语义-DSL

用于快速开发语义支持的领域特定语言 (DSL)

安装

pip install semdsl

使用

为了说明semdsl的使用，我们将创建一个简单的 LinkML 架构，用于部分 Clue棋盘游戏，特别是用于表示关于谁犯下了罪行、在哪里以及用什么的假设。

我们将使用 语法提示 注释 该架构，这些提示可用于生成DSL的语法。

```python
>>> schema = """
... id: https://example.org/clue
... name: clue
... imports:
...   - https://w3id.org/linkml/types
... classes:
...   ClueHypothesis:
...     attributes:
...       person:   # e.g. Colonel Mustard
...         annotations:
...           grammar.main: "WORD WORD"
...       location: # e.g. Kitchen
...         annotations:
...           grammar.main: "WORD"
...       weapon:   # e.g. Candlestick
...         annotations:
...           grammar.main: "WORD"
...     annotations:
...       grammar.main: >-
...         "<" person "in the" location "with the" weapon ">" 
... """

我们的想法是能够使用类似 <Colonel Mustard in the Kitchen with the Candlestick> 的字符串来表示假设。

然后我们可以使用 DSLEngine 类来加载架构并生成语法

>>> from semdsl import DSLEngine
>>> engine = DSLEngine()
>>> engine.load_schema(schema)
>>> print(engine.lark_serialization)
from lark import Lark
...
class_clue_hypothesis : "<" person "in the" location "with the" weapon ">"
person : WORD WORD
location : WORD
weapon : WORD
...

默认是Lark语法。

然后您可以使用生成的语法将序列化数据解析成符合架构的pydantic对象

>>> obj = engine.parse_as_object('<Colonel Mustard in the Kitchen with the Candlestick>')
>>> print(obj.location)
Kitchen
>>> print(obj.json())
   {"person": "Colonel Mustard", "location": "Kitchen", "weapon": "Candlestick"}

自动分配生成规则

在上一个示例中，我们看到了如何使用语法规则 注释 已存在的架构

然而，我们还可以从架构本身 生成 语法规则。

这是通过在类上使用 grammar.main 注解，然后在该类的属性上使用 grammar.auto 注解来完成的。例如

>>> schema = """
... id: https://example.org/clue
... name: clue
... imports:
...   - https://w3id.org/linkml/types
... classes:
...   ClueHypothesis:
...     attributes:
...       person:   # e.g. Colonel Mustard
...       location: # e.g. Kitchen
...       weapon:   # e.g. Candlestick
... """

现在我们将创建一个新的引擎并加载架构，并生成一个新的“功能式”语法

>>> engine = DSLEngine() ## create new DSLEngine
>>> engine.load_schema(schema)
>>> print(engine.lark_serialization)
from lark import Lark
...
class_clue_hypothesis : "ClueHypothesis(" slot_clue_hypothesis__person? slot_clue_hypothesis__location? slot_clue_hypothesis__weapon? ")"
slot_clue_hypothesis__person : "person=" TYPE_STRING
slot_clue_hypothesis__location : "location=" TYPE_STRING
slot_clue_hypothesis__weapon : "weapon=" TYPE_STRING
...

然后您可以使用生成的语法将字符串解析成对象

>>> obj = engine.parse_as_object('ClueHypothesis(person="Colonel Mustard" location="Kitchen" weapon="Candlestick")')
>>> print(obj.location)
Kitchen

添加额外的语义

您可以使用以下元模型元素

• 类URI
• 槽URI

将URI分配给您方案中的类和槽，这可以在RDF序列化中使用。

在此，我们扩展了我们的Clue方案，为主类的槽范围添加了类

>>> schema = """
... id: https://example.org/clue
... name: clue
... prefixes:
...   linkml: https://w3id.org/linkml/
...   clue: https://example.org/clue/
...   schema: http://schema.org/
...   prov: http://www.w3.org/ns/prov#
...   dbpedia: http://dbpedia.org/ontology/
... imports:
...   - linkml:types
... classes:
...   NamedThing:
...     class_uri: schema:Thing
...     attributes:
...       id:
...         identifier: true
...       range: uriorcurie
...   Person:
...     class_uri: schema:Person
...     is_a: NamedThing
...   Location:
...     class_uri: schema:Location
...     is_a: NamedThing
...   Weapon:
...     class_uri: dbpedia:Weapon
...     is_a: NamedThing
...   ClueHypothesis:
...     class_uri: prov:Action
...     tree_root: true
...     attributes:
...       person:   # e.g. Colonel Mustard
...         slot_uri: prov:wasAssociatedWith
...         range: Person
...         annotations:
...           grammar.main: TYPE_URIORCURIE
...       location: # e.g. Kitchen
...         slot_uri: prov:atLocation
...         range: Location
...         annotations:
...           grammar.main: TYPE_URIORCURIE
...       weapon:   # e.g. Candlestick
...         slot_uri: prov:used
...         range: Weapon
...         annotations:
...           grammar.main: TYPE_URIORCURIE
...     annotations:
...       grammar.main: >-
...         "<" person "in the" location "with the" weapon ">"
... """

现在解析并导出到文件。这次输入字符串使用CURIE表示Clue假设中的不同事物。

>>> engine = DSLEngine()
>>> engine.load_schema(schema)
>>> obj = engine.parse_as_object("< clue:ColonelMustard in the clue:Kitchen with the clue:Candlestick >")
>>> import yaml
>>> with open("tests/output/clue-output.yaml", "w", encoding="utf-8") as f:
...     yaml.dump(obj.dict(), f)

从这里我们可以使用LinkML将其转换为RDF序列化

cd clue-output.yaml
linkml-convert clue-output -s clue_model.yaml -t ttl

结果

@prefix clue: <https://example.org/clue/> .
@prefix prov: <http://www.w3.org/ns/prov#> .

[] a prov:Action ;
    prov:atLocation clue:Kitchen ;
    prov:used clue:Candlestick ;
    prov:wasAssociatedWith clue:ColonelMustard .

命令行界面

semdsl --help

限制

限于Lark语法

目前，semdsl仅支持Lark语法。该框架旨在允许可扩展性，例如扩展到ANTLR，但目前不支持。