extract(bits, [b | acc])
  end

  # this is the terminal condition when we don't have anything more to extract
  defp extract(<<>>, acc), do: acc |> Enum.reverse
end

IO.inspect Bits.extract("!!") # => [0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 1]
IO.inspect Bits.extract(<< 99 >>) #=> [0, 1, 1, 0, 0, 0, 1, 1]

The code is pretty self explanatory

2016-10-31

Elixir process timeout pitfall

If you taken a look at Elixir, you may have come across something like the below code

defmodule HardWorker do
  def work(id) do
    Process.sleep(id * 900)
    {:done, id}
  end
end

defmodule Runner do
  @total_timeout 1000

  def run do
    {us, _} = :timer.tc &work/0
    IO.puts "ELAPSED_TIME: #{us/1000}"
  end


  def work do
    tasks = Enum.map 1..10, fn id ->
      Task.async(HardWorker, :work, [id])
    end

    Enum.map(tasks, fn task ->
      Task.await(task, @total_timeout)
    end)
  end

end

Runner.run

Looks simple enough, we loop over and create 10 processes and then wait for them to finish. It also prints out a message ELAPSED_TIME: _ at the end where _ is the time taken for it to run all the processes.

Can you take a guess how long this runner will take in the worst case?

If you guessed 10 seconds, you are right! I didn’t guess 10 seconds when I first saw this kind of code. I expected it to exit after 1 second. However, the key here is that Task.await is called on 10 tasks and if the 10 tasks finish at the end of 1s, 2s, … 10s This code will run just fine.

This is a completely made up example but it should show you that running in parallel with timeouts is not just a Task.await away.

I have coded an example app with proper timeout handling and parallel processing at https://github.com/minhajuddin/parallel_elixir_workers Check it out.

Addendum

I posted this on the elixirforum and got some feedback about it.

tasks = Enum.map 1..10, fn id ->
  Task.async(HardWorker, :work, [id])
end

# at this point all tasks are running in parallel

Enum.map(tasks, fn task ->
  Task.await(task, @total_timeout)
end)

Let us take another look at the relevant code. Now, let us say that this is spawning processes P1 to P10 in that order. Let’s say tasks T1 to T10 are created for these processes. Now all these tasks are running concurrently.

Now, in the second Enum.map, in the first iteration the task is set to T1, so T1 has to finish before 1 second otherwise this code will timeout. However, while T1 is running T2..T10 are also running. So, when this code runs for T2 and waits for 1 second, T2 had been running for 2s. So, effectively T1 would be given a time of 1 second, T2 a time of 2 seconds and T3 a time of 3 seconds and so on.

This may be what you want. However, if you want all the tasks to finish executing within 1 second. You shouldn’t use Task.await. You can use Task.yield_many which takes a list of tasks and allows you to specify a timeout after which it returns with the results of whatever processes finished. The documentation for Task.yield_many has a very good example on how to use it.

@benwilson512 has a good example on this

..suppose you wrote the following code:

task = Task.async(fn -> Process.sleep(:infinity) end)

Process.sleep(5_000)
Task.await(task, 5_000)

How long before it times out? 10 seconds of course. But this is obvious and expected. This is exactly what you’re doing by making the Task.await calls consecutive. It’s just that instead of sleeping in the main process you’re waiting on a different task. Task.await is blocking, this is expected.

2016-10-28

How to control pianobar using global hotkeys using Tmux

I love pianobar. However, until yesterday I hated pausing and moving to the next video using pianobar. I had a small terminal dedicated for pianobar and every time I had to change the song or pause, I used to select the window and then hit the right shortcut. I love hotkeys, the allow you to control your stuff without opening windows. I also happen to use tmux a lot. And it hit me yesterday, I could have easily bound hotkeys to send the right key sequences to pianobar running a tmux session. Here is how I did it.

I use xmonad, so I wired up Windows + Shift + p to tmux send-keys -t scratch:1.0 p &> /tmp/null.log So, now whenever I hit the right hotkey it types the letter ‘p’ in the tmux session scratch window 1 and pane 0, where I have pianobar running.

I use xmonad, but you should be able to put these in a wrapper script and wire them up with any window manager or with unity.

-- relevant configuration
, ((modMask .|. shiftMask, xK_p    ), spawn "tmux send-keys -t scratch:1.0 p &> /tmp/null.log")  -- %! Pause pianobar
, ((modMask .|. shiftMask, xK_v    ), spawn "tmux send-keys -t scratch:1.0 n &> /tmp/null.log") -- %! next pianobar

, ((modMask, xK_c     ), spawn "mpc toggle") -- %! Pause mpd
, ((modMask, xK_z     ), spawn "mpc prev") -- %! previous in mpd
, ((modMask, xK_v     ), spawn "mpc next") -- %! next in mpd

2016-10-27

How to use pianobar with a socks5 proxy to play pandora

I love pandora, However, I live in India where pandora is doesn’t stream. I got around this by proxying over socks5. Here is how you can do it.

First you need access to a socks 5 proxy, If you have an ssh server running in the US or any country where pandora streams, you can spin up a proxy connection by running the following command ssh -D 1337 -f -C -q -N username@yourserver.com

Once you have this running you’ll need to change your pianobar config to make it use this proxy

# ~/.config/pianobar/config
password = yoursecretpasswordinplaintext
user = youremail
proxy = socks5://localhost:1337/

Once you have this setup, you can just run the pianobar command and it will start playing your favorite music.

2016-10-27

A simple ticker to receive tick events for every interval in Elixir

Go has very utilitarian ticker methods, for instance check: https://gobyexample.com/tickers

package main

import "time"
import "fmt"

func main() {

    // Tickers use a similar mechanism to timers: a
    // channel that is sent values. Here we'll use the
    // `range` builtin on the channel to iterate over
    // the values as they arrive every 500ms.
    ticker := time.NewTicker(time.Millisecond * 500)
    go func() {
        for t := range ticker.C {
            fmt.Println("Tick at", t)
        }
    }()

    // Tickers can be stopped like timers. Once a ticker
    // is stopped it won't receive any more values on its
    // channel. We'll stop ours after 1600ms.
    time.Sleep(time.Millisecond * 1600)
    ticker.Stop()
    fmt.Println("Ticker stopped")
}

These are very nice for running code at every interval. If you want something like this in Elixir, it can be implemented in a few lines of code.

This code allows you to create a Ticker process by calling Ticker.start with a recipient_pid which is the process which receives the :tick events, a tick_interval to specify how often a :tick event should be sent to the recipient_pid and finally a duration whose default is :infinity which means it will just keep ticking till the end of time. Once you set this up, the recipient will keep getting :tick events for every tick_interval. Go ahead and read the code, it is pretty self explanatory.

There is also erlang’s :timer.apply_interval(time, module, function, arguments) which will run some code for every interval of time. However, the code below doesn’t create overlapping executions.

I have also created a gist in the interest of collaboration here: https://gist.github.com/minhajuddin/064226d73d5648aa73364218e862a497

defmodule Ticker do
  require Logger
  # public api
  def start(recipient_pid, tick_interval, duration \\ :infinity) do
    # Process.monitor(pid) # what to do if the process is dead before this?
    # start a process whose only responsibility is to wait for the interval
    ticker_pid = spawn(__MODULE__, :loop, [recipient_pid, tick_interval, 0])
    # and send a tick to the recipient pid and loop back
    send(ticker_pid, :send_tick)
    schedule_terminate(ticker_pid, duration)
    # returns the pid of the ticker, which can be used to stop the ticker
    ticker_pid
  end

  def stop(ticker_pid) do
    send(ticker_pid, :terminate)
  end

  # internal api
  def loop(recipient_pid, tick_interval, current_index) do
    receive do
      :send_tick ->
        send(recipient_pid, {:tick, current_index}) # send the tick event
        Process.send_after(self, :send_tick, tick_interval) # schedule a self event after interval
        loop(recipient_pid, tick_interval, current_index + 1)
      :terminate ->
        :ok # terminating
        # NOTE: we could also optionally wire it up to send a last_tick event when it terminates
        send(recipient_pid, {:last_tick, current_index})
      oops ->
        Logger.error("received unexepcted message #{inspect oops}")
        loop(recipient_pid, tick_interval, current_index + 1)
    end
  end

  defp schedule_terminate(_pid, :infinity), do: :ok
  defp schedule_terminate(ticker_pid, duration), do: Process.send_after(ticker_pid, :terminate, duration)
end


defmodule Listener do
  def start do
    Ticker.start self, 500, 2000 # will send approximately 4 messages
  end

  def run do
    receive do
      {:tick, _index} = message ->
        IO.inspect(message)
        run
      {:last_tick, _index} = message ->
        IO.inspect(message)
        :ok
    end
  end
end

Listener.start
Listener.run
# will output
# => {:tick, 0}
# => {:tick, 1}
# => {:tick, 2}
# => {:tick, 3}
# => {:last_tick, 4}

2016-10-13

Lets encrypt auto renewal for ubuntu and nginx

Create a file called /etc/nginx/le_redirect_include.conf

# intercept the challenges
location '/.well-known/acme-challenge' {
  default_type "text/plain";
  root /usr/share/nginx/letsencrypt;
}

# redirect all traffic to the https version
location / {
  return 301 https://$host$request_uri;
}

In your redirect block include this file

server {
  server_name www.liveformhq.com liveformhq.com;
  include /etc/nginx/le_redirect_include.conf;
}

To generate the LE keys run the following

sudo mkdir -p /usr/share/nginx/letsencrypt
# generate the certificate
sudo letsencrypt certonly --webroot=/usr/share/nginx/letsencrypt --domain cosmicvent.com --domain www.cosmicvent.com
# reload nginx
sudo kill -s HUP $(cat /var/run/nginx.pid)

Put the following in your crontab

1 2	$ sudo crontab -e @weekly /usr/bin/letsencrypt &> /tmp/letsencrypt.log; sudo kill -s HUP $(cat /var/run/nginx.pid)

2016-07-05

Algorithm to compute downtime of a service/server

I am working on an open source side project called webmonitorhq.com It notifies you when your service goes down. It also stores the events when a service goes down and comes back up. And I wanted to show the uptime of a service for a duration of 24 hours, 7 days etc,.

This is the algorithm I came up with, Please point out any improvements that can be made to it. I’d love to hear them.

The prerequisite to this algorithm is that you have data for the UP events and the DOWN events

I have a table called events with an event string and an event_at datetime

events
id
event (UP or DOWN)
event_at (datetime of event occurence)

Algorithm to calculate downtime

Decide the duration (24 hours, 7 days, 30 days)
Select all the events in that duration
Add an UP event at the end of the duration
Add a inverse of the first event at the beginning of this duration e.g. if the first event is an UP add a DOWN and vice versa
Start from the first UP event after a DOWN event and subtract the DOWN event_at from the UP event_at, do this till you reach the end. This gives you the downtime
Subtract duration from downtime to get uptime duration

e.g.

24 hour duration. Current Time is 00hours
UPat1 DOWNat5 UPat10
UPat1 DOWNat5 UPat10 UPat24
DOWNat0 UPat1 DOWNat5 UPat10 UPat24
UPat1 - DOWNat0 + UPat10 - DOWNat5 Downtime = 1 + 5
24 - 6 => 18

2016-05-20

Elixir IO.inspect to debug pipelines

While writing multiple pipelines, you may want to debug the intermediate values. Just insert |> IO.inspect between your pipelines.

e.g. in the expression below:

:crypto.strong_rand_bytes(32)
  |> :base64.encode_to_string
  |> :base64.decode
  |> :base64.encode

If we want to check intermediate values we just need to add a |> IO.inspect

:crypto.strong_rand_bytes(32)
  |> IO.inspect
  |> :base64.encode_to_string
  |> IO.inspect
  |> :base64.decode
  |> IO.inspect
  |> :base64.encode
  |> IO.inspect

This will print all the intermediate values to the STDOUT. IO.inspect is a function which prints the input and returns it.