Skip to content
This repository has been archived by the owner on Oct 25, 2021. It is now read-only.

Latest commit

 

History

History
237 lines (183 loc) · 24.6 KB

schema-spec.md

File metadata and controls

237 lines (183 loc) · 24.6 KB
Raw

treecreeper schema specification

Note that yaml files are indented with two spaces

Types

Types are defined in individual files in a types subdirectory. Add a .yaml file named after your type e.g. Cookie.yaml, DNSRecord.yaml. Names of types should be singular, without punctuation, and with each word/abbreviation capitalised. The file expects the following properties:

name required default details examples
name yes The name of the type. This should be exactly the same as the name of the file. It must begin with an upper case letter and contain only upper and lower case letters CostCentre, FastlyInstance
pluralName no The pluralised version of the type name. If not specified, defaults to the type name with an s appended People
description yes Description of what kind of entity the type is. This can be spread over multiple lines and supports markdown. It is used to describe types in the graphql schema (and UIs powered by this such as graphiql) and in tc-ui
moreInformation no An optional extension to the description to allow an extended description for the type. This can be spread over multiple lines and supports markdown. It is used to provide additional information in the UI
properties yes An object containing one or more Property Definition. Each key in this object defines the property name, and must be a camel-cased string, i.e. only letters and numbers, beginning with a lower case letter, with capital letters typically used to mark the beginning of a new word property names: officeLocation, supportsHttp2
fieldsets no An object containing one or more Fieldset Definition. Each key in this object defines the fieldset name, and must be a camel-cased string, i.e. only letters and numbers, beginning with a lower case letter, with capital letters typically used to mark the beginning of a new word fieldset names: general, contactDetails
createPermissions no An array listing the codes of clients that are allowed to create instances of this type. If not specified, any client is allowed to - my-type-importer
minimumViableRecord no An array listing the names of properties of this type that are considered particularly important. This information is used by tc-ui to ensure this data is entered - deliveryTeam
inactiveRule no Many types have an isActive property, which is used to specify whether a particular record is active. Some types may use different properties to define this, and may define an inactiveRule object defining one or more property names and the values they can take that denote they are inactive lifecycleStage: Decommissioned

Property Definitions

Each property definition is an object with the following properties

name required default details examples
type yes Code on properties named 'code', otherwise undefined This defines what kind of data can be stored in the property. It can be the name of any primitive type or enum Paragraph, ServiceTier
label yes A short label to be used when displaying this property in forms etc.
description yes Description of what data is stored in the property. This is used to provide information in systems that allow viewing or modifying the data
required no true on properties named 'code', otherwise false Boolean defining whether or not the field is required
unique no true on properties named 'code', otherwise false Boolean indicating if values in this property must be unique across all records of this type
canIdentify no true on properties named 'code', otherwise false Boolean defining whether or not this field contains a 'pseudo id', i.e. a value that is unique, or almost unique, to a record, and can therefore be used in graphQL to query for a record e.g. a Person record's email property is a good example of a value that canIdentify
trueLabel no 'Yes' For properties with Boolean type, this defines the form label to use for the true option
falseLabel no 'No' For properties with Boolean type, this defines the form label to use for the false option
examples no Array of example values. This is not widely used but may e.g. be surfaced as placeholder text in a form
fieldset no 'misc' The name of a fieldset to group this field in. Defaults to 'misc' which will e.g. put the field in the Miscellaneous fieldset at the end of edit pages
useInSummary no true on properties named 'code', otherwise false Boolean indicating if the property is considered an essential property to use when constructing a summary of the record. At present, this is used when presenting search results
pattern no String which is the name of a string validation rule. CODE, PERMISSIVE

Note: All existing and any new properties, except for relationships, will automatically have the functionality to be filtered.

Primitive types

In treecreeper, you can define as many primitive types as you like. The names of the types are provided to tc-ui, and allow different behaviours to be implemented e.g. both 'Word' and 'Paragraph' might use the underlying 'String' type in GraphQL, but one might use a text input in tc-ui, while the other uses a textarea

To define a primitive type, create a primitive-types.yaml file. This contains a map from type names to underlying GraphQL types and tc-ui component choices, e.g.

Paragraph:
    graphql: String
    component: LargeText

By default, tc-schema-sdk will include the following primitive types:

type name graphql type component description
Code String Text Should be used for the code property of a type
String String Text Should be used for the name property of a type
Date Date Temporal Dates
Time Time Temporal Times
DateTime DateTime Temporal Fully qualified DateTimes
Int Int Number Integers
Float Float Number Decimal numbers
Boolean Boolean Boolean True or False

Relationship property definitions

All the options for property definitions apply to relationship property definitions unless a rule below explicitly says otherwise.

For relationships which should be annotated with their own properties, see the section on rich relationships

| name | required | default | details | examples | | ------------- | -------- | ---------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------------- | --- | --- | | type | true | Unlike for ordinary properties, this must be the name of some other type | CostCentre, System | | relationship | yes | | The name of the underlying neo4j relationship. This should be all upper case, with words separated by underscores. It is hidden from the public APIs, and requires some familiarity with the underlying data. See existing type yaml files for examples, or ask for guidance in the biz ops slack channel | HAS_TEAM_MEMBER, PAYS_FOR | | direction | yes | | Whether the underlying relationship should point away from – outgoing – or to – incoming – the record. Again, this requires familiarity with the underlying data. Please see existing examples or ask for help in the biz ops slack channel | outgoing, incoming | | hasMany | no | false | Boolean indicating whether or not the record can be related in this way to more than one other record. If in doubt, set to true, | | | | | showInactive | no | true | Boolean indicating whether it is preferable to show or hide relationships to inactive records. Can be used as a hint by any UI | | writeInactive | no | false | Boolean indicating whether it is possible to write relationships to inactive records. Can be used as a hint by any UI or system writing to Biz Ops, but does not, at this stage, prevent writing at the API level |

Custom cypher properties

Experimental Properties can contain the result of any cypher query run against the database. In order to do so add a cypher property to the property definition. This should contain a cypher query that returns either

  • one or more neo4j records, in which case the type of this property should be the same as the type of record returned, and hasMany: true may also be defined in order to indicate the return of multiple records
  • Some neo4j primitive value, in which case the type of this property should be a primitive type or an enum.

Note that, in most cases, you will want to 'anchor' your cypher query at the current record, so will need to use the this reference in it. See the GRANDstack docs for more details.

Fieldset Definitions

Fieldsets group properties with other properties that are related to them e.g. all the properties defining relationships between systems might be grouped in a relatedSystems fieldset. This structure is not used in the underlying data store, but is rather applied at a later stage by anything using the data. Fieldsets are defined using objects with the following properties

| name | required | default details | examples | | ----------- | -------- | --------------- | ------------------------------------------------------------ | --- | | heading | yes | | The text to use as a heading for the group of properties | | | description | no | | The text to use as a description for the group of properties | |

Example type file

name: Bus
pluralName: Buses
description: |
    Big red things
properties:
    code:
        type: Code
        required: true
        canIdentify: true
        canFilter: true
        description: 'Unique code/id for this item'
        label: Code
        fieldset: main
        examples:
            - LK60HPN
    isOperational:
        type: Boolean
        trueLabel: Bus runs on public roads
        falseLabel: Bus doesn't run or only runs off-road
fieldsets:
    main:
        heading: Main properties
        description: These are all essential to fill out.

String validation rules

These are expressed as regular expressions and are used to (optionally) validate values. Define a pattern in string-patterns.yaml by adding a property to the yaml file observing the following rules:

  • The property name must be in CONSTANT_CASE
  • The value must be either
    • a string that, when converted to a JavaScript regular expression, carries out the required string validation
    • a object with two properties, pattern and flags which are used to create a regular expression with flags

Enums

These are used to create enums/dropdown lists to restrict the values accepted in any field. For example ServiceTier is an enum with the values Platinum, Gold etc...

To define a new enum add a property to enums.yaml with either an array of values, or (if any of the values contain a number) a map of key/value pairs

Office:
  - OSB
  - Manila
  ...
StarRating:
  One: 1
  Two: 2
  ...

Add Description

To better understand the enum options, you can give them a description

Set as a key value pair:

Office:
    OSB: OSB description
    Manila: Manila description

Type hierarchy

To aid in developing user interfaces that must list all types, a type-hierarchy must be defined, and each type must appear in it. The type hierarchy is a map, with each property of the map defining a label and description for the grouping, and an ordered list of types that should be contained within it, e.g.:

politicalRegions
    label: Political Regions
    description: Areas of the world with some degree of political sovereignty
    types:
        - Country
        - Province
        - County
        - Borough

Rich relationships

Relationships may be annotated with their own properties e.g. a (Company)-[:EMPLOYS]->(Person) relationship may want to add a start and end date to the EMPLOYS relationship. Defining this in the schema is a 2 step process

1. Create a relationship type

This should be in its own yaml file in a relationships subdirectory.

Relationship Types are defined in individual files in a relationships. Add a .yaml file named after what you want to call your relationship type (the name is not very important, and is an implementation detail that will be hidden from the end user) e.g. EmploymentDetail.yaml. Names of types should be singular, without punctuation, and with each word/abbreviation capitalised. The file expects the following properties:

name required details examples
name yes The name of the relationship type. This should be exactly the same as the name of the file. It must begin with an upper case letter and contain only upper and lower case letters EmploymentDetail
relationship yes The name of the underlying neo4j relationship. This should be all upper case, with words separated by underscores. EMPLOYS
from yes The type the relationship starts at Company
to yes The type the relationship ends at Person
properties yes An object containing one or more Property Definition. Each key in this object defines the property name, and must be a camel-cased string, i.e. only letters and numbers, beginning with a lower case letter, with capital letters typically used to mark the beginning of a new word property names: officeLocation, supportsHttp2

2. Reference the relationship type from your types

In the example above, this would mean replacing the following:

properties:
  employees:
    label: Employees
    hasMany: true
    description: Employees of this company
    type: Person
    relationship: EMPLOYS
    direction: outgoing

with

properties:
  employees:
    label: Employees
    hasMany: true
    description: Employees of this company
    type: EmploymentDetail

Note that most properties remain unchanged - only the type and relationship properties need to change. In additional, if the relationship points between instances of the same type, it will be necessary to retain the direction property.