import { z } from 'zod'; const kind=z.enum(['guide','context','evidence']); export const validXmlText=(value:string)=>!/[\u0000-\u0008\u000b\u000c\u000e-\u001f\ufffe\uffff]/.test(value)&&!/[\ud800-\udfff]/u.test(value); const text=z.string().refine(validXmlText,'非法 XML 字符').trim().min(1).max(20000); export const documentInput=z.object({id:z.string().regex(/^user-[a-z0-9-]{1,74}$/),revision:z.number().int().min(1).max(2147483647),kind,title:text.max(200),summary:text.max(1000),applies:text,body:text,steps:z.array(text).max(100),checks:z.array(text).max(100),links:z.array(z.object({target:z.string().regex(/^[a-z0-9-]{1,79}$/),revision:z.number().int().min(1).max(2147483647),relation:kind,when:text.max(1000),label:text.max(200)}).strict()).max(100),tags:z.array(text.max(80)).max(20)}).strict(); // Deliberately narrow XML 1.0 grammar: no DTD, comments, CDATA, nested markup or arbitrary entities. export function parseDocumentXml(source:string){ if(!validXmlText(source)||source.includes(']]>'))throw Error('非法 XML 字符'); if(Buffer.byteLength(source)>131072)throw Error('XML 过大'); const trimXml=(v:string)=>v.replace(/^[ \t\r\n]+|[ \t\r\n]+$/g,''); let s=source.replace(/^<\?xml version="1.0" encoding="UTF-8"\?>[ \t\r\n]*/, ''); if(/<\?|{if(/&(?!amp;|lt;|gt;|quot;|apos;)/.test(v))throw Error('非法实体');return v.replace(/&(amp|lt|gt|quot|apos);/g,(_,k)=>({amp:'&',lt:'<',gt:'>',quot:'"',apos:"'"}[k as 'amp'|'lt'|'gt'|'quot'|'apos'] as string));}; const attrs=(v:string)=>{const result:Record={};while(trimXml(v)){const m=/^[ \t\r\n]+([a-z]+)="([^"<]*)"/.exec(v);if(!m||m[1] in result)throw Error('非法或重复属性');result[m[1]]=decode(m[2]);v=v.slice(m[0].length);}return result;}; const root=/^([\s\S]*)<\/guidance>[ \t\r\n]*$/.exec(s);if(!root)throw Error('非法根元素');const a=attrs(root[1]); if(Object.keys(a).sort().join(',')!=='id,kind,lang,revision,xmlns'||a.xmlns!=='urn:linked-guidance:0.1'||a.lang!=='zh-CN'||! /^[1-9][0-9]*$/.test(a.revision))throw Error('不支持的协议或属性'); const result:Record={id:a.id,kind:a.kind,revision:Number(a.revision),steps:[],checks:[],links:[],tags:[]};s=root[2]; while(trimXml(s)){const m=/^[ \t\r\n]*<(title|summary|applies|body|step|check|tag|link)((?:[ \t\r\n]+[a-z]+="[^"<]*")*[ \t\r\n]*)>([^<]*)<\/\1>/.exec(s);if(!m)throw Error('非法子元素');const [,tag,raw,value]=m;const at=attrs(raw);const v=decode(value); if(tag==='link'){if(Object.keys(at).sort().join(',')!=='relation,revision,target,when'||! /^[1-9][0-9]*$/.test(at.revision))throw Error('非法链接');(result.links as unknown[]).push({...at,revision:Number(at.revision),label:v});} else {if(Object.keys(at).length)throw Error('非法属性');if(tag==='step'||tag==='check'||tag==='tag')(result[tag==='step'?'steps':tag==='tag'?'tags':'checks'] as string[]).push(v);else{if(tag in result)throw Error('重复元素');result[tag]=v;}}s=s.slice(m[0].length);} return documentInput.parse(result); }