dsh-thesis 2.0.2 → 2.1.1

package/lib/paper/lib/lit-refs.js

@@ -0,0 +1,394 @@
+/**
+ * 候选文献解析：把 Undermind / Elicit / Consensus / Zotero / Google Scholar
+ * 导出的 `.bib`、`.ris`、`.csv` 读成统一的 `CandidateRef[]`。
+ *
+ * 只做**解析**，不做判断：解析结果一律视为「声称值」，必须再过
+ * `lit-verify.ts` 的权威库对账才能进 refs.bib。解析器不联网、不改写元数据，
+ * 解析不了的字段留空并给出警告（警告会出现在核验报告里）。
+ */
+const BIB_EXT = /\.(bib|bibtex)$/i;
+const RIS_EXT = /\.ris$/i;
+const CSV_EXT = /\.(csv|tsv)$/i;
+/** 按扩展名与内容特征判断格式；都不像时返回 null。 */
+export function detectRefFormat(fileName, text) {
+    if (BIB_EXT.test(fileName))
+        return 'bibtex';
+    if (RIS_EXT.test(fileName))
+        return 'ris';
+    if (CSV_EXT.test(fileName))
+        return 'csv';
+    if (/^\s*@\w+\s*\{/m.test(text))
+        return 'bibtex';
+    if (/^\s*TY\s{2}-\s/m.test(text))
+        return 'ris';
+    if (/^\s*(title|题名|"title")\s*[,;\t]/im.test(text))
+        return 'csv';
+    return null;
+}
+export function parseRefs(fileName, text, source) {
+    const format = detectRefFormat(fileName, text);
+    if (format === null) {
+        throw new Error(`无法识别 ${fileName} 的格式。支持 Undermind / Elicit / Consensus / Zotero 导出的 .bib、.ris、.csv。` +
+            `请在这些工具里选择「Export → BibTeX / RIS / CSV」后重试；不支持直接粘贴的散乱文本（那种只能逐条用 thesis_lit_verify 的 title/doi 参数核验）。`);
+    }
+    if (format === 'bibtex')
+        return { format, ...parseBibtex(text, source) };
+    if (format === 'ris')
+        return { format, ...parseRis(text, source) };
+    return { format, ...parseCsv(text, source) };
+}
+// ---------------------------------------------------------------------------
+// 作者拆分
+// ---------------------------------------------------------------------------
+/** 作者串 → 姓名数组。兼容 `A; B`、`A and B`、JSON 数组、`A, B, C`（≥3 个时的兜底）。 */
+export function splitAuthors(raw) {
+    const text = raw.trim();
+    if (text === '')
+        return [];
+    if (text.startsWith('[') && text.endsWith(']')) {
+        try {
+            const parsed = JSON.parse(text);
+            if (Array.isArray(parsed)) {
+                return parsed.map(item => (typeof item === 'string' ? item : '')).map(s => s.trim()).filter(s => s !== '');
+            }
+        }
+        catch {
+            // 不是合法 JSON：继续按分隔符拆。
+        }
+    }
+    const clean = text.replace(/\s+and\s+/gi, ';').replace(/\s*&\s*/g, ';');
+    const semicolon = clean.split(';').map(s => s.trim()).filter(s => s !== '');
+    if (semicolon.length > 1)
+        return semicolon;
+    if (semicolon.length === 1) {
+        // `Zhang, San, Li, Si` 这种没有分号的全名列表：只在 ≥4 段且段数成对时按逗号拆。
+        const parts = clean.split(',').map(s => s.trim()).filter(s => s !== '');
+        if (parts.length >= 4 && parts.length % 2 === 0) {
+            const grouped = [];
+            for (let i = 0; i < parts.length; i += 2)
+                grouped.push(`${parts[i]} ${parts[i + 1]}`);
+            return grouped;
+        }
+        return semicolon;
+    }
+    return [];
+}
+function toYear(raw) {
+    if (raw === undefined)
+        return null;
+    const m = /\b(1[89]\d{2}|20\d{2})\b/.exec(raw);
+    return m === null ? null : Number.parseInt(m[1], 10);
+}
+/** 扫出所有 `@type{key, body}`；用花括号配对找条目结尾，能容忍嵌套 `{}` 与多行字段。 */
+export function bibEntries(text) {
+    const entries = [];
+    const start = /@(\w+)\s*\{/g;
+    let match;
+    while ((match = start.exec(text)) !== null) {
+        const type = (match[1] ?? '').toLowerCase();
+        if (type === 'comment' || type === 'preamble' || type === 'string') {
+            start.lastIndex = match.index + match[0].length;
+            continue;
+        }
+        const open = text.indexOf('{', match.index);
+        let depth = 0;
+        let end = open;
+        for (; end < text.length; end += 1) {
+            const ch = text[end];
+            if (ch === '{')
+                depth += 1;
+            else if (ch === '}') {
+                depth -= 1;
+                if (depth === 0)
+                    break;
+            }
+        }
+        const inner = text.slice(open + 1, end);
+        const comma = inner.indexOf(',');
+        const key = (comma === -1 ? inner : inner.slice(0, comma)).trim();
+        const body = comma === -1 ? '' : inner.slice(comma + 1);
+        entries.push({ type, key, fields: parseBibFields(body) });
+        start.lastIndex = end + 1;
+    }
+    return entries;
+}
+function parseBibFields(body) {
+    const fields = new Map();
+    let i = 0;
+    while (i < body.length) {
+        // 跳过空白与逗号
+        while (i < body.length && /[\s,]/.test(body[i]))
+            i += 1;
+        const eq = body.indexOf('=', i);
+        if (eq === -1)
+            break;
+        const name = body.slice(i, eq).trim().toLowerCase();
+        i = eq + 1;
+        while (i < body.length && /\s/.test(body[i]))
+            i += 1;
+        let value = '';
+        if (body[i] === '{') {
+            let depth = 0;
+            const from = i;
+            for (; i < body.length; i += 1) {
+                if (body[i] === '{')
+                    depth += 1;
+                else if (body[i] === '}') {
+                    depth -= 1;
+                    if (depth === 0) {
+                        i += 1;
+                        break;
+                    }
+                }
+            }
+            value = body.slice(from + 1, i - 1);
+        }
+        else if (body[i] === '"') {
+            const from = i + 1;
+            i = body.indexOf('"', from);
+            if (i === -1)
+                i = body.length;
+            value = body.slice(from, i);
+            i += 1;
+        }
+        else {
+            const from = i;
+            while (i < body.length && !/[,}\n]/.test(body[i]))
+                i += 1;
+            value = body.slice(from, i);
+        }
+        if (name !== '')
+            fields.set(name, cleanBibValue(value));
+    }
+    return fields;
+}
+function cleanBibValue(raw) {
+    return raw
+        .replace(/\\&/g, '&')
+        .replace(/\\%/g, '%')
+        .replace(/[{}]/g, '')
+        .replace(/\s+/g, ' ')
+        .trim();
+}
+function parseBibtex(text, source) {
+    const warnings = [];
+    const candidates = [];
+    const entries = bibEntries(text);
+    entries.forEach((entry, index) => {
+        const title = entry.fields.get('title') ?? '';
+        if (title === '') {
+            warnings.push(`第 ${index + 1} 条（${entry.key || '无 key'}）没有 title，已跳过。`);
+            return;
+        }
+        const doi = entry.fields.get('doi');
+        const candidate = {
+            title,
+            authors: splitAuthors(entry.fields.get('author') ?? entry.fields.get('editor') ?? ''),
+            year: toYear(entry.fields.get('year')),
+            venue: entry.fields.get('journal') ?? entry.fields.get('booktitle') ?? entry.fields.get('publisher') ?? entry.fields.get('school') ?? '',
+            ...(doi !== undefined ? { doi } : {}),
+            ...(entry.fields.get('url') !== undefined ? { url: entry.fields.get('url') } : {}),
+            source,
+            ordinal: index + 1,
+        };
+        candidates.push(candidate);
+    });
+    return { candidates, warnings };
+}
+// ---------------------------------------------------------------------------
+// RIS
+// ---------------------------------------------------------------------------
+const RIS_FIELD = {
+    TI: 'title',
+    T1: 'title',
+    BT: 'title',
+    AU: 'author',
+    A1: 'author',
+    PY: 'year',
+    Y1: 'year',
+    JO: 'venue',
+    JF: 'venue',
+    T2: 'venue',
+    JA: 'venue',
+    DO: 'doi',
+    UR: 'url',
+    L2: 'url',
+};
+function parseRis(text, source) {
+    const warnings = [];
+    const candidates = [];
+    const records = [];
+    let current = new Map();
+    let open = false;
+    for (const line of text.split(/\r?\n/)) {
+        const m = /^([A-Z][A-Z0-9])\s{2}-\s?(.*)$/.exec(line);
+        if (m === null)
+            continue;
+        const tag = m[1];
+        const value = m[2].trim();
+        if (tag === 'TY') {
+            current = new Map();
+            open = true;
+            continue;
+        }
+        if (tag === 'ER') {
+            if (open && current.size > 0)
+                records.push(current);
+            open = false;
+            continue;
+        }
+        if (!open)
+            continue;
+        const field = RIS_FIELD[tag];
+        if (field === undefined)
+            continue;
+        current.set(field, [...(current.get(field) ?? []), value]);
+    }
+    records.forEach((record, index) => {
+        const title = record.get('title')?.[0]?.replace(/<[^>]+>/g, '').trim() ?? '';
+        if (title === '') {
+            warnings.push(`第 ${index + 1} 条 RIS 没有 TI（题名），已跳过。`);
+            return;
+        }
+        const doi = record.get('doi')?.[0];
+        const url = record.get('url')?.[0];
+        candidates.push({
+            title,
+            authors: (record.get('author') ?? []).flatMap(splitAuthors),
+            year: toYear(record.get('year')?.[0]),
+            venue: record.get('venue')?.[0] ?? '',
+            ...(doi !== undefined ? { doi } : {}),
+            ...(url !== undefined ? { url } : {}),
+            source,
+            ordinal: index + 1,
+        });
+    });
+    return { candidates, warnings };
+}
+// ---------------------------------------------------------------------------
+// CSV（Elicit / Consensus / Scopus 导出的表格）
+// ---------------------------------------------------------------------------
+/** RFC4180 风格解析：支持双引号包裹、`""` 转义、字段内换行。 */
+export function parseCsvRows(text, delimiter = ',') {
+    const rows = [];
+    let row = [];
+    let field = '';
+    let quoted = false;
+    for (let i = 0; i < text.length; i += 1) {
+        const ch = text[i];
+        if (quoted) {
+            if (ch === '"') {
+                if (text[i + 1] === '"') {
+                    field += '"';
+                    i += 1;
+                }
+                else {
+                    quoted = false;
+                }
+            }
+            else {
+                field += ch;
+            }
+            continue;
+        }
+        if (ch === '"') {
+            quoted = true;
+        }
+        else if (ch === delimiter) {
+            row.push(field);
+            field = '';
+        }
+        else if (ch === '\n' || ch === '\r') {
+            if (ch === '\r' && text[i + 1] === '\n')
+                i += 1;
+            row.push(field);
+            field = '';
+            if (row.some(cell => cell.trim() !== ''))
+                rows.push(row);
+            row = [];
+        }
+        else {
+            field += ch;
+        }
+    }
+    row.push(field);
+    if (row.some(cell => cell.trim() !== ''))
+        rows.push(row);
+    return rows;
+}
+const HEADER_ALIASES = {
+    title: ['title', '题名', '标题', 'paper title', 'name', 'document title'],
+    authors: ['authors', 'author', '作者', 'creators', 'author names'],
+    year: ['year', 'publication year', '出版年', '年份', 'published', 'date'],
+    venue: ['venue', 'journal', 'source title', 'source', 'publication', '期刊', '会议', 'published in', 'container title'],
+    doi: ['doi', 'doi link', 'doi url', 'digital object identifier'],
+    url: ['url', 'link', 'uri', 'pdf', 'landing page url'],
+};
+function normalizeHeader(raw) {
+    return raw.toLowerCase().replace(/[*_"]/g, '').replace(/\s+/g, ' ').trim();
+}
+function mapHeaders(header) {
+    const mapped = new Map();
+    header.forEach((cell, index) => {
+        const norm = normalizeHeader(cell);
+        for (const field of Object.keys(HEADER_ALIASES)) {
+            if (mapped.has(field))
+                continue;
+            if (HEADER_ALIASES[field].some(alias => norm === alias || norm.startsWith(`${alias} (`) || norm.startsWith(`${alias} -`))) {
+                mapped.set(field, index);
+            }
+        }
+    });
+    return mapped;
+}
+function detectDelimiter(text) {
+    const firstLine = text.split(/\r?\n/, 1)[0] ?? '';
+    const counts = [[',', 0], [';', 0], ['\t', 0]];
+    for (const ch of firstLine) {
+        const slot = counts.find(([d]) => d === ch);
+        if (slot !== undefined)
+            slot[1] += 1;
+    }
+    counts.sort((a, b) => b[1] - a[1]);
+    return counts[0][1] > 0 ? counts[0][0] : ',';
+}
+function parseCsv(text, source) {
+    const warnings = [];
+    const rows = parseCsvRows(text.replace(/^\uFEFF/, ''), detectDelimiter(text));
+    if (rows.length === 0)
+        return { candidates: [], warnings: ['文件为空。'] };
+    const header = rows[0];
+    const mapped = mapHeaders(header);
+    if (!mapped.has('title')) {
+        throw new Error(`CSV 表头里没有题名列（现有：${header.slice(0, 8).join(' / ')}）。` +
+            `请确认导出的是文献列表（列名含 Title/题名），而不是筛选结果或统计表。`);
+    }
+    const get = (row, field) => {
+        const index = mapped.get(field);
+        if (index === undefined)
+            return undefined;
+        const value = row[index]?.trim();
+        return value === undefined || value === '' ? undefined : value;
+    };
+    const candidates = [];
+    rows.slice(1).forEach((row, index) => {
+        const title = get(row, 'title');
+        if (title === undefined) {
+            warnings.push(`第 ${index + 2} 行没有题名，已跳过。`);
+            return;
+        }
+        const doi = get(row, 'doi');
+        const url = get(row, 'url');
+        candidates.push({
+            title,
+            authors: splitAuthors(get(row, 'authors') ?? ''),
+            year: toYear(get(row, 'year')),
+            venue: get(row, 'venue') ?? '',
+            ...(doi !== undefined ? { doi } : {}),
+            ...(url !== undefined ? { url } : {}),
+            source,
+            ordinal: index + 2,
+        });
+    });
+    return { candidates, warnings };
+}

package/lib/paper/lib/lit-verify.d.ts

@@ -0,0 +1,79 @@
+/**
+ * 文献核验：把「候选文献」（AI 研究助手导出、粘贴、学长给的 pdf 清单）逐条
+ * 拿到权威库对账，只有对得上的才允许进 refs.bib。
+ *
+ * 为什么需要这一层：Undermind / Elicit / Consensus 这类助手是**发现**工具，
+ * 它们的导出里既有真论文，也有识别错误的题名、错位的年份、编造的 DOI。
+ * 「零假文献」不能靠"我信任这个来源"，只能靠**逐条对账 + 失败即拒**。
+ *
+ * 权威库（全部免费、无需 key）：
+ * - Crossref  https://api.crossref.org/works/{doi} 与 ?query.bibliographic=
+ * - OpenAlex  https://api.openalex.org/works/doi:{doi} 与 ?search=
+ * - DataCite  https://api.datacite.org/dois/{doi}（数据集/学位论文/预印本）
+ * - PubMed    https://eutils.ncbi.nlm.nih.gov/entrez/eutils/（生物医学）
+ *
+ * 判定口径（写死，不给"感觉差不多"留口子）：
+ * - verified：权威库存在该文献，且题名相似度 ≥0.92、第一作者姓氏对得上、年份差 ≤1；
+ * - mismatch：权威库有记录但题名/作者/年份冲突（多半是张冠李戴，须人工看原文）；
+ * - not-found：权威库查无此文（DOI 不存在 / 题名检索无命中）——**疑似编造**；
+ * - unchecked：网络不可用或库限流，本轮没核验（fail closed：不许入库）。
+ */
+import { type FetchLike, type LitRecord } from './lit-api.ts';
+export type VerifyStatus = 'verified' | 'mismatch' | 'not-found' | 'invalid' | 'unchecked';
+/** 待核验的候选文献。字段是"声称值"，不保证为真。 */
+export interface CandidateRef {
+    readonly title: string;
+    readonly authors: readonly string[];
+    readonly year: number | null;
+    readonly venue: string;
+    readonly doi?: string;
+    readonly url?: string;
+    /** 来源标签：undermind / elicit / consensus / zotero / manual / paste… */
+    readonly source: string;
+    /** 在原文件中的位置（第几条/第几行），用于回指。 */
+    readonly ordinal?: number;
+}
+export interface ReferenceCheck {
+    readonly field: '题名' | '作者' | '年份' | 'DOI' | '出处';
+    readonly claimed: string;
+    readonly registered: string;
+    /** ok=一致；warn=有差异但可接受；fail=冲突（禁止入库） */
+    readonly severity: 'ok' | 'warn' | 'fail';
+}
+export interface VerifyScreen {
+    readonly registry: string;
+    readonly query: string;
+    readonly ok: boolean;
+    readonly detail: string;
+}
+export interface VerifyOutcome {
+    readonly candidate: CandidateRef;
+    readonly status: VerifyStatus;
+    /** 题名相似度 0..1（无对照记录时为 0）。 */
+    readonly score: number;
+    readonly record?: LitRecord;
+    readonly checks: readonly ReferenceCheck[];
+    readonly screens: readonly VerifyScreen[];
+    readonly message: string;
+}
+export declare const VERIFY_REPORT_REL = "08-\u5408\u89C4/\u6587\u732E\u6838\u9A8C\u62A5\u544A.md";
+/** 去掉 `https://doi.org/`、`doi:` 前缀与空白；统一小写。不是合法 DOI 形状时返回 undefined。 */
+export declare function normalizeDoi(raw: string | undefined): string | undefined;
+export declare function doiShapeValid(doi: string): boolean;
+/** 题名归一化：去重音、去标点、压缩空白、小写。中文按字切分。 */
+export declare function normalizeTitle(raw: string): string;
+/** 题名相似度：中文按双字 shingle，西文按词，取 Dice 系数。 */
+export declare function titleSimilarity(a: string, b: string): number;
+/** 姓氏：`Zhang, Wei` 取逗号前；西文取最后一个词；中文姓名取第一个字；去掉 Jr./van 等缀词。 */
+export declare function familyName(author: string): string;
+/**
+ * 核验单条候选文献。网络全部失败时返回 `unchecked`（**不是** verified）：
+ * 拿不到权威库就宁可不入库，这是「零假文献」的 fail-closed 语义。
+ */
+export declare function verifyCandidate(fetchImpl: FetchLike, candidate: CandidateRef, signal?: AbortSignal): Promise<VerifyOutcome>;
+/** 批量核验（顺序执行，礼貌对待免费库；单条失败不影响其它条）。 */
+export declare function verifyAll(fetchImpl: FetchLike, candidates: readonly CandidateRef[], signal?: AbortSignal): Promise<VerifyOutcome[]>;
+export declare function summarizeOutcomes(outcomes: readonly VerifyOutcome[]): string;
+export declare function renderOutcome(outcome: VerifyOutcome): string;
+export declare function renderVerifyReport(date: string, outcomes: readonly VerifyOutcome[], label: string): string;
+export declare const VERIFY_REPORT_HEADER = "# \u6587\u732E\u6838\u9A8C\u62A5\u544A\n\n> \u53E3\u5F84\uFF1A\u6BCF\u6761\u5019\u9009\u6587\u732E\u5FC5\u987B\u7531 Crossref / OpenAlex / DataCite / PubMed \u4E2D\u7684\u6743\u5A01\u5E93\u786E\u8BA4\u5B58\u5728\uFF0C\n> \u4E14\u9898\u540D\u76F8\u4F3C\u5EA6 \u22650.92\u3001\u7B2C\u4E00\u4F5C\u8005\u59D3\u6C0F\u4E00\u81F4\u3001\u5E74\u4EFD\u5DEE \u22641\uFF0C\u624D\u5141\u8BB8\u5199\u5165 02-\u6587\u732E/refs.bib\u3002\n> Undermind / Elicit / Consensus \u53EA\u8D1F\u8D23\u300C\u627E\u5230\u5019\u9009\u300D\uFF0C\u5165\u5E93\u4E00\u5F8B\u4EE5\u672C\u62A5\u544A\u7684\u6838\u9A8C\u7ED3\u8BBA\u4E3A\u51C6\u3002\n";