OSW Schema
OSW Schema [OSWab674d663a5b472f838d8e1eb43e6784] | |
---|---|
ID | OSWab674d663a5b472f838d8e1eb43e6784 |
UUID | ab674d66-3a5b-472f-838d-8e1eb43e6784 |
Label | OSW Schema |
Machine compatible name | OswSchema |
Statements (outgoing) | |
Statements (incoming) | |
|
|
Keywords |
Description
Documentation about to OSW data schema
Item | |
---|---|
Type(s)/Category(s) | Article |
CreativeWork |
---|
Article |
---|
Schema for the dynamical Composition of nested class structures
Introduction
Transforming Semantic Mediawiki into a knowledge base with structured data used to require PageForms and multiple template per class:
- Template to store the data and render the page
- Form to edit the page
- Form to query for instances of the class
- Template to render the query results
- Subtemplate to handle complex data
Reusing structures with this approach is difficult and storing data in wikicode leads to a significant lock-in regarding editing those data using existing stardards and tools.
The new concept is based on strictly storing data in json and use wikicode just for structured text and render templates. Insteaf of distributing content among multiple pages all relevent content is stored withing different slots of the same page. Additional, inheritance is highly supported and enables a broad reuse of any structure.
Overview
Why are Categories (Classes) different from Items (Instances)?
- Pro
- Reflects rdf(s) and owl standard
- Reflects OOP in Python and other programming languages, only (?) Javascript supports a similar concepts with prototypes
- Compatible with SMW features like rdf-export and Inferencing
- Contra
- User needs to decide before creating a term or move the term later to a different namespace (renaming/redirect)
Slots
Key | Model | Description |
---|---|---|
main | wikitext | default content slot, rendered between the page header and footer |
jsonschema | json | stored within a category (=class) page, defining the schema for the jsondata slot of any category member (instance) |
jsondata | json | structured data |
schema_template | text | stored within a metacategory/-class, contains a handlebars template to build the jsonschema of a class from its jsondata slot |
data_template | wikitext | stored within a category (=class) page, defining how the jsondata attributes of any category member (instance) are mapped to semantic properties |
header_template | wikitext | stored within a category (=class) page, renders the page header of any category member (instance) |
footer_template | wikitext | stored within a category (=class) page, renders the page footer of any category member (instance) |
header | wikitext | renders the page header |
footer | wikitext | renders the page footer |
template/internal | wikitext | hidden content, not rendered. Can be used to call parser functions or lua modules |
Meta-Schemas
Category (Class) ist the default Metacategory / -class for all categories / classes. Its slot schema_template contains a handlebars template that sets schema attributes like title, allOf, description, etc. from the user generated jsondata. Additional Metacategories can be created as subclasses of Category (Class) to simplify the creation of complex schemas, e. g. Category:OSWecff4345b4b049218f8d6628dc2f2f21. This feature is compareable to python metaclasses. Template:ELN/Editor/DrawIOMatecategories /-classes contain a handlebars template within the schema_template slot. The templated is evaluated with the jsondata-slot content to create / update the jsonschema-slot content of any derivated class on every edit.
Json-Schema
Base
https://json-schema.org/ (Draft 4)
A Jsonschema can reference other schemas. This is equivalent to [[Subcategory of::Entity]]
(Semantic Mediawiki) and owl:subclass_of
{
"title": "MyEntitySubclass",
"type": "object",
"allOf": [{"$ref": "/wiki/Category:Entity?action=raw&slot=jsonschema"}]
}
Json-Editor
https://github.com/json-editor/json-editor, which is used to render edit & query forms based on provided jsonschema, adds additional keywords and options.
- Basic features: see https://github.com/json-editor/json-editor#readme
- Further details: https://github.com/json-editor/json-editor/blob/master/README_ADDON.md
- Demos: https://json-editor.github.io/json-editor/, https://pmk65.github.io/jedemov2/dist/demo.html
- Examples: https://github.com/json-editor/json-editor/tree/master/tests/pages
Autocompletion
- https://github.com/trevoreyre/autocomplete
- https://github.com/corejavascript/typeahead.js/blob/master/doc/jquery_typeahead.md (currently not supported)
Enables autocompletion in input fields. Configuration see Additional options
Custom Extensions
Embedded i18n support:
keywords title
and description
can be extended with additional keywords title*
and description*
, which hold and object with lang-keys (de, en, etc.) pointing to the translated strings.
{
"title": "Default Title",
"title*": {"en": "Title (en)", "de": "Titel (de)"}
}
Additional keywords
Key | Alias | Subkeys | Value | Description | Note |
---|---|---|---|---|---|
eval_template | evaluation template for the current json object (while 'template' is used by jsoneditor in the UI) | eval_templates are expanded before the json data is passed to render templates and property mapping | |||
type | mediawiki | uses the wiki template parser. Cannot handle objects and arrays => non-literals get stripped | |||
" | mustache | uses the lua mustache template parser https://github.com/OpenSemanticLab/lustache/tree/scribunto-module-pages. Can handle objects and arrays | https://mustache.github.io/, https://stackblitz.com/edit/mustache-tester?file=index.js | ||
" | mustache-wikitext | applies mustache first, then wikitext | wikitext parts containing {{ need to be wrapped inside {{=<% %>=}} and<%={{ }}=%>
| ||
mode | <none> | the given template will be used to render the json object and store it's semantic data | |||
" | render | the given template will be used to render the json object | |||
" | store | the given template will be used to store semantic data | |||
value | <string> | ||||
page | <wiki page> | ||||
slot | |||||
url |
Key | Alias | Subkeys | Value | Description | Note |
---|---|---|---|---|---|
type | - | an array of category pages | if defined, the given category will be used to render the json object and store it's semantic data |
Additional options
Key | Subkeys | SubSubkeys | Value | Example | Description | Note |
---|---|---|---|---|---|---|
conditional_visible | ||||||
modes | <array> | ["default", "query"] | Display this field only in the selected modes of the editor | |||
conditional_hide | tbd | |||||
autocomplete | buildin option | |||||
mode | smw | query mode | for now only supports semantic mediawiki | |||
query | [[HasLabel::~*{{input}}*]]|?HasLabel=label
|
handlebars query template | ||||
range | Category:Item | creates a static query [[Category:Item]][[HasLabel::~*{{input}}*]]|?HasLabel=label
|
||||
property | Property:HasLabel | existing value of the property [[HasLabel::+]][[HasLabel::~*{{input}}*]]?HasLabel=value
|
||||
render_template | how to display query results in the suggestion list | |||||
type | <array> | ["handlebars", "wikitext"] | template engines are applied in the specified order. wikitext will result in parse-API calls, which is not recommanded for large result sets | |||
value | template string | [[{{result.fulltext}}]]
|
the actual template string. Pure handlebars templates can contain html tags like links (<a> ) and images (<img> ), wikitext templates need to use the wiki-syntax [[ ]]
|
wiki-links to categories need a : prefix: [[:{{result.fulltext}}]]
| ||
label_template | how to display the item after getting selected by the user | |||||
type | <array> | ["handlebars"] | only handlebars supported | |||
value | template string | result.printouts.label.[0]
|
||||
role | query | {"filter": "min|max|eq"} | {"filter": "min"} | Creates a semantic mediawiki query for a numerical property, e. g. if the property maps to "HasNumber", the filter is "min" and the user provided value is 3 this results in the query ">3Property "HasNumber" (as page type) with input value ">3" contains invalid characters or is incomplete and therefore can cause unexpected results during a query or annotation process." |
Autocomplete
Default setting (planned, see also Object Properties and Data / Annotation Properties)
- for data / annotation properties: existing values of the property:
{{#ask:[[<Property>::+]]|?<Property>=value}}
- for object properties: existing instances of the range category:
{{#ask:[[Category:<range>]]}}
Shortcuts
category
: Populates the field with instances of the given category (and its subcategories)
{
"type": "string",
"format": "autocomplete",
"options":{
"autocomplete": {
"category":"Category:X"
}
}
}
Remote / external data import
{
"...": {},
"data_source_maps": [
{
"id": "pubchem.ncbi.nlm.nih.gov",
"source": "https://pubchem.ncbi.nlm.nih.gov/rest/pug_view/data/compound/{{pubchem_cid}}/JSON",
"label": "PubChem",
"required": [
"pubchem_cid"
],
"object_map": {
"cas_numbers": "$..[?(@.TOCHeading==='CAS')].Information..Value.StringWithMarkup..String"
}
}
],
"properties": {
"pubchem_cid": {
"title": "PubChem CID",
"type": "string"
},
"cas_numbers": {
"type": "array",
"...": {}
},
"...": {}
}
}
Tests could be run on various playgrounds:
- Handlebars template: [1]
Search
Properties can be marked as inputs for the categories query form by adding the conditional_visible option. Therefor a context mapping to a SWM Property is mandatory (see #JSON-LD). With "hidden": true
the property is only visible in the query form. With "hidden": false
it's visible both in the edit and in the query form.
{
"@context":
"...",
{
"query_label": "Property:HasLabel"
}
],
"...": {},
"properties": {
"query_label": {
"type": "string",
"options": {
"hidden": true,
"conditional_visible": {
"modes": [
"query"
]
}
}
}
}
}
The default comperator for text properties is ~
(like), for all other properties =
(equal). The comperator can be defined with the role filter option, e. g. "min"
for >
.
{
"@context": [
{
"start_date_min": "Property:HasStartDateAndTime"
}
],
"...": "...",
"properties": {
"start_date_min": {
"type": "string",
"format": "datetime-local",
"options": {
"flatpickr": {},
"conditional_visible": {
"modes": "query"
},
"role": {
"query": {
"filter": "min"
}
}
}
}
}
}
JSON-LD
json-ld jsonschema: https://github.com/json-ld/json-ld.org/blob/main/schemas/jsonld-schema.json
Local playground: https://wiki-dev.open-semantic-lab.org/w/extensions/MwJson/json-ld/playground/index.html
References
json-ld should be embedded into jsonschema, but has its own referencing mechanism:
{
"@context": [
"/wiki/Category:Entity?action=raw&slot=jsonschema",
{
"Property": {"@id": "https://wiki-dev.open-semantic-lab.org/id/Property-3A", "@prefix": true},
"manufacturer": "Property:HasManufacturer"
}
],
"title": "MyEntitySubclass",
"type": "object",
"allOf": [{"$ref": "/wiki/Category:Entity?action=raw&slot=jsonschema"}],
"properties": {
"manufacturer": {
"type": "string"
}
}
}
Example: [4]
Property mapping
Properties with a local definition (SMW Property) are automatically mapped. Jsondata of an instance of the category could then be provided with an json-ld context:
{
"@context": [
{
"@version": 1.1,
"wiki": "https://wiki-dev.open-semantic-lab.org/id/",
"Property": {"@id": "wiki:Property-3A", "@prefix": true},
"@vocab": "https://wiki-dev.open-semantic-lab.org/id/Property-3A",
"myproperty": "Property:MyProperty",
"myproperty2": "wiki:Property-3AMyProperty",
"myproperty3": "MyProperty",
"Item": {"@id": "wiki:Item-3A", "@prefix": true},
"myObjectProperty": {"@id": "Property:MyObjectProperty", "@type": "@id"}
}
],
"myproperty": "Works by using '@prefix': true (preferred)",
"myproperty2": "Works by using 'wiki' prefix with terminating '/'",
"myproperty3": "Works by using '@vocab'",
"myObjectProperty": "Item:123456"
}
Currently there seems no way to express that a property has two ids (e. g. with "label": {"@id": ["property:HasLabel", "skos:prefLabel"]}
): https://github.com/json-ld/json-ld.org/issues/160
As a workaround, an additional context notation is provided: <property>* pointing to a list of additional "@id" mappings:
{
"@context": [
{
"@version": 1.1,
"skos": "https://www.w3.org/TR/skos-reference/",
"wiki": "https://wiki-dev.open-semantic-lab.org/id/",
"Property": {"@id": "wiki:Property-3A", "@prefix": true},
"label": "skos:prefLabel",
"label*": "Property:HasLabel",
"label**": "Property:Display_title_of"
}
],
"label": "Maps externally to skos:prefLabel and internally to Property:HasLabel"
}
Object Properties and Data / Annotation Properties
Properties default to data / annotation properties (value is a literal). Object properties (value is an identifier/reference to another object) can by defined by adding "@type": "@id"
.
Subobjects
If the value of a mapped property is an object (after expanding all eval_templates), it will get stored as a smw subobject with an id derivated from the field uuid
, a display title from label
and a category from type
(if provided).
Example: can be selected with [[MyObjectProperty.MyProperty::myvalue]]
{
"@context": [
{
"@version": 1.1,
"wiki": "https://wiki-dev.open-semantic-lab.org/id/",
"Property": {"@id": "wiki:Property-3A", "@prefix": true},
"myproperty": "Property:MyProperty",
"Item": {"@id": "wiki:Item-3A", "@prefix": true},
"myObjectProperty": {"@id": "Property:MyObjectProperty", "@type": "@id"}
}
],
"myObjectProperty": {
"uuid": "2ea5b605-c91f-4e5a-9559-3dff79fdd4a5",
"label": "MySubobject",
"myproperty": "myvalue"
}
}
Labels and i18n
i18n language keys can be embedded in to an label object to create a language tagged string
{
"@context": [
{
"@version": 1.1,
"skos": "https://www.w3.org/TR/skos-reference/",
"label": {"@id": "skos:prefLabel", "@container": "@language"},
"label2": {"@id": "skos:prefLabel"},
"label_text": {"@id": "@value"},
"label_lang_key": {"@id": "@language"}
}
],
"label": {"en": "'Text' gets transformed to 'Text@en' by applying @container"},
"label2": {"label_text": "'Text' gets transformed to 'Text@de' by subkeys @id's", "label_lang_key": "de"}
}
Recursive Parsing
Module:Category
Called from Category:<UUID>@template
Synchonize Category:<UUID>@jsondata.subclass_of with Category:<UUID>@jsonschema.allOfExpand Category:Category@header_template with jsondata parametersRender Category:<UUID>@mainExpand Category:Category@footer_template with jsondata parameters
Module:Entity
Called from item@template, item = Item:<UUID>
Recursion
- For each Item:<UUID>@jsondata.osl_category as category:
- For each category@jsondata.osl_category or category@jsonschema.allOf as supercategory:
- For each supercategory@jsondata.osl_category or supercategory@jsonschema.allOf as supersupercategory:
- ...
- Expand supercategory@header_template with item@jsondata parameters. Fallback: Render infobox
- Expand supercategory@data_template with item.jsondata parameters. Fallback: Use Json-LD mapping within category:jsonschema
- For each supercategory@jsondata.osl_category or supercategory@jsonschema.allOf as supersupercategory:
- If category@header_template: Expand category@header_template with item@jsondata parameters
- Else: Render infobox with all attribute-value pairs
- Expand category@data_template with item.jsondata parameters. Fallback: Use Json-LD mapping within category:jsonschema
- For each category@jsondata.osl_category or category@jsonschema.allOf as supercategory:
- Render item@main
- footer...
Data Storing
- template specified: use template
- category specified:
- category@data_template specified: use data_template
- Use Json-LD mapping
- mapping specified: Store semantic property
- Literal value: store value
- Object value
- Property has type text/code: Store json string
- osl_category / osl_template specifided: see below
- Don't store semantic property
- mapping specified: Store semantic property
Nested objects within item@jsondata are handled
- osl_category: same handling as the root object, but:
- Rendering with category@header_template. Fallback: Nested info box
- Data storing with category@data_template. Fallback: Creating a subobject with Json-LD mapping + (inverse) semantic relation to the root object
- osl_template: expand the template, return value is asigned to the property
Python Code Generation
see https://github.com/OpenSemanticLab/osw-python
Statement
s | p/s | o/p/s | o/p | o... |
---|---|---|---|---|
PCB | contains | Lead | ||
contains | HasMassConcentration | 10% | ||
HasMassConcentration | HasPrecision | 1% | ||
PCB | contains | Gold |
File Handling
Copy-Policy
drop: do not copy the file
copy: copy the file and store the reference to it
copy-ref: store the referece to the original file
ask-on-edit: store the reference but ask the user to copy the original file when he tries to edit it (current_page != creation_page)
Meta-Data
HasProject=project inherit: permissions from project
HasCreationPage=creation_page: wiki page within this file was created
HasEditPage=edit_page: wiki page within this file was edited
HasCreator=creator: initial creator of the file
HasEditor=editor: editors of the file
Links
jsondata
type |
| |||||
---|---|---|---|---|---|---|
uuid | "ab674d66-3a5b-472f-838d-8e1eb43e6784" | |||||
label |
| |||||
description |
| |||||
name | "OswSchema" | |||||
attachments |
|