Installing Lumen

Lumen is an alternative BEAM Implementation.
Warning that this is still at a very early stage. It promises the ability to build WebAssembly

I am not currently interested in using Rust to build it from sources, but there are now images published here

I used the following to install the tarball (I use a mac):

sudo tar xf lumen-0.1.0-nightly-x86_64-apple-darwin.tar.gz -C /opt

This deploys the bundle to: /opt/lumen

The compiler is now available at: /opt/lumen/bin/lumen

Here are more details on lumen:

On having a Useless Superpower

I am going to admit to having a minor superpower. I can hear high pitch sounds that most people can’t.

This is not at all useful as it results in me getting a headache in lots of shops or museums (and in one especially annoying instance a job interview). It’s typically a 50hz signal.

Minimalist Clustering Demo

This is based on code from a video that I have previously linked to.
However Elixir has made a few breaking changes since then.

This is going to demonstrate how to cluster two local Elixir nodes, deploy code from one to the other and then upgrade it while it is running.

You need to put the following code in a file called blabber.ex

defmodule Blabber do

  def start do

    spawn(fn -> loop(0) end)


  def loop(uptime) do

    receive do

      :stop ->

        IO.puts "Shutting blabber down ..."


      after 1000 ->

        IO.puts "Nice! #{uptime} seconds of bgu-free uptime on #{node()}."


    Blabber.loop(uptime + 1)



Next you need to start iex with a name and a cookie (in the same folder as blabber.ex):

iex --sname cat --cookie super-secret

In a second terminal start another iex session

iex --sname dog --cookie super-secret

From the first terminal connect the two nodes together:

Node.connect :"dog@MacBook-Pro"

Compile the code in the cat node:

c ("blabber.ex", ".")

We need to use this version of the compile function as by default since Elixir 1.5 the beam file is not output by default.

The process can be started on the cat node with:


Now you can transfer the compiled beam file to all linked nodes using:

nl Blabber

Now from the dog node you can start the process:


At this point we notice the typo in the code. It say bgu-free rather than bug-free.
Edit the source file save the change.

On the cat node recompile this:

c ("blabber.ex", ".")

This will immediately fix the problem on the cat node (and leave the count updating).

If you use:

nl Blabber

then the solution will be moved to the other node as well.

This demonstrates the simplest possible Clustering arrangement. It is simple to extend this to a move sophisticated service.

Twitter For Tech Support

Sometimes a companies official Tech Support channels can be a little slow.

If the official channels don’t work then you can turn to twitter. Raising a support issue on a social media site can cause the company to take action. Sometimes a DM to them will enable action that a support ticket does not.


This is a simple powershell script that can be used to get the frequency of the first letter from a sample file.

gc ‘./sample’ | %{ $_.substring(0,1) } | group

Running this over say the FTSE 100 symbol list returns:

Count Name                      Group
----- ----                      -----
   10 A                         {A, A, A, A...}
   11 B                         {B, B, B, B...}
    4 C                         {C, C, C, C}
    2 D                         {D, D}
    2 E                         {E, E}
    3 F                         {F, F, F}
    3 G                         {G, G, G}
    5 H                         {H, H, H, H...}
    8 I                         {I, I, I, I...}
    3 J                         {J, J, J}
    1 K                         {K}
    4 L                         {L, L, L, L}
    4 M                         {M, M, M, M}
    3 N                         {N, N, N}
    1 O                         {O}
    6 P                         {P, P, P, P...}
    8 R                         {R, R, R, R...}
   15 S                         {S, S, S, S...}
    2 T                         {T, T}
    2 U                         {U, U}
    1 V                         {V}
    2 W                         {W, W}

This highlights that the symbols are not uniformly spread across the alphabet.

A-F has 1/3 of the market as does P-Z

I found out this once when trying to use the ticker symbol to load balance market data across 3 servers.

Check the distribution of the data before you use a simple key.

Oddly the second letter is a better key:

    6 A                         {A, A, A, A...}
    3 B                         {B, B, B}
    4 C                         {C, C, C, C}
    5 D                         {D, D, D, D...}
    3 E                         {E, E, E}
    6 G                         {G, G, G, G...}
    4 H                         {H, H, H, H}
    3 I                         {I, I, I}
    2 K                         {K, K}
    8 L                         {L, L, L, L...}
    6 M                         {M, M, M, M...}
    7 N                         {N, N, N, N...}
    2 O                         {O, O}
    4 P                         {P, P, P, P}
    8 R                         {R, R, R, R...}
    9 S                         {S, S, S, S...}
    7 T                         {T, T, T, T...}
    2 U                         {U, U}
    6 V                         {V, V, V, V...}
    2 W                         {W, W}
    2 X                         {X, X}
    1 Z                         {Z}