gptdiff 0.1.10__py3-none-any.whl → 0.1.11__py3-none-any.whl

gptdiff/gptdiff.py

@@ -1,4 +1,9 @@
 #!/usr/bin/env python3
+from pathlib import Path
+import subprocess
+import hashlib
+import re
+
 
 import openai
 from openai import OpenAI
@@ -255,12 +260,20 @@ def build_environment(files_dict):
     return '\n'.join(env)
 
 def generate_diff(environment, goal, model=None, temperature=0.7, max_tokens=32000, api_key=None, base_url=None, prepend=None):
-    """API: Generate diff from environment and goal"""
+    """API: Generate a git diff from the environment and goal.
+
+If 'prepend' is provided, it should be a path to a file whose content will be
+prepended to the system prompt.
+    """
     if model is None:
         model = os.getenv('GPTDIFF_MODEL', 'deepseek-reasoner')
     if prepend:
-        prepend = load_prepend_file(args.prepend)
-        print("Including prepend",len(enc.encode(json.dumps(prepend))), "tokens")
+        if prepend.startswith("http://") or prepend.startswith("https://"):
+            import urllib.request
+            with urllib.request.urlopen(prepend) as response:
+                prepend = response.read().decode('utf-8')
+        else:
+            prepend = load_prepend_file(prepend)
     else:
         prepend = ""
     
@@ -343,18 +356,172 @@ def smartapply(diff_text, files, model=None, api_key=None, base_url=None):
 
     return files
 
-# Function to apply diff to project files
 def apply_diff(project_dir, diff_text):
-    diff_file = Path(project_dir) / "diff.patch"
-    with open(diff_file, 'w') as f:
-        f.write(diff_text)
+    """
+    Applies a unified diff (as generated by git diff) to the files in project_dir
+    using pure Python (without calling the external 'patch' command).
 
-    result = subprocess.run(["patch", "-p1", "-f", "--remove-empty-files", "--input", str(diff_file)], cwd=project_dir, capture_output=True, text=True)
-    if result.returncode != 0:
-        return False
-    else:
+    Handles file modifications, new file creation, and file deletions.
+
+    Returns:
+        True if at least one file was modified (or deleted/created) as a result of the patch,
+        False otherwise.
+    """
+    from pathlib import Path
+    import re, hashlib
+
+    def file_hash(filepath):
+        h = hashlib.sha256()
+        with open(filepath, "rb") as f:
+            h.update(f.read())
+        return h.hexdigest()
+
+    def apply_patch_to_file(file_path, patch):
+        """
+        Applies a unified diff patch (for a single file) to file_path.
+
+        Returns True if the patch was applied successfully, False otherwise.
+        """
+        # Read the original file lines; if the file doesn't exist, treat it as empty.
+        if file_path.exists():
+            original_lines = file_path.read_text(encoding="utf8").splitlines(keepends=True)
+        else:
+            original_lines = []
+        new_lines = []
+        current_index = 0
+
+        patch_lines = patch.splitlines()
+        # Regex for a hunk header, e.g., @@ -3,7 +3,6 @@
+        hunk_header_re = re.compile(r"^@@ -(\d+)(?:,(\d+))? \+(\d+)(?:,(\d+))? @@")
+        i = 0
+        while i < len(patch_lines):
+            line = patch_lines[i]
+            if line.startswith("@@"):
+                m = hunk_header_re.match(line)
+                if not m:
+                    print("Invalid hunk header:", line)
+                    return False
+                orig_start = int(m.group(1))
+                # orig_len = int(m.group(2)) if m.group(2) else 1  # not used explicitly
+                # new_start = int(m.group(3))
+                # new_len = int(m.group(4)) if m.group(4) else 1
+
+                # Copy unchanged lines before the hunk.
+                hunk_start_index = orig_start - 1  # diff headers are 1-indexed
+                if hunk_start_index > len(original_lines):
+                    print("Hunk start index beyond file length")
+                    return False
+                new_lines.extend(original_lines[current_index:hunk_start_index])
+                current_index = hunk_start_index
+
+                i += 1
+                # Process the hunk lines until the next hunk header.
+                while i < len(patch_lines) and not patch_lines[i].startswith("@@"):
+                    pline = patch_lines[i]
+                    if pline.startswith(" "):
+                        # Context line must match exactly.
+                        expected = pline[1:]
+                        if current_index >= len(original_lines):
+                            print("Context line expected but file ended")
+                            return False
+                        orig_line = original_lines[current_index].rstrip("\n")
+                        if orig_line != expected:
+                            print("Context line mismatch. Expected:", expected, "Got:", orig_line)
+                            return False
+                        new_lines.append(original_lines[current_index])
+                        current_index += 1
+                    elif pline.startswith("-"):
+                        # Removal line: verify and skip from original.
+                        expected = pline[1:]
+                        if current_index >= len(original_lines):
+                            print("Removal line expected but file ended")
+                            return False
+                        orig_line = original_lines[current_index].rstrip("\n")
+                        if orig_line != expected:
+                            print("Removal line mismatch. Expected:", expected, "Got:", orig_line)
+                            return False
+                        current_index += 1
+                    elif pline.startswith("+"):
+                        # Addition line: add to new_lines.
+                        new_lines.append(pline[1:] + "\n")
+                    else:
+                        print("Unexpected line in hunk:", pline)
+                        return False
+                    i += 1
+            else:
+                # Skip non-hunk header lines.
+                i += 1
+
+        # Append any remaining lines from the original file.
+        new_lines.extend(original_lines[current_index:])
+        # Ensure parent directories exist before writing the file.
+        file_path.parent.mkdir(parents=True, exist_ok=True)
+        # Write the new content back to the file.
+        file_path.write_text("".join(new_lines), encoding="utf8")
         return True
 
+    # Parse the diff into per-file patches.
+    file_patches = parse_diff_per_file(diff_text)
+    if not file_patches:
+        print("No file patches found in diff.")
+        return False
+
+    # Record original file hashes.
+    original_hashes = {}
+    for file_path, _ in file_patches:
+        target_file = Path(project_dir) / file_path
+        if target_file.exists():
+            original_hashes[file_path] = file_hash(target_file)
+        else:
+            original_hashes[file_path] = None
+
+    any_change = False
+    # Process each file patch.
+    for file_path, patch in file_patches:
+        target_file = Path(project_dir) / file_path
+        if "+++ /dev/null" in patch:
+            # Deletion patch: delete the file if it exists.
+            if target_file.exists():
+                target_file.unlink()
+                if not target_file.exists():
+                    any_change = True
+                else:
+                    print(f"Failed to delete file: {target_file}")
+                    return False
+        else:
+            # Modification or new file creation.
+            success = apply_patch_to_file(target_file, patch)
+            if not success:
+                print(f"Failed to apply patch to file: {target_file}")
+                return False
+
+    # Verify that at least one file was changed by comparing hashes.
+    for file_path, patch in file_patches:
+        target_file = Path(project_dir) / file_path
+        if "+++ /dev/null" in patch:
+            if not target_file.exists():
+                any_change = True
+            else:
+                print(f"Expected deletion but file still exists: {target_file}")
+                return False
+        else:
+            old_hash = original_hashes.get(file_path)
+            if target_file.exists():
+                new_hash = file_hash(target_file)
+                if old_hash != new_hash:
+                    any_change = True
+                else:
+                    print(f"No change detected in file: {target_file}")
+            else:
+                print(f"Expected modification or creation but file is missing: {target_file}")
+                return False
+
+    if not any_change:
+        print("Patch applied but no file modifications detected.")
+        return False
+    return True
+
+
 def parse_arguments():
     parser = argparse.ArgumentParser(description='Generate and optionally apply git diffs using GPT-4.')
     parser.add_argument('prompt', type=str, help='Prompt that runs on the codebase.')
@@ -399,41 +566,76 @@ def parse_diff_per_file(diff_text):
     Note:
         Uses 'b/' prefix detection from git diffs to determine target paths
     """
-    diffs = []
-    file_path = None
-    current_diff = []
-    from_path = None
-
-    for line in diff_text.split('\n'):
-        if line.startswith('diff --git'):
-            if current_diff and file_path is not None:
-                diffs.append((file_path, '\n'.join(current_diff)))
-            current_diff = [line]
-            file_path = None
-            from_path = None
-            parts = line.split()
-            if len(parts) >= 4:
-                b_path = parts[3]
-                file_path = b_path[2:] if b_path.startswith('b/') else b_path
-        else:
-            current_diff.append(line)
-            if line.startswith('--- '):
-                from_path = line[4:].strip()
-            elif line.startswith('+++ '):
-                to_path = line[4:].strip()
-                if to_path == '/dev/null':
-                    if from_path:
-                        # For deletions, use from_path after stripping 'a/' prefix
-                        file_path = from_path[2:] if from_path.startswith('a/') else from_path
-                else:
-                    # For normal cases, use to_path after stripping 'b/' prefix
-                    file_path = to_path[2:] if to_path.startswith('b/') else to_path
-
-    # Handle remaining diff content after loop
-    if current_diff and file_path is not None:
-        diffs.append((file_path, '\n'.join(current_diff)))
-
-    return diffs
+    header_re = re.compile(r'^(?:diff --git\s+)?(a/[^ ]+)\s+(b/[^ ]+)\s*$', re.MULTILINE)
+    lines = diff_text.splitlines()
+
+    # Check if any header line exists.
+    if not any(header_re.match(line) for line in lines):
+        # Fallback strategy: detect file headers starting with '--- a/' or '-- a/'
+        diffs = []
+        current_lines = []
+        current_file = None
+        deletion_mode = False
+        header_line_re = re.compile(r'^-{2,3}\s+a/(.+)$')
+
+        for line in lines:
+            if header_line_re.match(line):
+                if current_file is not None and current_lines:
+                    if deletion_mode and not any(l.startswith("+++ ") for l in current_lines):
+                        current_lines.append("+++ /dev/null")
+                    diffs.append((current_file, "\n".join(current_lines)))
+                current_lines = [line]
+                deletion_mode = False
+                file_from = header_line_re.match(line).group(1).strip()
+                current_file = file_from
+            else:
+                current_lines.append(line)
+                if "deleted file mode" in line:
+                    deletion_mode = True
+                if line.startswith("+++ "):
+                    parts = line.split()
+                    if len(parts) >= 2:
+                        file_to = parts[1].strip()
+                        if file_to != "/dev/null":
+                            current_file = file_to[2:] if (file_to.startswith("a/") or file_to.startswith("b/")) else file_to
+        if current_file is not None and current_lines:
+            if deletion_mode and not any(l.startswith("+++ ") for l in current_lines):
+                current_lines.append("+++ /dev/null")
+            diffs.append((current_file, "\n".join(current_lines)))
+        return diffs
+    else:
+        # Use header-based strategy.
+        diffs = []
+        current_lines = []
+        current_file = None
+        deletion_mode = False
+        for line in lines:
+            m = header_re.match(line)
+            if m:
+                if current_file is not None and current_lines:
+                    if deletion_mode and not any(l.startswith("+++ ") for l in current_lines):
+                        current_lines.append("+++ /dev/null")
+                    diffs.append((current_file, "\n".join(current_lines)))
+                current_lines = [line]
+                deletion_mode = False
+                file_from = m.group(1)  # e.g. "a/index.html"
+                file_to = m.group(2)    # e.g. "b/index.html"
+                current_file = file_to[2:] if file_to.startswith("b/") else file_to
+            else:
+                current_lines.append(line)
+                if "deleted file mode" in line:
+                    deletion_mode = True
+                if line.startswith("+++ "):
+                    parts = line.split()
+                    if len(parts) >= 2:
+                        file_to = parts[1].strip()
+                        if file_to != "/dev/null":
+                            current_file = file_to[2:] if (file_to.startswith("a/") or file_to.startswith("b/")) else file_to
+        if current_file is not None and current_lines:
+            if deletion_mode and not any(l.startswith("+++ ") for l in current_lines):
+                current_lines.append("+++ /dev/null")
+            diffs.append((current_file, "\n".join(current_lines)))
+        return diffs
 
 def call_llm_for_apply_with_think_tool_available(file_path, original_content, file_diff, model, api_key=None, base_url=None, extra_prompt=None, max_tokens=30000):
     parser = FlatXMLParser("think")
@@ -486,14 +688,12 @@ def call_llm_for_apply(file_path, original_content, file_diff, model, api_key=No
         ... )
         >>> print(updated)
         def new(): pass"""
-
     system_prompt = """Please apply the diff to this file. Return the result in a block. Write the entire file.
 
 1. Carefully apply all changes from the diff
 2. Preserve surrounding context that isn't changed
 3. Only return the final file content, do not add any additional markup and do not add a code block
 4. You must return the entire file. It overwrites the existing file."""
-
     user_prompt = f"""File: {file_path}
 File contents:
 <filecontents>
@@ -504,7 +704,6 @@ Diff to apply:
 <diff>
 {file_diff}
 </diff>"""
-
     if extra_prompt:
         user_prompt += f"\n\n{extra_prompt}"
     if model == "gemini-2.0-flash-thinking-exp-01-21":
@@ -513,7 +712,6 @@ Diff to apply:
         {"role": "system", "content": system_prompt},
         {"role": "user", "content": user_prompt},
     ]
-
     if api_key is None:
         api_key = os.getenv('GPTDIFF_LLM_API_KEY')
     if base_url is None:
@@ -525,7 +723,6 @@ Diff to apply:
         temperature=0.0,
         max_tokens=max_tokens)
     full_response = response.choices[0].message.content
-
     elapsed = time.time() - start_time
     minutes, seconds = divmod(int(elapsed), 60)
     time_str = f"{minutes}m {seconds}s" if minutes else f"{seconds}s"
@@ -602,8 +799,9 @@ def main():
         args.model = os.getenv('GPTDIFF_MODEL', 'deepseek-reasoner')
 
     if not args.call and not args.apply:
+        append = "\nInstead of using <diff> tags, use ```diff backticks."
         with open('prompt.txt', 'w') as f:
-            f.write(full_prompt)
+            f.write(full_prompt+append)
         print(f"Total tokens: {token_count:5d}")
         print(f"\033[1;32mNot calling GPT-4.\033[0m")  # Green color for success message
         print('Instead, wrote full prompt to prompt.txt. Use `xclip -selection clipboard < prompt.txt` then paste into chatgpt')
@@ -646,7 +844,6 @@ def main():
             print("\a")  # Terminal bell for completion notification
         return
 
-    # Output result
     elif args.apply:
         print("\nAttempting apply with the following diff:")
         print("\n<diff>")
@@ -728,3 +925,6 @@ def main():
     print(f"Completion tokens: {completion_tokens}")
     print(f"Total tokens: {total_tokens}")
     #print(f"Total cost: ${cost:.4f}")
+
+if __name__ == "__main__":
+    main()

{gptdiff-0.1.10.dist-info → gptdiff-0.1.11.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: gptdiff
-Version: 0.1.10
+Version: 0.1.11
 Summary: A tool to generate and apply git diffs using LLMs
 Author: 255labs
 Classifier: License :: OSI Approved :: MIT License
@@ -26,10 +26,15 @@ Dynamic: requires-dist
 Dynamic: summary
 
 # GPTDiff
+<!--
+GPTDiff: Create and apply diffs using AI.
+This tool leverages natural language instructions to modify project codebases.
+-->
 
-🚀 **Create and apply diffs with AI** - Modify projects with natural language
+🚀 **Create and apply diffs with AI**  
+Modify your project using plain English.
 
-More docs at [gptdiff.255labs.xyz](gptdiff.255labs.xyz)
+More documentation at [gptdiff.255labs.xyz](gptdiff.255labs.xyz)
 
 ### Example Usage of `gptdiff`
 
@@ -63,7 +68,8 @@ cd myproject
 gptdiff 'add hover effects to the buttons'
 ```
 
-Generates a prompt.txt file that you can copy and paste into an LLM
+Generates a prompt.txt file containing the full request.
+Copy and paste its content into your preferred LLM (e.g., ChatGPT) for further experimentation.
 
 ### Simple command line agent loops
 
@@ -147,7 +153,7 @@ First sign up for an API key at https://nano-gpt.com/api and generate your key.
 export GPTDIFF_LLM_API_KEY='your-api-key'
 # Optional: For switching API providers
 export GPTDIFF_MODEL='deepseek-reasoner'  # Set default model for all commands
-export GPTDIFF_LLM_BASE_URL='https://nano-gpt.com/api/v1/
+export GPTDIFF_LLM_BASE_URL='https://nano-gpt.com/api/v1/'
 ```
 
 #### Windows
@@ -171,7 +177,7 @@ Prevent files being appended to the prompt by adding them to `.gitignore` or `.g
 
 ### Command Line Usage
 
-After installing the package, you can use the `gptdiff` command in your terminal. cd into your codebase and run:
+After installing the package, use the `gptdiff` command in your terminal. Change directory into your codebase and run:
 
 ```bash
 gptdiff '<user_prompt>'
@@ -181,13 +187,7 @@ any files that are included in .gitignore are ignored when generating prompt.txt
 
 #### Specifying Additional Files
 
-You can specify additional files or directories to include in the prompt by adding them as arguments to the `gptdiff` command. If no additional files or directories are specified, the tool will default to using the current working directory.
-
-Example usage:
-
-```bash
-gptdiff 'make this change' src test
-```
+You may supply extra files or directories as arguments to the `gptdiff` command. If omitted, the tool defaults to the current working directory, excluding those matching ignore rules.
 
 #### Autopatch Changes
 
@@ -202,24 +202,20 @@ Preview changes without applying them by omitting the `--apply` flag when using
 ```bash
 gptdiff "Modernize database queries" --call
 ```
-<span style="color: #0066cc;">ℹ️ Diff preview generated - review changes before applying</span>
+<span style="color: #0066cc;">i️ Diff preview generated - review changes before applying</span>
 
 This often generates incorrect diffs that need to be manually merged.
 
 #### Smart Apply
 
-For more reliable patching of complex changes, use `smartapply` which processes each file's diff individually with the LLM:
+For robust handling of complex changes, use `smartapply`. It processes each file’s diff individually via the LLM, ensuring nuanced conflict resolution.
 
-```bash
-gptdiff 'refactor authentication system' --apply
-```
-
-### Completion Notification
+## Completion Notification
 
 Use the `--nobeep` option to disable the default completion beep:
 
 ```bash
-gptdiff '<user_prompt>' --beep
+gptdiff '<user_prompt>' --nobeep
 ```
 
 ## Local API Documentation
@@ -242,29 +238,25 @@ import os
 
 os.environ['GPTDIFF_LLM_API_KEY'] = 'your-api-key'
 
-# Create environment representation
-environment = '''
-File: main.py
-Content:
-def old_name():
-    print("Need renaming")
-'''
+# Create files dictionary
+files = {"main.py": "def old_name():\n    print('Need renaming')"}
+
+# Generate transformation diff using an environment string built from the files dictionary
+environment = ""
+for path, content in files.items():
+    environment += f"File: {path}\nContent:\n{content}\n"
 
-# Generate transformation diff
 diff = generate_diff(
     environment=environment,
     goal='Rename function to new_name()',
     model='deepseek-reasoner'
 )
 
-# Apply changes safely
-updated_environment = smartapply(
-    diff_text=diff,
-    environment_str=environment
-)
+# Apply changes safely using the files dict
+updated_files = smartapply(diff, files)
 
 print("Transformed codebase:")
-print(updated_environment)
+print(updated_files["main.py"])
 ```
 
 **Batch Processing Example:**

gptdiff-0.1.11.dist-info/RECORD

@@ -0,0 +1,8 @@
+gptdiff/__init__.py,sha256=yGjgwv7tNvH1ZLPsQyoo1CxpTOl1iCAwwDBp-_17ksQ,89
+gptdiff/gptdiff.py,sha256=HbnFkP1o5jQ-WIC99y-et_X6BeBsArtKqiFb2FG6X28,36694
+gptdiff-0.1.11.dist-info/LICENSE.txt,sha256=zCJk7yUYpMjFvlipi1dKtaljF8WdZ2NASndBYYbU8BY,1228
+gptdiff-0.1.11.dist-info/METADATA,sha256=pCTb2Mu4w0Y6HSAY5A8F8jwrJrLfq_SQw80PyQvRIXA,7799
+gptdiff-0.1.11.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
+gptdiff-0.1.11.dist-info/entry_points.txt,sha256=0yvXYEVAZFI-p32kQ4-h3qKVWS0a86jsM9FAwF89t9w,49
+gptdiff-0.1.11.dist-info/top_level.txt,sha256=XNkQkQGINaDndEwRxg8qToOrJ9coyfAb-EHrSUXzdCE,8
+gptdiff-0.1.11.dist-info/RECORD,,

gptdiff-0.1.10.dist-info/RECORD

@@ -1,8 +0,0 @@
-gptdiff/__init__.py,sha256=yGjgwv7tNvH1ZLPsQyoo1CxpTOl1iCAwwDBp-_17ksQ,89
-gptdiff/gptdiff.py,sha256=78_Y1ifKxCdC-e8TdKm3kKDklyV0K8S0fKyDdXhLNQs,27706
-gptdiff-0.1.10.dist-info/LICENSE.txt,sha256=zCJk7yUYpMjFvlipi1dKtaljF8WdZ2NASndBYYbU8BY,1228
-gptdiff-0.1.10.dist-info/METADATA,sha256=jw4gVLU2Gk7Iuqy-NdkeDlF9loiLfd7825lZxXIbQEY,7602
-gptdiff-0.1.10.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
-gptdiff-0.1.10.dist-info/entry_points.txt,sha256=0yvXYEVAZFI-p32kQ4-h3qKVWS0a86jsM9FAwF89t9w,49
-gptdiff-0.1.10.dist-info/top_level.txt,sha256=XNkQkQGINaDndEwRxg8qToOrJ9coyfAb-EHrSUXzdCE,8
-gptdiff-0.1.10.dist-info/RECORD,,