@feng0531/dsh-vcs 0.3.68 → 0.3.70

package/git-history.mjs

@@ -97,7 +97,7 @@ export function historyPage(rows,snapshot,offset,limit){
  return {snapshot,commits:rows.slice(0,limit),nextOffset:hasMore&&!truncated?offset+limit:null,truncated};
 }
 // --follow emits one selected-path name-status record per commit. Fail closed on unexpected framing.
-export function parseFollowHistory(text,path,maxRecords){
+export function parseFollowHistory(text,path,maxRecords,{similarity=false}={}){
  if(typeof text!=='string'||Buffer.byteLength(text,'utf8')>2*1024*1024||!Number.isInteger(maxRecords)||maxRecords<1||maxRecords>10101)throw Error('Invalid follow history bounds');
  if(!text)return [];if(!text.endsWith('\0'))throw Error('Truncated follow history');
  const fields=text.slice(0,-1).split('\0'),rows=[];let current=path;
@@ -109,10 +109,12 @@ export function parseFollowHistory(text,path,maxRecords){
   const oldPath=fields[i++],newPath=/^\n[RC]/.test(status)?fields[i++]:oldPath;
   if(!oldPath||!newPath||newPath!==current)throw Error('Ambiguous follow history path');
   if(/^\n[RC]/.test(status)&&Number(status.slice(2))>100)throw Error('Invalid follow similarity score');
-  const boundary=row.parents.length>1?'merge-first-parent':(/^\nC/.test(status)||(/^\nR/.test(status)&&status!=='\nR100'))?'copy-or-inexact-rename':null;
-  rows.push({...row,path:current,...(boundary?{boundary}:status==='\nR100'?{oldPath}:{})});
+  const rename=/^\nR/.test(status),score=rename?Number(status.slice(2)):null;
+  const accepted=rename&&(score===100||(similarity&&score>=50));
+  const boundary=row.parents.length>1?'merge-first-parent':(/^\nC/.test(status)||(rename&&!accepted))?'copy-or-inexact-rename':null;
+  rows.push({...row,path:current,...(boundary?{boundary}:accepted?{oldPath,...(similarity?{similarity:score}:{})}:{})});
   if(boundary)break;
-  if(status==='\nR100')current=oldPath;
+  if(accepted)current=oldPath;
   if(status==='\nA')break; // Stop at creation; never cross an older unrelated same-name file.
  }
  return rows;

package/index.mjs

@@ -34,11 +34,11 @@ const REQUEST_FIELDS = {
   'vcs/compare': ['sessionId', 'repositoryId', 'mode', 'id'],
   'vcs/history': ['sessionId', 'repositoryId', 'snapshot', 'offset', 'limit', 'search'],
   'vcs/commit': ['sessionId', 'repositoryId', 'commit', 'parentIndex'],
-  'vcs/commit-compare': ['sessionId', 'repositoryId', 'commit', 'parentIndex', 'id'],
+  'vcs/commit-compare': ['sessionId', 'repositoryId', 'commit', 'parentIndex', 'id', 'large'],
   'vcs/file-history': ['sessionId', 'repositoryId', 'commit', 'parentIndex', 'id', 'offset', 'limit', 'follow'],
   'vcs/blame': ['sessionId', 'repositoryId', 'commit', 'parentIndex', 'id', 'startLine', 'lineLimit'],
   'vcs/revision-changes': ['sessionId', 'repositoryId', 'base', 'target'],
-  'vcs/revision-compare': ['sessionId', 'repositoryId', 'base', 'target', 'id'],
+  'vcs/revision-compare': ['sessionId', 'repositoryId', 'base', 'target', 'id', 'large'],
   'vcs/revision-image': ['sessionId', 'repositoryId', 'base', 'target', 'id', 'side'],
   'vcs/tree': ['sessionId', 'repositoryId', 'commit'],
   'vcs/tree-file': ['sessionId', 'repositoryId', 'commit', 'path'],
@@ -68,6 +68,7 @@ function validatePayload(endpoint, payload) {
       if(typeof payload.commit!=='string'||!/^(?:[a-f0-9]{40}|[a-f0-9]{64})$/.test(payload.commit))throw invalid('Invalid commit id.');
       if(payload.parentIndex!==undefined&&(!Number.isInteger(payload.parentIndex)||payload.parentIndex<0||payload.parentIndex>100))throw invalid('Invalid parent index.');
     }
+    if(['vcs/commit-compare','vcs/revision-compare'].includes(endpoint)&&payload.large!==undefined&&typeof payload.large!=='boolean')throw invalid('Invalid large comparison mode.');
     if(endpoint==='vcs/tree-segment'&&payload.offset!==undefined&&(!Number.isInteger(payload.offset)||payload.offset<0||payload.offset>16777216))throw invalid('Invalid segment offset.');
     if(endpoint==='vcs/blame'&&((payload.startLine!==undefined&&(!Number.isInteger(payload.startLine)||payload.startLine<1||payload.startLine>100001))||(payload.lineLimit!==undefined&&(!Number.isInteger(payload.lineLimit)||payload.lineLimit<1||payload.lineLimit>500))))throw invalid('Invalid blame line window.');
     if(endpoint==='vcs/file-history'&&payload.follow!==undefined&&typeof payload.follow!=='boolean')throw invalid('Invalid rename follow mode.');
@@ -111,9 +112,9 @@ async function readRepository(api, endpoint, repository, payload, {cwd, mode, si
     case 'vcs/revision-changes':
       return api.getRevisionChanges({...repository}, {base: payload.base, target: payload.target, signal});
     case 'vcs/revision-compare':
-      return api.getRevisionComparison({...repository}, {base: payload.base, target: payload.target, id: payload.id, signal});
+      return api.getRevisionComparison({...repository}, {base: payload.base, target: payload.target, id: payload.id, ...(payload.large !== undefined ? {large: payload.large} : {}), signal});
     case 'vcs/commit-compare':
-      return api.getCommitComparison({...repository}, {commit: payload.commit, parentIndex: payload.parentIndex ?? 0, id: payload.id, signal});
+      return api.getCommitComparison({...repository}, {commit: payload.commit, parentIndex: payload.parentIndex ?? 0, id: payload.id, ...(payload.large !== undefined ? {large: payload.large} : {}), signal});
     case 'vcs/commit':
       return api.getCommitDetails({...repository}, {commit: payload.commit, parentIndex: payload.parentIndex ?? 0, signal});
     case 'vcs/history':
@@ -194,7 +195,7 @@ export function createHandler(ctx, api = adapter, maxActive = 4, { now = Date.no
       // Both adapter operations revalidate the canonical root; comparison also validates
       // the change ID against fresh status, so do not duplicate a full status scan here.
       if(['vcs/history','vcs/commit','vcs/commit-compare','vcs/revision-changes','vcs/revision-compare','vcs/file-history','vcs/blame','vcs/tree','vcs/tree-file','vcs/tree-segment','vcs/commit-image','vcs/revision-image','vcs/workspace-image','vcs/references'].includes(endpoint)&&repository.type!=='git')throw invalid('History currently supports Git only.');
-      if(endpoint==='vcs/tree-segment'){if(activeSegments>=1)return failure('vcs/busy','A text segment is already loading. Please retry.');activeSegments++;segmentSlot=true;}
+      if(endpoint==='vcs/tree-segment'||(['vcs/commit-compare','vcs/revision-compare'].includes(endpoint)&&payload.large===true)){if(activeSegments>=1)return failure('vcs/busy','A text segment is already loading. Please retry.');activeSegments++;segmentSlot=true;}
       if(['vcs/commit-image','vcs/revision-image','vcs/workspace-image'].includes(endpoint)){
         if(activeImages>=2)return failure('vcs/busy','Too many image requests. Please retry.');
         activeImages++;imageSlot=true;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@feng0531/dsh-vcs",
-  "version": "0.3.68",
+  "version": "0.3.70",
   "private": false,
   "publishConfig": { "access": "public", "registry": "https://registry.npmjs.org/" },
   "type": "module",

package/text-content.mjs

@@ -1,5 +1,29 @@
 /** Pure bounded byte decoding shared by working and committed previews. */
 export const MAX_TEXT = 2 * 1024 * 1024;
+export const MAX_LARGE_TEXT = 8 * 1024 * 1024;
+
+/** Explicit historical mode only: complete UTF-8, never a truncated Diff input. */
+export function decodeLargeText(buffer, signal) {
+  signal?.throwIfAborted();
+  if (buffer.length > MAX_LARGE_TEXT) throw new Error('Large comparison exceeds 8 MiB per side.');
+  if (buffer.some(byte => byte < 32 && ![9, 10, 12, 13].includes(byte))) {
+    throw new Error('Large comparison requires UTF-8 text without binary control bytes.');
+  }
+  let text;
+  try { text = new TextDecoder('utf-8', {fatal: true}).decode(buffer); }
+  catch { throw new Error('Large comparison requires valid UTF-8 text.'); }
+  let lines = 1, length = 0;
+  for (let index = 0; index < text.length; index++) {
+    if (index % 65536 === 0) signal?.throwIfAborted();
+    if (text[index] === '\n' || text[index] === '\r') {
+      if (text[index] === '\r' && text[index + 1] === '\n') index++;
+      if (++lines > 100000) throw new Error('Large comparison exceeds 100000 lines per side.');
+      length = 0;
+    } else if (++length > 65536) throw new Error('Large comparison exceeds 65536 characters per line.');
+  }
+  signal?.throwIfAborted();
+  return {text, encoding: 'UTF-8', totalBytes: buffer.length, lineCount: lines};
+}
 
 export function decode(buffer) {
   if (buffer.length > MAX_TEXT) {

package/vcs.mjs

@@ -8,7 +8,7 @@ import path from 'node:path';
 import { createHash } from 'node:crypto';
 import { XMLParser, XMLValidator } from 'fast-xml-parser';
 
-import {decode, MAX_TEXT} from './text-content.mjs';
+import {decode, MAX_TEXT, MAX_LARGE_TEXT, decodeLargeText} from './text-content.mjs';
 import {inside, confined} from './repository-path.mjs';
 const xml = new XMLParser({ ignoreAttributes: false, attributeNamePrefix: '', parseTagValue: false, parseAttributeValue: false, trimValues: false, processEntities: true, isArray: name => ['entry', 'target', 'property'].includes(name) });
 
@@ -386,10 +386,27 @@ export async function listFileHistory(repo,{commit,parentIndex=0,id,offset=0,lim
   const entry=details.changes.find(row=>row.id===id);
   if(!entry)throw new Error('Change is not part of selected commit');
   if(follow){
-   const count=offset+limit+1;
-   const output=await git(repo.root,['--literal-pathspecs','log','-z','--follow','--first-parent','--diff-merges=first-parent','--find-renames=100%','--name-status','--no-ext-diff','--no-textconv','--no-show-signature','--encoding=UTF-8','--max-count='+count,'--format='+HISTORY_FORMAT,commit,'--',entry.path],MAX_TEXT,{signal});
-   const rows=parseFollowHistory(output.toString('utf8'),entry.path,count).slice(offset);
-   return {...historyPage(rows,commit,offset,limit),path:entry.path,followsRenames:true,followPolicy:'exact-first-parent'};
+   const count=offset+limit+1,deadline=Date.now()+TIMEOUT;
+   const followOptions=()=>{signal?.throwIfAborted();const timeoutMs=deadline-Date.now();if(timeoutMs<=0)throw Object.assign(Error('Rename tracing timed out'),{code:'TIMEOUT'});return {signal,timeoutMs};};
+   const output=await git(repo.root,['--literal-pathspecs','log','-z','--follow','--first-parent','--diff-merges=first-parent','--find-renames=50%','-l100','--name-status','--no-ext-diff','--no-textconv','--no-show-signature','--encoding=UTF-8','--max-count='+count,'--format='+HISTORY_FORMAT,commit,'--',entry.path],MAX_TEXT,followOptions());
+   const chain=parseFollowHistory(output.toString('utf8'),entry.path,count,{similarity:true});
+   let checks=0;
+   for(let index=0;index<chain.length;index++){
+    const row=chain[index];if(!row.oldPath)continue;
+    let boundary;
+    if(++checks>32)boundary='rename-check-limit';
+    else {
+     // Confirm the whole transition, not Git's selected-path heuristic alone.
+     // A single deletion and addition excludes competing deleted sources.
+     const raw=await git(repo.root,['diff-tree','--no-commit-id','--name-status','-z','-r','--no-renames','--no-ext-diff','--no-textconv',row.parents[0],row.id,'--'],MAX_TEXT,followOptions());
+     const changes=parseCommitChanges(raw.toString('utf8'),row.id);
+     const removed=changes.filter(change=>change.status==='deleted'),added=changes.filter(change=>change.status==='added');
+     if(removed.length!==1||added.length!==1||removed[0].path!==row.oldPath||added[0].path!==row.path)boundary='ambiguous-rename';
+    }
+    if(boundary){chain[index]={...row,boundary};chain.length=index+1;break;}
+   }
+   const rows=chain.slice(offset);
+   return {...historyPage(rows,commit,offset,limit),path:entry.path,followsRenames:true,followPolicy:'unique-similarity-first-parent'};
   }
   const text=await git(repo.root,['--literal-pathspecs','log','-z','--no-show-signature','--encoding=UTF-8','--topo-order','--max-count='+String(limit+1),'--skip='+String(offset),'--format='+HISTORY_FORMAT,commit,'--',entry.path],MAX_TEXT,{signal});
   return {...historyPage(parseHistory(text.toString('utf8'),limit+1),commit,offset,limit),path:entry.path,followsRenames:false};
@@ -455,35 +472,48 @@ async function working(root, relative, svnLink = false) {
   try { const info = await handle.stat(); if (!info.isFile()) throw new Error('Not a regular file'); const buffer = Buffer.alloc(MAX_TEXT + 1); const { bytesRead } = await handle.read(buffer, 0, buffer.length, 0); await confined(root, relative); return decode(buffer.subarray(0, bytesRead)); } finally { await handle.close(); }
 }
 async function gitBlob(root, revision, file) { return content(() => git(root, ['show', '--no-ext-diff', '--no-textconv', revision + ':' + file], MAX_TEXT)); }
-async function historicalBlob(root,revision,file,signal){
+async function historicalBlob(root,revision,file,signal,large=false){
   const listing=(await git(root,['ls-tree','-z',revision,'--',file],MAX_TEXT,{signal})).toString('utf8');
   const records=listing.split(String.fromCharCode(0)).filter(Boolean);
   const record=records.find(row=>row.slice(row.indexOf('	')+1)===file);
   if(!record)throw new Error('Historical path missing');
   const [mode,type,oid]=record.slice(0,record.indexOf('	')).split(' ');
+  if(large&&(!['100644','100755'].includes(mode)||type!=='blob'))throw new Error('Large comparison requires a committed regular file.');
   if(mode==='160000')return {text:oid,notice:'Submodule commit reference; repository contents are not loaded.'};
   if(type!=='blob'||!['100644','100755','120000'].includes(mode))return {text:'',notice:'Unsupported historical object type.'};
+  if (large) {
+    const rawSize = (await git(root, ['cat-file', '-s', oid], MAX_TEXT, {signal})).toString('utf8').trim();
+    if (!/^\d+$/.test(rawSize)) throw new Error('Invalid historical object size.');
+    const size = Number(rawSize);
+    if (size > MAX_LARGE_TEXT) throw new Error('Large comparison exceeds 8 MiB per side.');
+    const buffer = await git(root, ['cat-file', 'blob', oid], MAX_LARGE_TEXT, {signal});
+    if (buffer.length !== size) throw new Error('Historical object size changed.');
+    return decodeLargeText(buffer, signal);
+  }
+
   const result=await content(()=>git(root,['cat-file','blob',oid],MAX_TEXT,{signal}));
   if(mode==='120000')result.notice=[result.notice,'Symbolic link target text; not followed.'].filter(Boolean).join(' ');
   return result;
 }
-export async function getCommitComparison(repo,{commit,parentIndex=0,id,signal}={}){
+export async function getCommitComparison(repo,{commit,parentIndex=0,id,signal,large=false}={}){
+  if(typeof large!=='boolean')throw new Error('Invalid large comparison mode');
   if(typeof id!=='string'||!/^[a-f0-9]{64}$/.test(id))throw new Error('Invalid historical change id');
   const details=await commitDetails(repo,{commit,parentIndex,signal});
   const entry=details.changes.find(row=>row.id===id);
   if(!entry)throw new Error('Historical change is not part of selected commit and parent');
-  const left=details.parent&&entry.status!=='added'?await historicalBlob(repo.root,details.parent,entry.oldPath||entry.path,signal):{text:''};
-  const right=entry.status!=='deleted'?await historicalBlob(repo.root,commit,entry.path,signal):{text:''};
-  return {...entry,left:{...left,label:details.parent||'Empty tree'},right:{...right,label:commit},binary:!!(left.binary||right.binary),notice:[left.notice,right.notice].filter(Boolean).join(' ')};
+  const left=details.parent&&entry.status!=='added'?await historicalBlob(repo.root,details.parent,entry.oldPath||entry.path,signal,large):{text:''};
+  const right=entry.status!=='deleted'?await historicalBlob(repo.root,commit,entry.path,signal,large):{text:''};
+  return {...entry,large,left:{...left,label:details.parent||'Empty tree'},right:{...right,label:commit},binary:!!(left.binary||right.binary),notice:[left.notice,right.notice].filter(Boolean).join(' ')};
 }
-export async function getRevisionComparison(repo,{base,target,id,signal}={}){
+export async function getRevisionComparison(repo,{base,target,id,signal,large=false}={}){
+  if(typeof large!=='boolean')throw new Error('Invalid large comparison mode');
   if(typeof id!=='string'||!/^[a-f0-9]{64}$/.test(id))throw new Error('Invalid revision change id');
   const details=await getRevisionChanges(repo,{base,target,signal});
   const entry=details.changes.find(row=>row.id===id);
   if(!entry)throw new Error('Change is not part of selected revision pair');
-  const left=entry.status!=='added'?await historicalBlob(repo.root,base,entry.oldPath||entry.path,signal):{text:''};
-  const right=entry.status!=='deleted'?await historicalBlob(repo.root,target,entry.path,signal):{text:''};
-  return {...entry,base,target,left:{...left,label:base},right:{...right,label:target},binary:!!(left.binary||right.binary),notice:[left.notice,right.notice].filter(Boolean).join(' ')};
+  const left=entry.status!=='added'?await historicalBlob(repo.root,base,entry.oldPath||entry.path,signal,large):{text:''};
+  const right=entry.status!=='deleted'?await historicalBlob(repo.root,target,entry.path,signal,large):{text:''};
+  return {...entry,base,target,large,left:{...left,label:base},right:{...right,label:target},binary:!!(left.binary||right.binary),notice:[left.notice,right.notice].filter(Boolean).join(' ')};
 }
 async function properties(root, file, base) {
   const result = parseXML(await svn(root, ['proplist', '--xml', '--verbose', ...(base ? ['--revision', 'BASE'] : []), '--', file + '@']));