Enhance clipboard parsing to convert file-like code blocks into new file diffs

Author: robertpiosik

README.md

@@ -76,9 +76,9 @@ It uses a simple [content script](https://github.com/robertpiosik/CodeWebChat/bl
 - Qwen
 - Together
 - Yuanbao
-- Z.AI
+- Z
 
-> <small>**Legal disclaimer:** The injected, yellow [Apply response with CWC] button is not a means of automatic output extraction, it's an alias for the original _copy to clipboard_ button—changes integration process uses clipboard-stored chat response text.</small>
+> <small>**Legal disclaimer:** The injected, yellow _Apply response with CWC_ button is not a means of automatic output extraction, it's an alias for the original _copy to clipboard_ button—changes integration process uses clipboard-stored chat response text.</small>
 
 ## <span style="background-color: #fbb100; color: black; padding: 0.2em 0.6em; border-radius: 999px">API Tools</span>
 

packages/browser/src/content-scripts/send-prompt-content-script/chatbots/grok.ts

@@ -99,7 +99,7 @@ export const grok: Chatbot = {
         get_code_blocks: (t) => t.querySelectorAll('code'),
         perform_copy: (f) => {
           const copy_button = f.querySelector(
-            'button:nth-child(2)'
+            'button:nth-child(4)'
           ) as HTMLElement
           if (!copy_button) {
             report_initialization_error({

packages/vscode/package.json

@@ -2,7 +2,7 @@
   "name": "gemini-coder",
   "displayName": "Code Web Chat",
   "description": "CWC initializes your favorite chatbot with code and instructions, hands-free!",
-  "version": "1.408.0",
+  "version": "1.409.0",
   "scripts": {
     "build": "npx vsce package --no-dependencies",
     "vscode:prepublish": "rimraf out && npm run compile",

packages/vscode/src/commands/apply-chat-response-command/utils/clipboard-parser/clipboard-parser.spec.ts

@@ -213,19 +213,6 @@ describe('clipboard-parser', () => {
       )
     })
 
-    //     it('should handle quoted filenames', () => {
-    //       const text = load_clipboard_text('quoted-filenames.txt')
-    //       const result = parse_multiple_files({
-    //         response: text,
-    //         is_single_root_folder_workspace: true
-    //       })
-
-    //       expect(result).toHaveLength(1)
-    //       expect(result[0].file_path).toBe('src/utils.py')
-    //       expect(result[0].content).toBe(`def add(a, b):
-    //     return a + b`)
-    //     })
-
     it('should parse uncommented filename format', () => {
       const text = load_test_case_file(
         'uncommented-filename',
@@ -724,6 +711,27 @@ describe('clipboard-parser', () => {
       )
     })
 
+    it('should parse multiple diff files format in variant k', () => {
+      const text = load_test_case_file(
+        'diff-multiple-files-variant-k',
+        'diff-multiple-files-variant-k.txt'
+      )
+      const result = parse_response(text, true)
+
+      expect(result.type).toBe('patches')
+      expect(result.patches).toHaveLength(2)
+
+      expect(result.patches![0].file_path).toBe('src/lorem.ts')
+      expect(result.patches![1].file_path).toBe('src/ipsum.ts')
+
+      expect(result.patches![0].content).toBe(
+        load_test_case_file('diff-multiple-files-variant-k', 'file-1.txt')
+      )
+      expect(result.patches![1].content).toBe(
+        load_test_case_file('diff-multiple-files-variant-k', 'file-2.txt')
+      )
+    })
+
     it('should parse multiple diff files format with inner triple backticks', () => {
       const text = load_test_case_file(
         'diff-inner-triple-backticks',

packages/vscode/src/commands/apply-chat-response-command/utils/clipboard-parser/clipboard-parser.ts

@@ -23,13 +23,13 @@ export interface ClipboardContent {
   code_completion?: ClipboardCodeCompletion
 }
 
-const extract_workspace_and_path = (
-  raw_file_path: string,
-  is_single_root_folder_workspace = false
-): { workspace_name?: string; relative_path: string } => {
-  const file_path = raw_file_path.replace(/\\/g, '/')
+const extract_workspace_and_path = (params: {
+  raw_file_path: string
+  is_single_root_folder_workspace?: boolean
+}): { workspace_name?: string; relative_path: string } => {
+  const file_path = params.raw_file_path.replace(/\\/g, '/')
   // If workspace has only one root folder, don't try to extract workspace name
-  if (is_single_root_folder_workspace) {
+  if (params.is_single_root_folder_workspace) {
     return { relative_path: file_path }
   }
 
@@ -113,10 +113,11 @@ export const parse_code_completion = (params: {
           const completion_info = extract_path_and_position(first_line_of_block)
           if (completion_info) {
             const { workspace_name, relative_path } =
-              extract_workspace_and_path(
-                completion_info.path,
-                params.is_single_root_folder_workspace
-              )
+              extract_workspace_and_path({
+                raw_file_path: completion_info.path,
+                is_single_root_folder_workspace:
+                  params.is_single_root_folder_workspace
+              })
             return {
               file_path: relative_path,
               content: cleanup_api_response({
@@ -203,10 +204,11 @@ export const parse_multiple_files = (params: {
         if (extracted_filename) {
           state = 'CONTENT'
           top_level_xml_file_mode = true
-          const { workspace_name, relative_path } = extract_workspace_and_path(
-            extracted_filename,
-            params.is_single_root_folder_workspace
-          )
+          const { workspace_name, relative_path } = extract_workspace_and_path({
+            raw_file_path: extracted_filename,
+            is_single_root_folder_workspace:
+              params.is_single_root_folder_workspace
+          })
           current_file_name = relative_path
           if (workspace_name) {
             current_workspace_name = workspace_name
@@ -340,10 +342,11 @@ export const parse_multiple_files = (params: {
           const extracted_filename = extract_file_path_from_xml(line)
           if (extracted_filename) {
             const { workspace_name, relative_path } =
-              extract_workspace_and_path(
-                extracted_filename,
-                params.is_single_root_folder_workspace
-              )
+              extract_workspace_and_path({
+                raw_file_path: extracted_filename,
+                is_single_root_folder_workspace:
+                  params.is_single_root_folder_workspace
+              })
             current_file_name = relative_path
             if (workspace_name) {
               current_workspace_name = workspace_name
@@ -394,10 +397,11 @@ export const parse_multiple_files = (params: {
 
           if (extracted_filename) {
             const { workspace_name, relative_path } =
-              extract_workspace_and_path(
-                extracted_filename,
-                params.is_single_root_folder_workspace
-              )
+              extract_workspace_and_path({
+                raw_file_path: extracted_filename,
+                is_single_root_folder_workspace:
+                  params.is_single_root_folder_workspace
+              })
             current_file_name = relative_path
             if (workspace_name) {
               current_workspace_name = workspace_name
@@ -498,10 +502,10 @@ export const parse_file_content_only = (params: {
   const extracted_filename = extract_path_from_line_of_code(first_line)
   if (!extracted_filename) return null
 
-  const { workspace_name, relative_path } = extract_workspace_and_path(
-    extracted_filename,
-    params.is_single_root_folder_workspace
-  )
+  const { workspace_name, relative_path } = extract_workspace_and_path({
+    raw_file_path: extracted_filename,
+    is_single_root_folder_workspace: params.is_single_root_folder_workspace
+  })
 
   const content = lines.slice(1).join('\n')
   const cleaned_content = content

packages/vscode/src/commands/apply-chat-response-command/utils/clipboard-parser/extract-diff-patches/extract-diffs.ts

@@ -105,56 +105,134 @@ const process_collected_patch_lines = (
   return patch
 }
 
-const extract_code_block_patches = (normalized_text: string): Diff[] => {
-  const patches: Diff[] = []
-  const lines = normalized_text.split('\n')
-
-  // Parse from end to beginning to find code blocks
-  const code_blocks: { start: number; end: number }[] = []
-
-  // First pass: find all closing ``` and work backwards to find opening ```diff
-  for (let i = lines.length - 1; i >= 0; i--) {
-    const trimmed_line = lines[i].trim()
+const convert_code_block_to_new_file_diff = (lines: string[]): Diff | null => {
+  if (lines.length == 0) {
+    return null
+  }
 
-    if (trimmed_line == '```') {
-      // Found closing backticks, now look backwards for opening
-      for (let j = i - 1; j >= 0; j--) {
-        const opening_line = lines[j].trim()
-        if (opening_line == '```diff') {
-          code_blocks.unshift({ start: j, end: i })
-          i = j // Skip to this position to avoid overlapping blocks
-          break
+  // Regex to find file path. It can be commented or not.
+  // It handles paths with slashes, backslashes, dots, alphanumerics, underscores, and hyphens.
+  const path_regex = /(?:(?:\/\/|#|--|<!--)\s*)?([\w./\\-]+)/
+
+  let file_path: string | undefined
+  let path_line_index = -1
+
+  // Look for a file path in the first few lines of the code block
+  for (let i = 0; i < Math.min(lines.length, 5); i++) {
+    const line = lines[i].trim()
+    const match = line.match(path_regex)
+
+    if (match && match[1]) {
+      const potential_path = match[1]
+
+      if (
+        (potential_path.includes('.') || potential_path.includes('/')) &&
+        !potential_path.includes(' ')
+      ) {
+        const rest_of_line = line
+          .substring(line.indexOf(potential_path) + potential_path.length)
+          .trim()
+        // e.g. `25:5` for line and column, or `-->` for html comments
+        const is_just_path_and_location = /^(?:\d+:\d+)?\s*(-->)?\s*$/.test(
+          rest_of_line
+        )
+
+        if (is_just_path_and_location) {
+          file_path = potential_path.replace(/\\/g, '/') // normalize backslashes
+          path_line_index = i
+          break // Found it, stop searching.
         }
       }
     }
   }
 
-  // Handle unclosed block at the end
-  let last_processed_line = -1
-  if (code_blocks.length > 0) {
-    last_processed_line = code_blocks[code_blocks.length - 1].end
+  if (!file_path) {
+    return null
   }
 
-  let last_opener_index = -1
+  const content_lines = lines.filter((_, index) => index !== path_line_index)
+
+  const patch_lines = content_lines.map((line) => `+${line}`)
+  const patch_content = [
+    `--- /dev/null`,
+    `+++ b/${file_path}`,
+    `@@ -0,0 +1,${content_lines.length} @@`,
+    ...patch_lines
+  ].join('\n')
+
+  return {
+    file_path,
+    content: patch_content + '\n'
+  }
+}
+
+const extract_all_code_block_patches = (normalized_text: string): Diff[] => {
+  const patches: Diff[] = []
+  const lines = normalized_text.split('\n')
+
+  const code_block_regex = /^```(\w*)/
+
+  const code_blocks: { start: number; end: number; type: string }[] = []
+  const stack: { start: number; type: string }[] = []
+
   for (let i = 0; i < lines.length; i++) {
-    const trimmed_line = lines[i].trim()
-    if (trimmed_line == '```diff') {
-      last_opener_index = i
+    const line = lines[i]
+    const trimmed_line = line.trim()
+    const match = trimmed_line.match(code_block_regex)
+
+    if (!match) continue
+
+    const lang = match[1] || ''
+
+    if (stack.length > 0 && stack[stack.length - 1].type == 'diff') {
+      if (line.match(/^[+\- ]/)) {
+        continue
+      }
+    }
+
+    if (lang == '' && stack.length > 0 && trimmed_line == '```') {
+      const open_block = stack.pop()!
+      code_blocks.push({
+        start: open_block.start,
+        end: i,
+        type: open_block.type
+      })
+    } else {
+      stack.push({ start: i, type: lang })
     }
   }
 
-  if (last_opener_index > last_processed_line) {
+  const closed_blocks = [...code_blocks].sort((a, b) => a.start - b.start)
+
+  for (const open_block of stack) {
+    let end_line = lines.length
+    for (const closed_block of closed_blocks) {
+      if (closed_block.start > open_block.start) {
+        end_line = closed_block.start
+        break
+      }
+    }
     code_blocks.push({
-      start: last_opener_index,
-      end: lines.length
+      start: open_block.start,
+      end: end_line,
+      type: open_block.type
     })
   }
 
+  code_blocks.sort((a, b) => a.start - b.start)
+
   // Process each found code block
   for (const block of code_blocks) {
     const block_lines = lines.slice(block.start + 1, block.end) // Exclude the ``` lines
-    const processed_patches = parse_multiple_raw_patches(block_lines)
-    patches.push(...processed_patches)
+    if (block.type == 'diff') {
+      const processed_patches = parse_multiple_raw_patches(block_lines)
+      patches.push(...processed_patches)
+    } else {
+      const patch = convert_code_block_to_new_file_diff(block_lines)
+      if (patch) {
+        patches.push(patch)
+      }
+    }
   }
 
   return patches
@@ -202,10 +280,13 @@ export const extract_diffs = (clipboard_text: string): Diff[] => {
   const normalized_text = clipboard_text.replace(/\r\n/g, '\n')
   const lines = normalized_text.split('\n')
 
-  const uses_code_blocks = lines.some((line) => line.trim() == '```diff')
+  const code_block_regex = /^```(\w*)/
+  const uses_code_blocks = lines.some((line) =>
+    code_block_regex.test(line.trim())
+  )
 
   if (uses_code_blocks) {
-    return extract_code_block_patches(normalized_text)
+    return extract_all_code_block_patches(normalized_text)
   } else {
     return parse_multiple_raw_patches(lines)
   }

packages/vscode/src/commands/apply-chat-response-command/utils/clipboard-parser/test-cases/diff-multiple-files-variant-k/diff-multiple-files-variant-k.txt

@@ -0,0 +1,13 @@
+```html
+// src/lorem.ts
+<div>lorem ipsum</div>
+```
+
+```diff
+--- src/ipsum.ts
++++ src/ipsum.ts
+@@ -1,3 +1,3 @@
+ console.log("hello")
+-console.log("old ipsum")
++console.log("new ipsum")
+```

packages/vscode/src/commands/apply-chat-response-command/utils/clipboard-parser/test-cases/diff-multiple-files-variant-k/file-1.txt

@@ -0,0 +1,4 @@
+--- /dev/null
++++ b/src/lorem.ts
+@@ -0,0 +1,1 @@
++<div>lorem ipsum</div>

packages/vscode/src/commands/apply-chat-response-command/utils/clipboard-parser/test-cases/diff-multiple-files-variant-k/file-2.txt

@@ -0,0 +1,6 @@
+--- a/src/ipsum.ts
++++ b/src/ipsum.ts
+@@ -1,3 +1,3 @@
+ console.log("hello")
+-console.log("old ipsum")
++console.log("new ipsum")