openchamber/packages/ui/src/lib/diff/patchFileDiff.ts

import {
  parseDiffFromFile,
  parsePatchFiles,
  processFile,
  trimPatchContext,
  type FileDiffMetadata,
} from '@pierre/diffs';

const PATCH_DIFF_CACHE_LIMIT = 64;
const DEFAULT_PATCH_CONTEXT_LINES = 3;
const patchFileDiffCache = new Map<string, FileDiffMetadata>();

export const fileDiffFromPatch = (
  file: string,
  patch: string,
  contextLines = DEFAULT_PATCH_CONTEXT_LINES
): FileDiffMetadata => {
  const key = `${file}\0${contextLines}\0${patch}`;
  const cached = patchFileDiffCache.get(key);
  if (cached) {
    patchFileDiffCache.delete(key);
    patchFileDiffCache.set(key, cached);
    return cached;
  }

  const completeContents = completePatchContents(patch);
  const value = completeContents
    ? (processFile(trimPatchContext(patch, contextLines), {
      oldFile: { name: file, contents: completeContents.before },
      newFile: { name: file, contents: completeContents.after },
    }) ?? emptyFileDiff(file))
    : (parsePatchFiles(withPatchHeader(file, patch))[0]?.files[0] ?? emptyFileDiff(file));
  patchFileDiffCache.set(key, value);

  while (patchFileDiffCache.size > PATCH_DIFF_CACHE_LIMIT) {
    const firstKey = patchFileDiffCache.keys().next().value;
    if (!firstKey) break;
    patchFileDiffCache.delete(firstKey);
  }

  return value;
};

const withPatchHeader = (file: string, patch: string): string => {
  if (!patch.trim()) {
    return patch;
  }

  if (patch.startsWith('diff --git ') || /^--- [^\n]*\r?\n\+\+\+ /m.test(patch)) {
    return patch;
  }

  return `Index: ${file}\n===================================================================\n--- ${file}\t\n+++ ${file}\t\n${patch}`;
};

const completePatchContents = (patch: string): { before: string; after: string } | undefined => {
  if (!patch.startsWith('diff --git ') && !/^--- [^\n]*\t?\r?\n\+\+\+ [^\n]*\t?(?:\r?\n|$)/m.test(patch)) {
    return undefined;
  }

  const hunkMatches = [...patch.matchAll(/^@@\s+-(\d+)(?:,\d+)?\s+\+(\d+)(?:,\d+)?\s+@@[^\n]*(?:\r?\n|$)/gm)];
  if (hunkMatches.length !== 1) {
    return undefined;
  }

  const hunk = hunkMatches[0];
  const oldStart = Number.parseInt(hunk[1] ?? '', 10);
  const newStart = Number.parseInt(hunk[2] ?? '', 10);
  if (oldStart > 1 || newStart > 1 || !Number.isFinite(oldStart) || !Number.isFinite(newStart)) {
    return undefined;
  }

  const hunkStart = (hunk.index ?? 0) + hunk[0].length;
  const body = patch.slice(hunkStart);
  const before: Array<{ text: string; newline: boolean }> = [];
  const after: Array<{ text: string; newline: boolean }> = [];
  let previous: '-' | '+' | ' ' | undefined;

  for (const rawLine of body.split(/\r?\n/)) {
    if (rawLine.startsWith('diff --git ') || rawLine.startsWith('@@ ')) {
      break;
    }

    if (rawLine.startsWith('\\')) {
      if (previous === '-' || previous === ' ') {
        const value = before.at(-1);
        if (value) value.newline = false;
      }
      if (previous === '+' || previous === ' ') {
        const value = after.at(-1);
        if (value) value.newline = false;
      }
      continue;
    }

    if (rawLine.startsWith('-')) {
      before.push({ text: rawLine.slice(1), newline: true });
      previous = '-';
      continue;
    }

    if (rawLine.startsWith('+')) {
      after.push({ text: rawLine.slice(1), newline: true });
      previous = '+';
      continue;
    }

    if (!rawLine.startsWith(' ')) {
      continue;
    }

    before.push({ text: rawLine.slice(1), newline: true });
    after.push({ text: rawLine.slice(1), newline: true });
    previous = ' ';
  }

  return {
    before: joinPatchLines(before),
    after: joinPatchLines(after),
  };
};

const joinPatchLines = (lines: Array<{ text: string; newline: boolean }>): string =>
  lines.map((line) => line.text + (line.newline ? '\n' : '')).join('');

const emptyFileDiff = (file: string): FileDiffMetadata =>
  parseDiffFromFile({ name: file, contents: '' }, { name: file, contents: '' });

/**
 * Split a unified diff patch for a single file into standalone per-hunk patches.
 *
 * Each returned patch preserves the original file header (everything before the
 * first `@@` hunk header) plus exactly one hunk, producing a patch that can be
 * fed to `git apply` on its own.
 *
 * Returns an empty array when no hunk headers are present.
 */
export const splitPatchIntoHunks = (patch: string): string[] => {
  if (!patch) return [];

  const lines = patch.split(/\r?\n/);
  const hunkHeaderRegex = /^@@\s/;
  const headerLines: string[] = [];
  let firstHunk = 0;
  while (firstHunk < lines.length && !hunkHeaderRegex.test(lines[firstHunk] ?? '')) {
    headerLines.push(lines[firstHunk]);
    firstHunk += 1;
  }

  if (firstHunk >= lines.length) {
    return [];
  }

  const hunks: string[][] = [];
  for (let index = firstHunk; index < lines.length; index += 1) {
    const line = lines[index];
    if (hunkHeaderRegex.test(line ?? '')) {
      hunks.push([...headerLines, line]);
    } else if (hunks.length > 0) {
      hunks[hunks.length - 1].push(line ?? '');
    }
  }

  return hunks.map((hunkLines) => hunkLines.join('\n'))
    .filter((hunk) => hunk.trim().length > 0)
    .map((hunk) => (hunk.endsWith('\n') ? hunk : `${hunk}\n`));
};

/**
 * Extract a standalone patch for a single hunk by zero-based index.
 *
 * Returns `null` when the index is out of range or the patch has no hunks.
 */
export const extractHunkPatch = (patch: string, hunkIndex: number): string | null => {
  if (!Number.isInteger(hunkIndex) || hunkIndex < 0) return null;
  const hunks = splitPatchIntoHunks(patch);
  return hunks[hunkIndex] ?? null;
};