Core Schema Notation (CSN)
CSN (pronounced as "Season") is a notation for compact representations of CDS models — tailored to serve as an optimized format to share and interpret models with minimal footprint and dependencies.
It's similar to JSON Schema but goes beyond JSON's abilities, in order to capture full-blown Entity-Relationship Models and Extensions. This makes CSN models a perfect source to generate target models, such as OData/EDM or OpenAPI interfaces, as well as persistence models for SQL or NoSQL databases.
Anatomy
A CSN model in JSON:
{
"requires": [ "@sap/cds/common", "./db/schema" ],
"definitions": {
"some.type": { "type": "cds.String", "length": 11 },
"another.type": { "type": "some.type" },
"structured.type": { "elements": {
"foo": { "type": "cds.Integer" },
"bar": { "type": "cds.String" }
}}
},
"extensions": [
{ "extend":"Foo", "elements":{
"bar": { "type": "cds.String" }
}}
]
}
The same model in YAML:
requires:
- @sap/cds/common
- ./db/schema
definitions:
some.type: {type: cds.String, length: 11}
another.type: {type: some.type }
structured.type:
elements:
foo: {type: cds.Integer}
bar: {type: cds.String}
extensions: [
- extend: Foo
elements:
bar: {type: cds.String}
]
The same model as a plain JavaScript object:
({
requires:[ '@sap/cds/common', './db/schema' ],
definitions: {
'some.type': { type:"cds.String", length:11 },
'another.type': { type:"some.type" },
'structured.type': { elements: {
'foo': { type:"cds.Integer" },
'bar': { type:"cds.String" }
}}
},
extensions: [
{ extend:'Foo', elements:{
'bar': { type:"cds.String" }
}
],
})
For the remainder of this spec, you see examples in plain JavaScript representation with the following conventions:
({property:...}) // a CSN-specified property name
({'name':...}) // a definition's declared name
"value" // a string value, including referred names
11, true // number and boolean literal values
Properties
requires
– an array listing imported modelsdefinitions
– a dictionary of named definitionsextensions
– an array of unnamed aspectsi18n
– a dictionary of dictionaries of text translations
All properties are optional
For example, one model could contain a few definitions, while another one only contains some extensions.
References are case-sensitive
All references in properties like type
or target
use exactly the same notation regarding casing as their targets' names. To avoid problems when translating models to case-insensitive environments like SQL databases, avoid case-significant names and references. For example, avoid two different definitions in the same scope whose names only differ in casing, such as foo
and Foo
.
Literals
There are several places where literals can show up in models, such as in SQL expressions, calculated fields, or annotations.
Standard literals are represented as in JSON:
Kind | Example |
---|---|
Globals | true , false , null |
Numbers1 | 11 or 2.4 |
Strings | "foo" |
Dates2 | "2016-11-24" |
Times2 | "16:11Z" |
DateTimes2 | "2016-11-24T16:11Z" |
Records | {"foo":<literal>, ...} |
Arrays | [<literal>, ...] |
In addition, CSN specifies these special forms for references, expressions, and enum
symbols:
Kind | Example |
---|---|
Unparsed Expressions | {"=":"foo.bar < 9"} |
Enum symbols3 | {"#":"asc"} |
Remarks
1 This is as in JSON and shares the same issues when decimals are mapped to doubles with potential rounding errors. The same applies to Integer64. Use strings to avoid that, if applicable.
2 Also, as in JSON, dates, and times are represented just as strings as specified in ISO 8601; consumers are assumed to know the types and handle the values correctly.
3 As enum symbols are equal to their values, it frequently suffices to just provide them as strings. Similar to time and dates in CSN and JSON, the consumers are assumed to know the types and handle the values correctly. The
{"#":...}
syntax option is to serve cases where you have to distinguish the kind only based on the provided value, for example, in untyped annotations.
Definitions
Each entry in the definitions
dictionary is essentially a type definition. The name is the absolute, fully qualified name of the definition, and the value is a record with the definition details.
Example
({definitions:{
'Name': {type:"cds.String"},
'Currency': {type:"cds.String", length:3},
'USD': {type:"Currency"},
'Amount': {elements:{
'value': {type:"cds.Decimal", precision:11, scale:3},
'currency': {type:"Currency"},
}},
'SortOrder':{enum:{ 'asc':{}, 'desc':{} }}
}})
The name of a definition is its key in the enclosing dictionary, like in definitions
for top-level entries or in elements
for structured types and entities.
Names must:
- Be nonempty strings.
- Neither start, nor end with
.
or::
. - Not contain substrings
..
or:::
. - Not contain the substring
::
more than once.
Properties
kind
– one ofcontext
,service
,entity
,type
,action
,function
, orannotation
type
– an optional base type that this definition is derived fromelements
– optional dictionary of elements in case of structured types
Property kind
is always omitted for elements and can be omitted for top-level type definitions. These examples are semantically equivalent:
Foo1 = { type:"cds.String" }
Foo2 = { type:"cds.String", kind:"type" }
Type Definitions
Custom-defined types are entries in definitions
with an optional property kind
="type"
and the following properties.
Property | Used for |
---|---|
type | Scalar Types, Structured Types, and Associations |
elements | Structured Types |
items | Arrayed Types |
enum | Enumeration Types |
Example
({definitions: {
'scalar.type': {type:"cds.String", length:3 },
'struct.type': {elements:{'foo': {type:"cds.Integer"}}},
'arrayed.type': {items:{type:"cds.Integer"}},
'enum.type': {enum:{ 'asc':{}, 'desc':{} }}
}})
Properties
kind
– omitted or"type"
type
– the base type, this definition is derived fromelements
– optional element definitions for structured types.items
– optional definition of item types for arrayed types.enum
– an optional dictionary of enum members for enumeration types.value
– a constant literal value or calculation expressiondefault
– a default value or expressionlocalized
= true if this type was declared like foo : localized String...
– other type-specific properties, for example, a String'slength
Scalar Types
Scalar types always have property type
specified, plus optional type-specific parameter properties.
({definitions:{
'scalar.type': {type:"cds.String", length:3 },
}})
See the CDL reference docs for an overview of CDS' built-in types.
While in CDS sources you can refer to these types without prefix, they always have to be specified with their fully qualified names in CSN, for example:
({definitions: {
'Foo': { type:"cds.Integer" },
'Bar': { type:"cds.Decimal", precision:11, scale:3 },
}})
Structured Types
Structured types are signified by the presence of an elements
property. The value of elements
is a dictionary of elements
. The name is the local name of the element and the values in turn are Type Definitions.
The optional property includes
contains a list of fully qualified entity-, aspect-, or type-names. Elements, actions, and annotations from those definitions are then copied into the structured type.
({definitions:{
'structured.type': {elements:{
'foo': {type:"cds.Integer"},
'bar': {type:"cds.String"}
}}
}})
Arrayed Types
Arrayed types are signified by the presence of a property items
. The value of which is in turn a type definition that specifies the arrayed items' type.
({definitions:{
'arrayed.type': {items:{type:"cds.Integer"}}
}})
Enumeration Types
The enum
property is a dictionary of enum member elements with the name being the enum symbol and the value being a CQN literal value expression. The literal expression optionally specifies a constant val
as a literal plus optional annotations. An enumeration type can specify an explicit type
(for example, Decimal) but can also omit it and refer from given enumeration values, or String as default.
({definitions:{
'Gender': {enum:{
'male':{},
'female':{},
'non_binary': {
val: 'non-binary'
}
}},
'Status': {enum:{
'submitted': {val:1},
'fulfilled': {val:2}
}},
'Rating': {type:"cds.Decimal", enum:{
'low': {val:0},
'medium': {val:50},
'high': {val:100}
}}
}})
Entity Definitions
Entities are structured types with kind ='entity'
. In addition, one or more elements usually have property key
set to true, to flag the entity's primary key.
Example
({definitions:{
'Products': {kind:"entity", elements:{
'ID': {type:"cds.Integer", key:true},
'title': {type:"cds.String", notNull:true},
'price': {type:"Amount", virtual:true},
}}
}})
Properties
kind
– is always"entity"
elements
– as in Structured Types, optionally equipped with one or more of these boolean properties:key
– signifies that the element is (part of) the primary keyvirtual
– has this element ignored in generic persistence mappingnotNull
– the not null constraint as in SQL
includes
– as in Structured Types
View Definitions
Views are entities defined as projections on underlying entities. In CSN, views are signified by the presence of property query
, which captures the projection as a CQN expression.
Example
({definitions:{
'Foo': { kind:"entity", query: {
SELECT:{
from: {ref:['Bar']},
columns: [ {ref:['title']}, {ref:['price']} ]
}
}}
}})
Properties
kind
– mandatory; always"entity"
query
– the parsed query in CQN formatelements
– optional elements signature, omitted and inferredparams
– optional parameters
Views with Declared Signatures
Views with declared signatures have the additional property elements
filled in as in entities:
({definitions:{
'with.declared.signature': {kind:"entity",
elements: {
'title': {type:"cds.String"},
'price': {type:"Amount"}
},
query: { SELECT:{...} },
}
}})
Views with Parameters
Views with parameters have an additional property params
– an optional dictionary of parameter type definitions:
({definitions:{
'with.params': {kind:"entity",
params: { 'ID': { type: 'cds.Integer' } },
query: { SELECT:{...} },
}
}})
Projections
Use the projection
property for views if you don't need the full power of SQL. See as projection on
in CDL for restrictions.
({ definitions: {
'Foo': { kind: "entity",
projection: {
from: { ref: ['Bar'] },
columns: [ '*' ]
}
}
}})
Properties
kind
– mandatory; always"entity"
projection
– the parsed query; equivalent toquery.SELECT
, see CQNelements
– optional elements signature, omitted and inferred
Associations
Associations are like scalar type definitions with type
being cds.Association
or cds.Composition
plus additional properties specifying the association's target
and optional information like on
conditions or foreign keys
.
Basic to-one Associations
The basic form of associations are to-one associations to a designated target:
({definitions:{
'Books': { kind:"entity", elements:{
'author': { type:"cds.Association", target:"Authors" },
}},
//> an association type-def
'Currency': { type:"cds.Association", target:"Currencies" },
}})
With Specified cardinality
Add property cardinality
to explicitly specify a to-one or to-many relationship:
({definitions:{
'Authors': { kind:"entity", elements:{
'books': { type:"cds.Association", target:"Books", cardinality:{max:"*"} },
}},
}})
Property cardinality
is an object {src?,min?,max}
with...
src
set to1
give a hint to database optimizers, that a source entity always existsmin
specifying the target's minimum cardinality – default:0
max
specifying the target's maximum cardinality – default:1
In summary, the default cardinality is [0..1], which means to-one.
With Specified on
Condition
So-called unmanaged associations have an explicitly specified on
condition:
({definitions:{
'Authors': { kind:"entity", elements:{
'books': { type:"cds.Association", target:"Books", cardinality{max:"*"},
on: [{ref:['books', 'author']}, '=', {ref:['$self']}]
},
}}
}})
With Specified keys
Managed to-one associations automatically use the target's designated primary key
elements. You can overrule this by explicitly specifying alternative target properties to be used in the keys
property:
({definitions:{
'Books': {kind:"entity", elements:{
'genre': {type:"cds.Association", target:"Genres", keys:[
{ref:["category"], as:"cat"},
{ref:["name"]},
]},
}},
}})
Property keys
has the format and mechanisms of CQN projections.
Annotations
Annotations are represented as properties, prefixed with @
. This format applies to type/entity-level annotations as well as to element-level ones.
Example
({definitions:{
'Employees': {kind:"entity",
'@title':"Mitarbeiter",
'@readonly':true,
elements:{
'firstname': {type:"cds.String", '@title':"Vorname"},
'surname': {type:"cds.String", '@title':"Nachname"},
}
},
}})
Annotations are used to add custom information to definitions, the prefixed @
acts as a protection against conflicts with built-in/standard properties. They're flat lists of key-value pairs, with keys being fully qualified property names and values being represented as introduced in the section Literals and Expressions.
Aspects
In parsed-only models, the top-level property extensions
holds an array of unapplied extensions or annotations (→ see also Aspects in CDL). The entries are of this form:
ext = { extend|annotate: <name>, <property>: <value>, … }
with:
extend
orannotate
referring to the definition to be extended or annotated<property>
being the property that should be extended, for example,elements
if an entity should be extended with further elements
Extend with <named aspect>
The most basic form allows to express an extension of a named definition with another named definition (→ see Named Aspects):
csn = { extensions:[
{ extend:"TargetDefinition", includes:["NamedAspect"]}
]}
Extend with <anonymous aspect>
The form { extend:<target>, <property>: <value>, … }
allows to add elements to an existing struct definition as well as to add or override annotations of the target definition:
csn = { extensions:[
// extend Foo with @foo { ..., bar: String; }
{
extend: "Foo",
'@foo': true,
elements: {
// adds a new element 'bar'
bar: { type: "cds.String", '@bar': true },
}
},
]}
annotate with <anonymous aspect>
The form { annotate:<target>, <property>: <value>, … }
allows to add or override annotations of the target definition as well as those of nested elements:
csn = {extensions:[
// annotate Foo with @foo;
{ annotate:"Foo", '@foo':true },
// annotate Foo with @foo { boo @boo }
{ annotate:"Foo", '@foo':true, elements: {
// annotates existing element 'boo'
boo: {'@boo':true },
}},
]}
Services
Services are definitions with kind ='service'
:
({definitions:{
'MyOrders': {kind:"service"}
}})
Actions / Functions
Entity definitions (for bound actions/functions) can have an additional property actions
. The keys of these actions
are the (local) names of actions/functions. Unbound actions/functions of a service are represented as top level definitions.
Example:
({definitions:{
'OrderService': {kind:"service"},
'OrderService.Orders': {kind:"entity", elements:{...}, actions:{
'validate': {kind:"function",
returns: {type: "cds.Boolean"}
}
}},
'OrderService.cancelOrder': {kind:"action",
params:{
'orderID': {type:"cds.Integer"},
'reason': {type:"cds.String"},
},
returns: {elements:{
'ack': {enum:{ 'succeeded':{}, 'failed':{} }},
'msg': {type:"cds.String"},
}}
}
}}
}})
Properties
kind
– either"action"
or"function"
as in ODataparams
– a dictionary with the values being Type Definitionsreturns
– a Type Definition describing the response
Note: The definition of the response can be a reference to a declared type or the inline definition of a new (structured) type.
Imports
The requires
property lists other models to import definitions from. It is the CSN equivalent of the CDL using
directive.
Example
({
requires: [ '@sap/cds/common', './db/schema' ],
// [...]
})
As in Node.js the filenames are either absolute module names or relative filenames, starting with ./
or ../
.
i18n
A CSN may optionally contain a top-level i18n
property, which can contain translated texts. The expected structure is as follows:
({
i18n: {
'language-key': {
'text-key': "some string"
}
}
})
This data must be written and handled by the application, there's no out-of-the-box support for this by CAP.