General Information

This is an introduction for people new to JSON and Markdown, if you are already familiar with JSON schemas already, move along now, nothing to see here.

Headers

Please make sure all property markdown files in documentation/properties/ folder start with the following header:


--- 
layout:            property
title:            "name of the data item as it appears in the schemas (ie. lowercase)"
schemas:        [a list of schemas where this data_item is used]
---

JSON

JSON is short for "Javascript Object Notation" and defines the format for an "object". An object in programming terms can be a variable, data structure or a function (basically anything). More commonly, in Object-Oriented programming an object typically refers to an Instance of a Class. That is to say, a specific description of something more general. Like a Goat being an instance of the Class "Animal".

For our purposes, we can just consider JSON to be the data and a JSON-schema to just define the format for that data, ie. what it needs to look like so the computer can understand where things are when it receives it.

Properties, names and values

A property in JSON notation is a name-value pair and list of properties is what makes up the data in our schemas. What you would normally call a "field" (like interest_rate) is the name of the property and what we might call an attribute is the value of the property.

One property

{"name": value}

List of properties

{"name1": "value1", "name3": "value3", "name2": "value2"}

JSON is just text, so the formatting is purely visual, even the order (see above) does not matter. As things get more complicated however, it is considered best practice to format your JSON to look more human-friendly:

{
  "name1": "value1",
  "name3": "value3",
  "name2": "value2"
}

You can visit: http://jsonprettyprint.com/ to help with this formatting.

JSON-Schema

While any data sent/received in JSON format will always look like the above, we need a way to make sure it is what we are expecting. For this we define a JSON-Schema which basically describes what the JSON formatted data should look like.

Description

What if we want to add a little more information to describe our property? This is where the "description" parameter comes in. We add some curly brackets where the "value" is supposed to be and add a description as so:

{"name": {"description": "A little blurb about this property"} }

or

{
  "name": {
   "description": "A little blurb about this property"
  }
}

Types

Now that we have described our property value, let's go a step further and narrow it down to a specific type. Is it a number, a word, a list? The "type" parameter allows us to specify exactly this. So now our schema would look something like this:

{
  "name": {
   "description": "A little blurb about this property",
   "type": "number"
  }
}

JSON has 7 standard types that we can use:

A null value means the value is unknown. Note that this is different from an empty or zero field. If the field is empty, undefined or does not exist, then it can simply be omitted (unless it is indicated as being "required" in which case you should provide a suitable default value).

An integer is a number without a fraction or exponent part.

ex 1. 4
ex 2. 26908289076124561671021

A number is a number with or without a fraction or exponent part.

ex 1. 26908289076124561671021
ex 2. 269082.89076124561671021

A string is a list of characters (except " or \ ) inside "quotes". You can think of a string as a word. Note that the "word" can also contain numbers (like your national insurance number). But also note that numbers represented as strings need to be converted back to numbers if you want to add or multiply them.

ex 1. "sheep"
ex 2. "AS546NB8"

A boolean is simple true or false flag. Note that the true/false flag is lowercase and not inside "quotes".

ex 1. true
ex 2. false

If you were wondering, the word boolean comes from a founding father of modern logic, the English mathematician George Boole.

An array is a list of the other types, separated by commas and inside square brackets [ ].

ex 1. [2, 235, 34634, 34]
ex 2. ["sheep", "sheep_dog", "fox"]

An object is a JSON object, or in other words, the thing we are defining. So this allows for nesting of objects within objects. This is valid JSON, but adds time and complexity in the decoding/parsing process so generally should be avoided.

{
  "farm_id": "E2G2K3LSJENJ4J3K10H",
  "animals": {
    "goat": 2,
    "sheep": 7,
    "sheep_dog": 1
  },
  "farm_owner": "peter"
}

Restrictions

Once we have given a description, a type and maybe a format for our property value we can implement some sanity checks by applying further restrictions.

Restrictions can come in the form of visual presentation like formats or enums, or they can come in the form of simple quantitative sanity-checks. Both are extremely useful to narrow down the possibilities of what might be considered as "valid data" according to the schema. These restrictions not only ensure that bad data is caught at the most granular level, but it also ensures that common semantics are used to define the same thing. It is the first step towards a harmonised standard. In other words, if a loan currency is US Dollar, let's agree to call it "USD" instead of "US_Dollar", "dollar-USA", "011" or "$US."

Restrictions can be difficult to implement as you need to consider all potential edge cases. Account balances are generally positive but sometimes can be negative, too. Leveraging widely used ISO or IFRS standards are therefore a great way to ensure you have considered the full spectrum of possible values. It also means that, more often than not, firms will already be familiar with and recording data in line with these standards.

Formats

How would you represent a date value like 31 August 2014?

  • We could make use of a string type: "31 August 2014" but then "Aug 31 2014" would also be valid and we would have trouble ordering our data by date.
  • We could accept an integer type: 31082014 but then 1235 would also be a valid integer. And even if we restricted it to 8-digits, we would have to remember to make sure single digit months have a leading zeros and still this doesn't tell us if it is 31 August 2014 or 14 February 8013.

You get the idea, sometimes we need more information. This is where the format parameter comes in.

The following is the list of the formats specified in the JSON Schema specification:

The one we use most commonly from this is the "date-time" format and it basically means your dates and timestamps need to be a valid string but also look like this: YYYY-MM-DDTHH:MM:SSZ in accordance with ISO 8601.

So for our example, if this is your schema:

{"animal_birthday": {
  "description": "The recorded birthday for the animal.",
  "type": "string",
  "format": "date-time"
  }
}

then this would be a valid input:

{"animal_birthday": "1995-06-07T13:00:00Z"}

and this would be an invalid input:

{"animal_birthday": "1995-6-7T13:00:00Z"}

Enums

enum is short for enumeration which is short for "a complete, ordered listing of all the items in a collection." All that means is a list (an array) of possible values our value can have. Enums are typically used with string types to limit the range of possible strings that are considered valid for a property value.

Again, it is important to consider edge cases and hence where relevant, it is advisable to leave an "other" or "none" value so that and edge case can be temporarily mapped to a generic parameter. Particularly, in the case of an open source project, you can envisage enum lists getting longer as the schemas evolve through contributions.

So for example, if this is your schema:

{"animal_type": {
  "description": "The type of animal on the farm.",
  "type": "string",
  "enum": ["sheep", "goat", "cow", "other"]
  }
}

then this would be a valid input:

{"animal_type": "goat"}

and this would be an invalid input:

{"animal_type": "horse"}

More information

The JSON spec can be found here and JSON-schema spec cane be found here.


results matching ""

    No results matching ""