205 lines
8.8 KiB
TypeScript
205 lines
8.8 KiB
TypeScript
import { describe, expect, it } from 'vitest'
|
||
import { loadDocxBundle } from '../../../../src/main/services/ai/human/tools/docx/adapters/ooxml/nodeIndex'
|
||
import { buildDocxNetReview } from '../../../../src/main/services/ai/human/tools/docx/review/netReview'
|
||
import { graphemes } from '../../../../src/main/services/ai/human/tools/docx/review/inlineDiff'
|
||
import {
|
||
paragraphXml as p,
|
||
reviewBuffer,
|
||
reviewRuns,
|
||
storyText
|
||
} from '../../../harness/backend/docxReviewAssertions'
|
||
|
||
const compare = async (
|
||
before: string,
|
||
after: string,
|
||
oldExtras: Record<string, string | Buffer> = {},
|
||
newExtras = oldExtras
|
||
) => {
|
||
const source = await loadDocxBundle(await reviewBuffer(before, oldExtras))
|
||
const working = await loadDocxBundle(await reviewBuffer(after, newExtras))
|
||
const result = await buildDocxNetReview(source, working)
|
||
return { ...result, document: working.documentDom, zip: working.zip }
|
||
}
|
||
describe('DOCX first/final net review', () => {
|
||
it.each([
|
||
[
|
||
['title', 'date', 'conclusion', 'section', 'risk', 'tail'],
|
||
['title', 'date', 'risk', 'conclusion', 'section', 'tail']
|
||
],
|
||
[
|
||
['title', 'date', 'conclusion', 'section', 'risk', 'tail'],
|
||
['title', 'date', 'section', 'risk', 'conclusion', 'tail']
|
||
],
|
||
[
|
||
['A', 'B', 'C'],
|
||
['B', 'A', 'C']
|
||
],
|
||
[
|
||
['A', 'B', 'C', 'D'],
|
||
['C', 'D', 'A', 'B']
|
||
],
|
||
[
|
||
['A', 'same', 'B', 'same', 'C'],
|
||
['A', 'same', 'same', 'B', 'C']
|
||
],
|
||
[
|
||
['风险旧稿', '结论', '尾段'],
|
||
['结论', '尾段', '风险新稿']
|
||
],
|
||
[
|
||
['A', 'B'],
|
||
['B', 'new', 'A']
|
||
],
|
||
[
|
||
['A', 'A', 'B'],
|
||
['A', 'B', 'B']
|
||
]
|
||
])('restores both orders for %j → %j', async (before, after) => {
|
||
const result = await compare(before.map((t) => p(t)).join(''), after.map((t) => p(t)).join(''))
|
||
expect(storyText(result.document, false)).toEqual(before)
|
||
expect(storyText(result.document, true)).toEqual(after)
|
||
})
|
||
it.each([
|
||
['尾款126万元', '尾款128万元'],
|
||
['甲乙丙丁戊', '甲新丙好戊'],
|
||
['a b\tZ', 'a b\tQ'],
|
||
['👩🏽💻é🙂', '👩🏽💻é🙃'],
|
||
['abc', ''],
|
||
['', 'new'],
|
||
['123abc！？', '123ABC？'],
|
||
['重复重复', '重复变化重复']
|
||
])('character diff preserves %j and %j', async (before, after) => {
|
||
const result = await compare(p(before), p(after))
|
||
expect(storyText(result.document, false)).toEqual([before])
|
||
expect(storyText(result.document, true)).toEqual([after])
|
||
expect(result.stats).toEqual({ changeCount: 1, insertCount: 1, deleteCount: 1 })
|
||
expect(
|
||
reviewRuns(result.document)
|
||
.filter((r) => r.kind === 'deleted')
|
||
.every((r) => r.xml.includes('strike'))
|
||
).toBe(true)
|
||
})
|
||
it('splits by grapheme, including a cluster crossing run boundaries', async () => {
|
||
expect(graphemes('👩🏽💻é🇨🇳')).toEqual(['👩🏽💻', 'é', '🇨🇳'])
|
||
const result = await compare(
|
||
'<w:p><w:r><w:t>e</w:t></w:r><w:r><w:t>́旧</w:t></w:r></w:p>',
|
||
p('é新')
|
||
)
|
||
expect(
|
||
reviewRuns(result.document)
|
||
.filter((r) => r.kind === 'deleted')
|
||
.map((r) => r.text)
|
||
.join('')
|
||
).toBe('旧')
|
||
expect(storyText(result.document, true)).toEqual(['é新'])
|
||
})
|
||
it('ignores arbitrary run splits and node identities', async () => {
|
||
const result = await compare(
|
||
p('一致'),
|
||
'<w:p w:rsidR="ABCD"><w:r><w:t>一</w:t></w:r><w:r><w:t>致</w:t></w:r></w:p>'
|
||
)
|
||
expect(result.stats.changeCount).toBe(0)
|
||
expect(reviewRuns(result.document).every((r) => !r.kind)).toBe(true)
|
||
})
|
||
it('marks only characters whose format changes, preserving their foreground/background', async () => {
|
||
const result = await compare(
|
||
'<w:p><w:r><w:t>前</w:t></w:r><w:r><w:rPr><w:color w:val="223344"/><w:shd w:fill="FFFF00"/></w:rPr><w:t>中</w:t></w:r><w:r><w:t>后</w:t></w:r></w:p>',
|
||
'<w:p><w:r><w:t>前</w:t></w:r><w:r><w:rPr><w:color w:val="223344"/><w:shd w:fill="00FFFF"/><w:b/></w:rPr><w:t>中</w:t></w:r><w:r><w:t>后</w:t></w:r></w:p>'
|
||
)
|
||
const runs = reviewRuns(result.document)
|
||
expect(runs.filter((r) => r.kind).map((r) => r.text)).toEqual(['中'])
|
||
expect(runs[1]!.xml).toContain('223344')
|
||
expect(runs[1]!.xml).toContain('00FFFF')
|
||
expect(runs[1]!.xml).toContain('w:bdr')
|
||
expect(result.stats).toEqual({ changeCount: 1, insertCount: 0, deleteCount: 0 })
|
||
expect(storyText(result.document)).toEqual(['前中后'])
|
||
})
|
||
it('aligns repeated text with the retained formatting when simplification removes an earlier occurrence', async () => {
|
||
const before =
|
||
'<w:p><w:r><w:t>本产品是一套面向社区公开资料整理与分享场景的 </w:t></w:r><w:r><w:rPr><w:b/></w:rPr><w:t>公开资料整理工作台</w:t></w:r><w:r><w:t>，支持资料检索。</w:t></w:r></w:p>'
|
||
const after =
|
||
'<w:p><w:r><w:t>本产品是一套</w:t></w:r><w:r><w:rPr><w:b/></w:rPr><w:t>公开资料整理工作台</w:t></w:r><w:r><w:t>，支持资料检索。</w:t></w:r></w:p>'
|
||
const result = await compare(before, after)
|
||
expect(storyText(result.document, false)).toEqual([
|
||
'本产品是一套面向社区公开资料整理与分享场景的 公开资料整理工作台，支持资料检索。'
|
||
])
|
||
expect(storyText(result.document, true)).toEqual([
|
||
'本产品是一套公开资料整理工作台，支持资料检索。'
|
||
])
|
||
expect(reviewRuns(result.document).filter((run) => run.kind === 'formatted')).toEqual([])
|
||
expect(
|
||
reviewRuns(result.document)
|
||
.filter((run) => run.kind === 'deleted')
|
||
.map((run) => run.text)
|
||
.join('')
|
||
).toBe('面向社区公开资料整理与分享场景的 ')
|
||
})
|
||
it('keeps real format changes when matching repeated formatting would remove unchanged text', async () => {
|
||
const result = await compare(
|
||
'<w:p><w:r><w:rPr><w:b/></w:rPr><w:t>同</w:t></w:r><w:r><w:t>同</w:t></w:r></w:p>',
|
||
'<w:p><w:r><w:t>同</w:t></w:r><w:r><w:rPr><w:b/></w:rPr><w:t>同</w:t></w:r></w:p>'
|
||
)
|
||
const runs = reviewRuns(result.document)
|
||
expect(runs.every((run) => run.kind === 'formatted')).toBe(true)
|
||
expect(runs.map((run) => run.text).join('')).toBe('同同')
|
||
expect(result.stats.insertCount).toBe(0)
|
||
expect(result.stats.deleteCount).toBe(0)
|
||
expect(storyText(result.document, false)).toEqual(['同同'])
|
||
expect(storyText(result.document, true)).toEqual(['同同'])
|
||
})
|
||
it('uses border when background has been removed', async () => {
|
||
const result = await compare(p('背景', '<w:shd w:fill="FFFF00"/>'), p('背景'))
|
||
expect(
|
||
reviewRuns(result.document).every((r) => r.xml.includes('w:bdr') && !r.xml.includes('E9F7EF'))
|
||
).toBe(true)
|
||
})
|
||
it('does not split complex fields into visible characters', async () => {
|
||
const field = (result: string) =>
|
||
`<w:p><w:r><w:fldChar w:fldCharType="begin"/></w:r><w:r><w:instrText>DATE</w:instrText></w:r><w:r><w:fldChar w:fldCharType="separate"/></w:r><w:r><w:t>${result}</w:t></w:r><w:r><w:fldChar w:fldCharType="end"/></w:r></w:p>`
|
||
const result = await compare(field('2025'), field('2026'))
|
||
expect(result.document.getElementsByTagName('w:instrText').length).toBe(2)
|
||
expect(storyText(result.document, false)).toEqual(['2025'])
|
||
expect(storyText(result.document, true)).toEqual(['2026'])
|
||
})
|
||
it('bounds output/time for a long, fragmented repeated paragraph', async () => {
|
||
const text = '重复测试'.repeat(3000)
|
||
const fragmented = `<w:p>${Array.from(text)
|
||
.map((char) => `<w:r><w:t>${char}</w:t></w:r>`)
|
||
.join('')}</w:p>`
|
||
const start = Date.now(),
|
||
result = await compare(fragmented, p(text.slice(0, -1) + '新'))
|
||
expect(Date.now() - start).toBeLessThan(15000)
|
||
expect(storyText(result.document, true)).toEqual([text.slice(0, -1) + '新'])
|
||
expect(reviewRuns(result.document).length).toBeLessThan(10)
|
||
}, 20000)
|
||
it('keeps unchanged characters when a large edit exceeds the fast alignment limit', async () => {
|
||
const before = '前缀' + '甲乙'.repeat(2000) + '后缀'
|
||
const after = '前缀' + '丙乙'.repeat(2000) + '后缀'
|
||
const start = Date.now(),
|
||
result = await compare(p(before), p(after))
|
||
const runs = reviewRuns(result.document)
|
||
expect(
|
||
runs
|
||
.filter((r) => r.kind === 'deleted')
|
||
.map((r) => r.text)
|
||
.join('')
|
||
).toBe('甲'.repeat(2000))
|
||
expect(
|
||
runs
|
||
.filter((r) => r.kind === 'inserted')
|
||
.map((r) => r.text)
|
||
.join('')
|
||
).toBe('丙'.repeat(2000))
|
||
expect(
|
||
runs
|
||
.filter((r) => !r.kind)
|
||
.map((r) => r.text)
|
||
.join('')
|
||
).toBe('前缀' + '乙'.repeat(2000) + '后缀')
|
||
expect(storyText(result.document, false)).toEqual([before])
|
||
expect(storyText(result.document, true)).toEqual([after])
|
||
expect(Date.now() - start).toBeLessThan(15000)
|
||
expect(runs.length).toBeLessThanOrEqual(before.length * 2)
|
||
}, 20000)
|
||
})
|