This is a work through an discussion of Metaprogramming Elixir.

The entry point of metaprogramming in Elixir is the quite macro.

quote do: 1 + 2
{:+, [context: Elixir, import: Kernel], [1, 2]}

This allows the transformation of a code block into the abstract syntax tree (AST). This can then be manipulated like any other data structure and return to the generated code. The practical upshot of this is that half of the Elixir language is actually macros. You can if you need to write your own macros to perform complex operations simply.

There are recommendations to limit the use of macros to things that you can’t solve with normal functions. It’s possible with macros to create a powerful equivalent of inheritance, but must be used sparingly to avoid magical code.

Most of this work takes place at compile time rather than at runtime.

You use require to import macros.

General warning: in languages that have implicit returns adding logging at the end of a function masks the result. Capture the result, log then return result.

unquote is a technique where a macro can read variables from the hosts context. This is how you pass parameters into a macro.

This is how to wrap a callback api in promises:

// Sample function with callback
function clientFactory() {
  return {
    request: (name, callback) => {
      callback(name)
    }
  }
}

let client = clientFactory()

client.request('a', console.log)

// Wrapping a callback into a promise

function returnCallbackResult(input) {
  let promise = new Promise(
    resolver => {
      let client = clientFactory()
      client.request(input, (a) => {
        resolver(a)
      })
  })
  return promise
}

returnCallbackResult('b').then( console.log )

// How to wrap a promise

function chainingPromise(input) {
  let promise = new Promise(
    resolver => {
      input.then( (a) => resolver(`wrapping ${a}`))
    }
  )
  return promise
}

chainingPromise( returnCallbackResult( 'c' ) ).then(console.log)

I have finally finished working through the example in the book.

https://github.com/chriseyre2000/mastery ( 94529b1)

This is a great introduction to OTP covering some details that I have not seen elsewhere. In particular the ability to import an OTP project and not have it auto start will be useful elsewhere.

The last few chapters do seem rushed, and the example code does not quite match what you get when it is typed in. However it is a great example of building a domain model in Elixir and then adding on the external links later. This would count as Hexagonal Architecture.

This is more of a training course that a book. You would have trouble simply reading through without working on the exercise. This does take a significant amount of time.

We are now onto writing the outside in tests.

I have got as far as the first test. This is failing with authentication errors.

[error] Postgrex.Protocol (#PID<0.251.0>) failed to connect: ** (Postgrex.Error) FATAL 28P01 (invalid_password) password authentication failed for user "chris*****"

iex -S mix came to the rescue here:

h Ecto.Adapters.SQL.Sandbox.checkout

Checks a connection out for the given repo.

The process calling checkout/2 will own the connection until it calls checkin/2
or until it crashes when then the connection will be automatically reclaimed by
the pool. 

## Options

  • :sandbox - when true the connection is wrapped in a transaction.
    Defaults to true.
  • :isolation - set the query to the given isolation level.
  • :ownership_timeout - limits how long the connection can be owned.
    Defaults to the value in your repo config in config/config.exs (or
    preferably in config/test.exs), or 60000 ms if not set. The timeout exists
    for sanity checking purposes, to ensure there is no connection leakage, and
    can be bumped whenever necessary.

This was the important clue that I had not got my test database configured correctly.

This is the new version of config/test.exs

use Mix.Config

config :mastery_persistence, MasteryPersistence.Repo, 
                             database: "mastery_test", 
                             hostname: "localhost", 
                             port: 54320, 
                             pool: Ecto.Adapters.SQL.Sandbox,
                             username: "postgres",
                             password: "postgres"

config :mastery, :persistence_fn, &Mastery.Persistence.record_response/2

I had “forgotten” the username here.

I now have the persistence tests implemented: https://github.com:chriseyre2000/mastery (dc6c5bf)

I am now onto chapters 7 – 9 which deal with OTP. This is the really interesting part of the book. I am unsure that my sample code is accurate as there are very few tests and the supplied sample code has diverged from that which you enter if you are following along. I can’t tell if I have made a mistake or if the supplied code is wrong (my suspicion is a little of both). This has taught me a lot about debugging Elixir applications which will be useful. What looked like a cryptic error message was actually precise and detailed where the genserver received a Map when it was expecting an atom. Adding guard clauses is a great way to flush out these problems.

The book is written as a long exercise where you add code snippets to your application to update. Given that I am using an epub format I can’t simply cut and paste the code, so I am typing it in. This is interesting since epub adds a trailing - when wrapping lines which takes a while to get used to. This has a certain error rate so I frequently use `iex -S mix` to check for typos. This works well for sections where the required pieces are defined first. These later chapters were written outside-in (a technique that I like, but normally back with tests) so this does not help.

One of the downsides of these long exercises is that it can be hard to find a mistake in a snippet that has been updated several times.

The database additions have slowed me down as I don’t currently have a local working postgres version. The samples for the database completely miss out the credentials required, but this information is available elsewhere.

Here is the docker-compose.yml that I got to work (note that the port is marked host:container):

version: "3"
services:
  db:
    image: "postgres:11"
    environment:
      - POSTGRES_PASSWORD=postgres
    container_name: "my_postgres"
    ports:
      - "54320:5432"
    volumes:
      - my_dbdata:/var/lib/postgresql/data
volumes:
  my_dbdata: {}

In practice we would use a better username and password. I can have a portable database setup by using: docker-compose up -d db

Here is the config that I used to get the database setup:

use Mix.Config

config :mastery_persistence, MasteryPersistence.Repo,
  database: "mastery_dev",
  hostname: "localhost",
  port: 54320,
  username: "postgres",
  password: "postgres"

The use of OTP fails to mention the default 5 second timeout behaviour built into OTP requests. It does cover using timeouts to handle late responses and has a good explination of the use of via.

Poncho projects are an interesting concept. You just create a full mix application in a subdirectory (or a parallel directory). This allows it to be moved into a distinct repository later on should it be needed, but is less overhead than an umbrella project.

Now on to integrating the mastery_persistence into mastery. This part looks a bit rushed. They are asking us to change files that have not yet been created (config/config.exs).

I have worked through the examples and pushed the finished details to:

github.com:chriseyre2000/mastery

I am currently at: 6963329

All that remains is the testing chapter and going back and making sure everything works.

I am again working through this book.

I started again with the real edition. The major change is that it is now focused on Worker Bees rather than Wildabeast’s.

So far I have made it to the testing chapter at the end of part one.

It appears that the samples have not been updated to match the latest code.

The testing techniques are really good. Given that variables are immutable then test fixtures really are reusable. The context part of ExUnit is very clear.

I understand delaying the tests to keep the book to a reasonable length, but event a single character typo can break functionality and be hard to fix later. I had misspelt the key on a lookup and though that the tests were broken.

Here is the current state https://github.com/chriseyre2000/mastery at f4d9f35

This is a follow up to part one

I have made some more progress on this project and we are now down to 60 errors per day from the previous thousand. Given that we are integrating with other a hundred system this is more acceptable. Once a breakthrough was made to clear the highest frequency problem, the rest were fixed easily.

A lot of the existing log messages lacked context. They typically included what method was being used and the stacktrace. What was missing was enough information to recreate the problem.

My previous claim about not using a debugger no longer stands. I was forced to use this to identify paths through some code that had not been developed via TDD (it had some acceptance tests but the details were obscured). We are still removing the old Betamax tests – these consisted of recordings of a production run that are replayed. This is a good start, but is painful to adapt.

Most agile teams following some form of Scrum hold a daily standup.

Typically this is held in the morning at a time when the whole team can be in. I have also heard of end of day standups. These work better for teams that work remotly across timezones.

There are several forms of this meeting.

The basic is the three questions:

– What did you do yesterday

– What are you going to do today

– Any blockers

Another that I prefer is to walk the project board from right to left. Duscuss all cards that are not yet done. This will ensure that the board is upto date plus can show when someone is working off the plan.

I have also worked on a team that held two standups a day. A full one in the morning and a second quick version after lunch to cover changes.

I recently saw a tweet with a diagram using the wardley map value chains to document some software.

Another call out for @swardley maps as @suksr talks about using DDD to reduce cost of poor quality software #BFSServerless pic.twitter.com/ajL5HMdDt4

— Will Hamill (@willhamill) January 24, 2020

This looks to be an interesting way of viewing a set of products.

Given my use of graphviz I thought that I would give it a go.

Here is the repo: https://github.com/chriseyre2000/diagrams

This includes the dot file for the above diagram.

The idea of the value stream is to show the dependencies from the visible to the invisible. The horizontal axis has the custom items on the left and the commodity ones on the right.

I will be experimenting with this for a while.