YAML for Non-programmers

27 Aug 2018

It's been a while...

Introduction

I've been fiddling with home automation for a few years, but only recently found my way to http://www.home-assistant.io - a project with great developers and a great community. I've been hanging about the official Discord Chat Server, and try to give something back by helping people when I can.

One thing I noticed that people often struggle with is the YAML configuration files. A strange choice for the kind of quasi-programming you may want to do when automating your home appliances.

In this post I have tried to describe how YAML works, and how I think about the way it represents basic data structures. I hope it can be useful to someone...

Dictionaries and lists

To understand YAML, you need to understand what it's describing. YAML works with two main data structures - Dictionaries and Lists.

Lists are easy to understand. They're just a list - ordered collections of things. There are two things you need to remember though. Let me show them to you in a list :)

  • Lists are ordered. If you make a list with a dog, a cat and an elephant, the dog will be first, the elephant will be last and the cat will be in between.
  • Lists can be lists of anything; strings, integers, booleans, or even dictionaries or lists.

A list of lists might sound weird, but just think about the lists you have in your home.

  • A shopping list
  • A todo list
  • A list of things to pack for the vacation next week
  • The list of passwords on a post-it under your keyboard

Dictionaries (or dicts for short) are not much more complicated. You've probably encountered them in some way or another, but perhaps with a different name. In different programming languages they can be called dictionaries, hashes, maps, hash tables, tables, collections or even just objects. The technical name is Associative Array.

Regardless of name, the concept is simple. A dictionary is an unordered collection of entries - where each entry has a Key and a Value. The Key can be thought of as a name or title for the entry, and the Value is the data of the entry. Remember:

  • Keys are unique. Two entries in the same dict can not have the same key.
  • Values can be anything, just like the items in a list - Strings, integers, booleans, dictionaries or lists.

Let's look at some sample dictionaries:

Monday: Sausage and beans
Tuesday: Fish
Wednesday: French onion soup
Thursday: Pea soup and pancakes
Friday: Pizza

Each day is labeled by a key, and has a value - what are you going to eat that day. Note that while you could add another Wednesday to the end of the list, it wouldn't really make sense. Thus keys have to be unique. The values doesn't however. It would make perfect sense to eat pizza again on Saturday.

Since keys are unique, their order is not important:

Wednesday: French onion soup
Friday: Pizza
Monday: Sausage and beans
Tuesday: Fish
Thursday: Pea soup and pancakes

This dictionary contains exactly the same data as the one above. A clear difference from a list, where the order itself is a part of the data.

Another dictionary:

Name: Thomas Lovén
Email: thomasloven@example.com

That's a dictionary. Looks kind of like a database of a sort, doesn't it? Like the address book in your email program? Ah! But don't get fooled. The address book is a list, not a dictionary. However - the items in the list are dictionaries.

Let's add on to that dict:

Name: Thomas Lovén
Email: thomasloven@example.com
Hobbies: singing, woodworking, home automation

Now we added an entry to the dict where the value is a list. I have three hobbies. This illustrates that the value of a dictionary entry can be anything. Even a dictionary:

Name: Thomas Lovén
Email: thomasloven@example.com
Hobbies: singing, woodworking, home automation
Phones: Home: +46 (0)XX XXXXXX, Work: +40 (0)XX XXXXXX

And remember that lists can contain dictionaries too...

Name: Thomas Lovén
Email: thomasloven@example.com
Hobbies: singing, woodworking, home automation
Phones: Home: +46 (0)XX XXXXXX, Work: +40 (0)XX XXXXXX
Children: Name: N, Age: 3 ; Name: H, Age: 1

But at this point things are getting advanced. It's hard to keep track of what is a dict and what is a list and what contains what...

If only there was a language to describe these concepts... a sort of Markup Language, if you will... but who needs Yet Another one of those?

Let me tell you about

JSON

Javascript Structured Object Notation - JSON. You thought I was going to say YAML, didn't you? Soon...

JSON is a simple way of writing down the concepts described above, which can be easily understood both by humans and by computers.

JSON describes object and arrays, there are objects and arrays, but let's call them dictionaries and lists instead.

lists are surrounded by square brackets and contain items separated by commas. The items can be strings, numbers, dictionaries, lists or any of the magic values true, false, or null.

Dictionaries are surrounded by curly braces and contain entries separated by commas. The key and value of each entry is separated by a colon. Keys must be strings, but values can be anything that can be in a list.

Let's look at our dict from above in JSON format:

{
  Name: "Thomas Lovén",
  Email: "thomasloven@example.com",
  Hobbies: [
    "singing",
    "woodworking",
    "home automation"
    ]
  Phones: {
    Home: "+46 (0)XX XXXXXX",
    Work: "+46 (0)XX XXXXXX"
  }
  Children: [
    { Name: "N", Age: 3},
    { Name: "H", Age: 1}
    ]
}

I added some line breaks and indentations to make it more pretty, but this is much easier to read. Even the last entry with the nested dict-list-dict about my children. Two things to note:

  • You don't need quotes around the keys, but you do around string values. If you want whitespace in a key (which is entirely OK) it must be quoted, though.
  • The indentations and newlines I added, and in fact any whitespace not in quotes, is ignored.

OK. Now you understand one markup language. Let's learn something different.

YAML

YAML Ain't a Markup Language - but it's pretty darn close, to be honest.

While probably not historically accurate, YAML can be seen as an evolution or superset of JSON. In fact, any valid JSON is also valid YAML. That might be useful to remember. There are some notable differences, though.

First of all, YAML doesn't require braces. Instead items in lists are separated by newlines where each item starts with a dash:

- Item 1
- Item 2
- Item 3 is a long one that stretches over multiple lines.
  The new item won't start until we get to a line that starts
  with a dash, like the one below this one.
- Item 4
-
  - Item 5a
  - Item 5b
  - Item 5c
- Item 5

Some things to note:

  • There are no quotes. In YAML quotes are pretty much optional.This can be a blessing and a curse. For example "true" is a string, but true is a boolean value.
  • Indentation is important. Item 3 in the list stretches over multiple lines. Each line after the first one is indented (with an equal number of spaces - NOT tabs). The same is true for Item 5, which is a list. Each item in the sub-list is indented with an equal number of spaced.
  • The items of the list are not of the same type. Most are strings, but item 5 is a list.

I'll just take the opportunity to say this again: Indentation is important. Probably the most important concept of YAML.

Slightly simplified one can say that by indenting line, those line become a continuation of the line above, until the next de-dented line.

Dictionaries are also written one entry per line, as key: value:

Name: Thomas Lovén
Email: thomasloven@example.com
Hobbies:
  - singing
  - woodworking
  - home automation
Phones:
  Home: +46 (0)XX XXXXXX
  Work: +46 (0)XX XXXXXX
Children:
  -
    Name: N
    Age: 3
  - Name: H
    Age: 1

Things to note:

  • The space after the comma is important.
  • The value of Hobbies is a list. Like above, each line of the value is indented by an equal number of spaces.
  • The value of Phones is a dictionary. The same indentation rules apply.
  • The value of Children is a list where each item is a dictionary. So each line in each dictionary is indented twice.
  • The second entry in the list of children uses a contracted form, where the first key-value pair of the dictionary is put on the same line that signifies the list item. More on this later.

And that's all the basics of data representation using YAML.

On indentation

As I've been trying to help people with their configurations on the homeassistant Discord server, I have found one problem which occurs more than any other. Indentation errors.

Indentation is important

It must be correct or the YAML either won't be accepted by the parser or will describe something entirely different from what you intended.

The only advice I can give is to think carefully about the structure of the data you are trying to represent. What is your object? Is it a dictionary or a list? Where is it contained? Is it freestanding? Is it an entry in a dict? Is it an item in a list? What is it's parent? What are it's children? What are it's siblings?

If you have a complex structure - say some nested cards in lovelace - it might help to make a drawing on actual paper of what you want.

In the YAML dictionary sample above, I used a contracted form in my list of dicts. This is common practice, but may be a bit confusing at first since it makes the indentation unclear.

If might be easier to understand the structure of the document if you use the expanded form:

Children:
  -
    Name: N
    Age: 3
  -
    Name: H
    Age: 1

Also, remember that dictionary entries are unordered. The YAML below describes the same thing as that above, but might be more confusing.

Children:
  - Name: N
    Age: 3
  - Age: 1
    Name: H

Often software generated YAML (such as what's produced by lovelace-gen) write the entries out in such a way that the keys are ordered alphabetically.

Advanced topics

Comments

Adding comments to your code makes it easier to understand. Both to other people, and - more importantly - to your self when you return to it in six months because something stopped working.

In YAML, comments begin with a number sign #, last until the end of the line and are simply ignored by the parser.

# A dictionary about me
Name: Thomas Lovén
Email: thomasloven@example.com # This isn't really my email
Hobbies:
# Just some of the ways I like to waste time
  - singing # choir, mostly
  - woodworking

If you want a number sign in your entry, the entry must be quoted.

Spaces and colons

As mentioned, YAML doesn't require quotes around strings, but they are allowed. Quotes can be useful to tweak the parsing. Imagine for example the following list:

- Halflife 1
- Halflife 2
- Halflife 2: Episode Two

This is a list of strings, right? Wrong. The third entry is a dictionary with a single entry "Halflife 2" with value "Episode Two" (keys can contain spaces, by the way).

To fix this, you can use quotes:

- Halflife 1
- Halflife 2
- "Halflife 2: Episode Two"

Using JSON

There's a reason I went through JSON to explain YAML. As I said, all JSON is also valid YAML. This allows for compact notation:

Name: Thomas Lovén
Email: thomasloven@example.com
Hobbies: [singing, woodworking, home automation]
Phones:
  Home: +46 (0)XX XXXXXX
  Work: +46 (0)XX XXXXXX
Children:
  - {Name: N, Age: 3}
  - {Name: H, Age: 1}

I mention this because you just might run into it sometime. I like to use it in my configurations to bring down the line count, but it's easy to go overboard and make the data hard to read instead. In the end it's a matter of taste.

Note that there are still no quotes. That's OK as long as you don't want commas, } or ] in the value.

Further, YAML allows single line comma separated lists without braces.

Hobbies: singing, woodworking, home automation

This also means that a string containing a comma must be quoted.

Merging

Dictionaries can be merged using the merging operator: <<. For example:

a key: a value
b key: b value
<<: {d key: d value, e key: e value}

will be parsed as

a key: a value
b key: b value
c key: c value
d key: d value

and so will

a key: a value
b key: b value
<<:
  c key: c value
  d key: d value

In short, the << operator takes a dictionary as its value, and merges it into the parent dictionary.

Node anchors

Merging is very convenient when used in combination with node anchors.

Node anchors are a way of saving a dictionary, and reusing it later

my_dict: &my_dict
  a: 1
  b: 2
  c: 3

In this case &my_dict is NOT the value of my_dict, but a node anchor - as signified by the ampersand &.

The anchor saves the value for later reuse and can be recalled any number of times using an asterisk *:

a dictionary: &saved
  a: 1
  b: 2
  c: 3

another dictionary: *saved

a list:
  - *saved
  - *saved

This will be parsed as:

a dictionary:
  a: 1
  b: 2
  c: 3

another dictionary:
  a: 1
  b: 2
  c: 3

a list:
  -
    a: 1
    b: 2
    c: 3
  -
    a: 1
    b: 2
    c: 3

You can also merge an anchor if you want to add more entries to the dict:

base: &base
  a: 1
  b: 2

extended version:
  <<: *base
  c: 3

You should be able to guess how this parses out.

Now, for my final trick:

Merging while defining.

The problem with the above examples is that you need to put the definition somewhere. The YAML snippets above will have the dictionary entries a dictionaryand base defined and set no matter what. Sometimes that's impractical, which is why you often see the following in homeassistant packages:

homeassistant:
  customize:
    package.node_anchors:
      common: &common
        key1: val1
        key2: val2

    sensor.my_sensor:
      <<: *common
      icon: mdi:temp
    sensor.another_sensor:
      <<: *common
      icon: mdi:home

The package.node_anchors entry in the customize dictionary contains a dictionary of stuff that is simply ignored. Anything you put there will have no effect on the package, so it's a great place to define anchors.

Another possibility is to put the definition in the first place it is used, and merge it immediately:

homeassistant:
  customize:
    sensor.my_sensor:
      <<: &common {key1: val1, key2: val2}
      icon: mdi:temp
    sensor.another_sensor:
      <<: *common
      icon: mdi:temp

Not all YAML parsers allow this, but it seems to work with homeassistant.

Comments

comments powered by Disqus
© 2012 Thomas Lovén - @thomasloven - GitHub