270 lines
13 KiB
TypeScript
270 lines
13 KiB
TypeScript
import type { BlockView } from '../blockView'
|
||
import type { IndexEntry, PropertyRow, RuleContext, Segment } from '../contract'
|
||
import { describe, expect, it } from 'vitest'
|
||
import { fileAttributes } from '../attributes'
|
||
import { toBlockViews } from '../blockView'
|
||
import { diffBlockTrees } from '../diff/blockDiff'
|
||
import { createPythonAdapter } from '../languages/python'
|
||
import { flatSegments, propertyLines, readFixture, WASM_DIR, withBlocks } from '../testSupport'
|
||
import { pythonZhTranslator } from '../translators/pythonZh'
|
||
import { celeryRule } from './celery'
|
||
import { dataclassRule } from './dataclass'
|
||
import { enumRule } from './enum'
|
||
import { fastapiRule } from './fastapi'
|
||
import { MODEL_INDEX, pydanticRule } from './pydantic'
|
||
import { sqlalchemyRule } from './sqlalchemy'
|
||
|
||
const py = createPythonAdapter(WASM_DIR)
|
||
const AIGC = readFixture('aigcMetadata.py')
|
||
const SHARE = readFixture('createShareRequest.py')
|
||
/** 和扩展里注册的顺序一样。 */
|
||
const RULES = [sqlalchemyRule, celeryRule, enumRule, dataclassRule, pydanticRule, fastapiRule]
|
||
|
||
async function views(source: string, reports: string[] = [], lookup?: RuleContext['lookup']): Promise<BlockView[]> {
|
||
return withBlocks(py, 'python', source, blocks =>
|
||
toBlockViews(blocks, { source, language: py, markers: [] }, pythonZhTranslator, { rules: RULES, report: message => reports.push(message), lookup }))
|
||
}
|
||
|
||
async function indexEntries(source: string): Promise<IndexEntry[]> {
|
||
return withBlocks(py, 'python', source, blocks => MODEL_INDEX.entries(blocks[0].node.tree.rootNode, { path: 'sample.py', source, exists: () => false }))
|
||
}
|
||
|
||
function properties(view: BlockView | undefined): PropertyRow[] {
|
||
if (!view || !('properties' in view))
|
||
throw new Error(`块 ${view?.id} 不是属性表`)
|
||
return view.properties
|
||
}
|
||
|
||
function named(list: BlockView[], name: string): BlockView {
|
||
const found = list.find(view => 'properties' in view && propertyLines(view.properties)[0] === `名称: ${name}`)
|
||
if (!found)
|
||
throw new Error(`没有 ${name}`)
|
||
return found
|
||
}
|
||
|
||
/** 属性表里所有带下标的词（连表格里的格子）。 */
|
||
function linkedSegments(rows: PropertyRow[]): Segment[] {
|
||
const pieces = rows.flatMap(row => [
|
||
...('markdown' in row.value ? [] : row.value),
|
||
...(row.items ?? []).flatMap(item => item.cells.flat()),
|
||
])
|
||
return flatSegments(pieces).filter(segment => segment.offset !== undefined)
|
||
}
|
||
|
||
describe('pydantic 库规则（Python）', () => {
|
||
it('luffy 的 AIGCMetadata：数据模型卡片，7 个无默认值字段，选项是严格、禁止多余字段；model_config 不单成一块', async () => {
|
||
const reports: string[] = []
|
||
const list = await views(AIGC, reports)
|
||
expect(reports).toEqual([])
|
||
const metadata = named(list, 'AIGCMetadata')
|
||
expect(metadata.kind).toBe('pydantic.model')
|
||
expect(metadata.badge).toEqual({ label: '数据模型', color: { dark: '#c678dd', light: '#8b3db5' }, icon: 'braces' })
|
||
expect(propertyLines(properties(metadata))).toEqual([
|
||
'名称: AIGCMetadata',
|
||
'说明: 生成内容来源标识。',
|
||
'访问级别: 公共',
|
||
'继承: BaseModel',
|
||
'选项: 严格 · 禁止多余字段',
|
||
'字段: 7 个',
|
||
' Label | 字符串 | | ',
|
||
' ContentProducer | 字符串 | | ',
|
||
' ProduceID | 字符串 | | ',
|
||
' ReservedCode1 | 字符串 | | ',
|
||
' ContentPropagator | 字符串 | | ',
|
||
' PropagateID | 字符串 | | ',
|
||
' ReservedCode2 | 字符串 | | ',
|
||
])
|
||
expect(properties(metadata).at(-1)?.columns).toEqual(['字段', '类型', '默认值', '说明'])
|
||
expect(metadata.children).toEqual([])
|
||
})
|
||
|
||
it('luffy 的 TextTrace：可空字段的默认值是空', async () => {
|
||
const trace = named(await views(AIGC), 'TextTrace')
|
||
expect(propertyLines(properties(trace))).toEqual([
|
||
'名称: TextTrace',
|
||
'说明: 生成文本及其来源标识。',
|
||
'访问级别: 公共',
|
||
'继承: BaseModel',
|
||
'字段: 7 个',
|
||
' content_key | 字符串 | | ',
|
||
' text | 字符串 | | ',
|
||
' reasoning_content | 字符串 | 空 | 空 | ',
|
||
' metadata | AIGCMetadataEnvelope | | ',
|
||
' model_id | 字符串 | 空 | 空 | ',
|
||
' served_model | 字符串 | 空 | 空 | ',
|
||
' generated_at | datetime | 空 | 空 | ',
|
||
])
|
||
})
|
||
|
||
it('luffy 的 CreateShareRequest：Field 的说明和长度约束写进行，校验方法仍是子块', async () => {
|
||
const request = named(await views(SHARE), 'CreateShareRequest')
|
||
expect(propertyLines(properties(request))).toEqual([
|
||
'名称: CreateShareRequest',
|
||
'访问级别: 公共',
|
||
'继承: BaseModel',
|
||
'选项: 可从对象读取',
|
||
'字段: 6 个',
|
||
' target_type | ShareTargetType | ShareTargetType.SESSION_ARTIFACT | ',
|
||
' session_id | 字符串 | 至少 1 个字符 | 作品所属 session id',
|
||
' preview_url | 字符串 | 至少 1 个字符 · 最长 2048 | 作品当前 preview URL(前端从预览态取)',
|
||
' title | 字符串 | 至少 1 个字符 · 最长 255 | 分享卡片标题',
|
||
' description | 字符串 | 空 | 空 · 最长 500 | ',
|
||
' thumbnail_url | 字符串 | 空 | 空 · 最长 2048 | ',
|
||
])
|
||
const lines = SHARE.split('\n')
|
||
expect(request.children.map(child => [child.kind, lines[child.startLine].trim()])).toEqual([
|
||
['function', '@field_validator("preview_url", "thumbnail_url")'],
|
||
])
|
||
expect(request.code).toContain('def validate_http_scheme')
|
||
})
|
||
|
||
it('同一文件里继承模型的也是模型；TypedDict、NamedTuple、普通类、没导入 pydantic 的不是', async () => {
|
||
const withChild = `${AIGC}\n\nclass Child(AIGCMetadata):\n extra: int = 1\n`
|
||
const child = named(await views(withChild), 'Child')
|
||
expect(child.kind).toBe('pydantic.model')
|
||
expect(propertyLines(properties(child)).slice(0, 4)).toEqual([
|
||
'名称: Child',
|
||
'访问级别: 公共',
|
||
'继承: AIGCMetadata',
|
||
'字段: 1 个',
|
||
])
|
||
expect(propertyLines(properties(child)).at(-1)).toBe(' extra | 整数 | 1 | ')
|
||
|
||
const others = [
|
||
'from typing import NamedTuple, TypedDict',
|
||
'from pydantic import BaseModel',
|
||
'',
|
||
'class ViaDict(TypedDict):',
|
||
' x: str',
|
||
'',
|
||
'class ViaTuple(NamedTuple):',
|
||
' x: str',
|
||
'',
|
||
'class Plain:',
|
||
' x: str',
|
||
'',
|
||
'class Model(BaseModel):',
|
||
' x: str',
|
||
'',
|
||
].join('\n')
|
||
expect((await views(others)).map(view => view.kind)).toEqual(['import', 'import', 'class', 'class', 'class', 'pydantic.model'])
|
||
expect((await views('class Model(BaseModel):\n x: str\n')).map(view => view.kind)).toEqual(['class'])
|
||
})
|
||
|
||
it('别名、Field 的默认值和约束、说明优先于注释；ClassVar 和没注解的类属性留下', async () => {
|
||
const source = [
|
||
'from pydantic import BaseModel as Model, Field',
|
||
'from typing import ClassVar',
|
||
'',
|
||
'class Limits(Model):',
|
||
' """界限。"""',
|
||
' registry: ClassVar[dict] = {}',
|
||
' # 个数',
|
||
' n: int = Field(default=0, ge=0, le=10)',
|
||
' # 忽略这行',
|
||
' name: str = Field(min_length=1, max_length=255, description="名字")',
|
||
' tags: list[str] = Field(default_factory=list)',
|
||
' step: int = Field(default=3, multiple_of=3, pattern="[0-9]+", unknown=1)',
|
||
' label = "a"',
|
||
'',
|
||
' def ping(self) -> None:',
|
||
' return None',
|
||
'',
|
||
].join('\n')
|
||
const limits = named(await views(source), 'Limits')
|
||
expect(propertyLines(properties(limits))).toEqual([
|
||
'名称: Limits',
|
||
'说明: 界限。',
|
||
'访问级别: 公共',
|
||
'继承: Model',
|
||
'字段: 4 个',
|
||
' n | 整数 | 0 · ≥ 0 · ≤ 10 | 个数',
|
||
' name | 字符串 | 至少 1 个字符 · 最长 255 | 名字',
|
||
' tags | 列表[字符串] | 每次新建 list() | ',
|
||
' step | 整数 | 3 · 3 的倍数 · 匹配 "[0-9]+" · unknown=1 | ',
|
||
])
|
||
const lines = source.split('\n')
|
||
expect(limits.children.map(child => lines[child.startLine].trim())).toEqual([
|
||
'registry: ClassVar[dict] = {}',
|
||
'label = "a"',
|
||
'def ping(self) -> None:',
|
||
])
|
||
})
|
||
|
||
it('筛选：种类是 pydantic.model、叫「数据模型」；类型里的名字带着源码下标', async () => {
|
||
const list = await views(AIGC)
|
||
const metadata = named(list, 'AIGCMetadata')
|
||
expect(metadata.attributes).toEqual({ kind: 'pydantic.model', access: 'public' })
|
||
expect(fileAttributes(list, pythonZhTranslator.attributeNames).find(attribute => attribute.id === 'kind')?.values)
|
||
.toContainEqual({ id: 'pydantic.model', name: '数据模型', count: 2 })
|
||
const linked = linkedSegments(properties(metadata))
|
||
expect(linked.map(segment => segment.text)).toEqual(expect.arrayContaining(['AIGCMetadata', 'BaseModel', 'Label']))
|
||
for (const segment of linked)
|
||
expect(AIGC.slice(segment.offset, (segment.offset ?? 0) + segment.text.length), segment.text).toBe(segment.text)
|
||
const trace = named(list, 'TextTrace')
|
||
const typeNames = linkedSegments(properties(trace)).filter(segment => segment.text === 'AIGCMetadataEnvelope')
|
||
expect(typeNames.map(segment => AIGC.slice(segment.offset, (segment.offset ?? 0) + segment.text.length))).toEqual(['AIGCMetadataEnvelope'])
|
||
const share = named(await views(SHARE), 'CreateShareRequest')
|
||
expect(linkedSegments(properties(share)).map(segment => segment.text)).toEqual(expect.arrayContaining(['ShareTargetType']))
|
||
})
|
||
|
||
it('别的文件里的父类：索引认本地能确定的模型；查不到先当普通类，查到再换成数据模型卡片', async () => {
|
||
expect(MODEL_INDEX.mentions(AIGC)).toBe(true)
|
||
expect(MODEL_INDEX.mentions('from luffy_agent.model.aigc import AIGCMetadata\n\nclass Child(AIGCMetadata):\n x: int\n')).toBe(false)
|
||
const entries = await indexEntries(`${AIGC}\n\nclass Plain:\n x: str\n`)
|
||
expect(entries.map(entry => [entry.key, entry.value])).toEqual([['AIGCMetadata', 'model'], ['TextTrace', 'model']])
|
||
const alias = await indexEntries('from pydantic import BaseModel as Model\n\nclass Alias(Model):\n x: int\n')
|
||
expect(alias.map(entry => entry.key)).toEqual(['Alias'])
|
||
|
||
const child = [
|
||
'from luffy_agent.model.aigc import AIGCMetadata',
|
||
'',
|
||
'class Child(AIGCMetadata):',
|
||
' extra: int = 1',
|
||
'',
|
||
'class Other(NotAModel):',
|
||
' x: int',
|
||
'',
|
||
].join('\n')
|
||
const pending = await views(child)
|
||
expect(pending.map(view => view.kind)).toEqual(['import', 'class', 'class'])
|
||
|
||
const model = (index: { id: string }, node: { text: string }): boolean => index.id === MODEL_INDEX.id && node.text === 'AIGCMetadata'
|
||
const found = await views(child, [], (index, node) => model(index, node)
|
||
? { key: node.text, name: { startIndex: node.startIndex, endIndex: node.endIndex }, value: 'model' }
|
||
: undefined)
|
||
const card = named(found, 'Child')
|
||
expect(card.kind).toBe('pydantic.model')
|
||
expect(propertyLines(properties(card)).slice(-2)).toEqual(['字段: 1 个', ' extra | 整数 | 1 | '])
|
||
expect(named(found, 'Other').kind).toBe('class')
|
||
})
|
||
|
||
it('给模型加一个字段、改一个字段的默认值：字段行标新增，默认值那一格标改过', async () => {
|
||
const after = AIGC
|
||
.replace(' ReservedCode2: str\n', ' ReservedCode2: str\n Extra: str\n')
|
||
.replace(' reasoning_content: str | None = None\n', ' reasoning_content: str | None = ""\n')
|
||
expect(after).not.toBe(AIGC)
|
||
const before = await views(AIGC)
|
||
const now = await views(after)
|
||
const diff = diffBlockTrees({ blocks: before, source: AIGC }, { blocks: now, source: after })
|
||
const metadata = named(now, 'AIGCMetadata')
|
||
const trace = named(now, 'TextTrace')
|
||
const metadataChange = diff.blocks[metadata.anchor]
|
||
const traceChange = diff.blocks[trace.anchor]
|
||
expect(metadataChange).toMatchObject({ change: 'mod', own: true })
|
||
expect(traceChange).toMatchObject({ change: 'mod', own: true })
|
||
|
||
const fieldsIndex = properties(metadata).findIndex(row => row.label === '字段')
|
||
const added = metadataChange.marks?.rows?.changed[fieldsIndex].items?.changed
|
||
expect(added?.[7]).toEqual({ added: true })
|
||
expect(Object.keys(added ?? {}).filter(index => added?.[Number(index)].added)).toEqual(['7'])
|
||
|
||
const traceFields = properties(trace).findIndex(row => row.label === '字段')
|
||
const cell = traceChange.marks?.rows?.changed[traceFields].items?.changed[2].cells?.[2]
|
||
expect(cell?.map(part => part.change)).toEqual(['del', 'add'])
|
||
expect(cell?.[0].segments.map(segment => segment.text).join('')).toBe('空')
|
||
expect(cell?.[1].segments.map(segment => segment.text).join('')).toBe('""')
|
||
expect(diff.stats).toEqual({ add: 0, mod: 2, del: 0 })
|
||
})
|
||
})
|