Merge pull request #494 from Cloud-Code-AI/493-feat-update-code-quali…

…ty-prediction

feat: updated the code review prompt and added usage for retriever

.experiments/code_review/evaluate.py

@@ -26,19 +26,23 @@ def compare_issues(ground_truth, model_issues):
             description_similarity = calculate_similarity(
                 gt_issue["description"], model_issue["description"]
             )
+            # actual_code_similarity = calculate_similarity(
+            #     gt_issue.get("current_code",''), model_issue.get("current_code", '')
+            # )
             if (
-                description_similarity > 0.1
+                description_similarity > 0.25
                 and gt_issue["file_path"] == model_issue["file_path"]
-                and abs(
-                    int(gt_issue.get("start_line", 0))
-                    - int(model_issue.get("start_line", -10))
-                )
-                <= 1
-                and abs(
-                    int(gt_issue.get("end_line", 0))
-                    - int(model_issue.get("end_line", -10))
-                )
-                <= 1
+                # and actual_code_similarity > 0.5
+                # and abs(
+                #     int(gt_issue.get("start_line", 0))
+                #     - int(model_issue.get("start_line", -10))
+                # )
+                # <= 1
+                # and abs(
+                #     int(gt_issue.get("end_line", 0))
+                #     - int(model_issue.get("end_line", -10))
+                # )
+                # <= 1
                 and abs(
                     int(gt_issue.get("severity", 0))
                     - int(model_issue.get("severity", -10))

kaizen/helpers/parser.py

@@ -179,9 +179,11 @@ def patch_to_separate_chunks(patch_text):
     return "\n".join(output)
 
 
-def format_change(old_num, new_num, change_type, content):
-    old_num_str = f"{old_num:<4}" if old_num is not None else "    "
-    new_num_str = f"{new_num:<4}" if new_num is not None else "    "
+def format_change(old_num, new_num, change_type, content, ignore_deletions=False):
+    old_num_str = f"{old_num:<5}" if old_num is not None else "_    "
+    new_num_str = f"{new_num:<5}" if new_num is not None else "_    "
+    if ignore_deletions:
+        old_num_str = "_    "
     return f"{old_num_str} {new_num_str} {change_type} {content}"
 
 
@@ -235,19 +237,31 @@ def patch_to_combined_chunks(patch_text, ignore_deletions=False):
         elif line.startswith("-"):
             content = line[1:]
             if not ignore_deletions:
-                changes.append(format_change(removal_line_num, None, "-1:[-]", content))
+                changes.append(
+                    format_change(
+                        removal_line_num, None, "-1:[-]", content, ignore_deletions
+                    )
+                )
             removal_line_num += 1
             unedited_removal_num = removal_line_num
         elif line.startswith("+"):
             content = line[1:]
-            changes.append(format_change(None, addition_line_num, "+1:[+]", content))
+            changes.append(
+                format_change(
+                    None, addition_line_num, "+1:[+]", content, ignore_deletions
+                )
+            )
             addition_line_num += 1
             unedited_addition_num = addition_line_num
         else:
             content = line
             changes.append(
                 format_change(
-                    unedited_removal_num, unedited_addition_num, " 0:[.]", content
+                    unedited_removal_num,
+                    unedited_addition_num,
+                    " 0:[.]",
+                    content,
+                    ignore_deletions,
                 )
             )
             unedited_removal_num += 1

kaizen/llms/prompts/code_review_prompts.py

@@ -24,6 +24,16 @@
   ]
 }}
 
+## Code Quality Percentage Guidelines:
+- Consider the following factors when calculating the final percentage:
+  - Code readability and maintainability
+  - Adherence to best practices and coding standards
+  - Presence of comments and documentation
+  - Efficient use of resources
+  - Proper error handling and edge case consideration
+- Adjust the final percentage based on overall code quality assessment
+- Ensure the final percentage is between 0 and 100
+
 ## Guidelines:
 - Provide specific feedback with file paths and line numbers
 - Use markdown for code snippets. Make sure all code is following the original indentations.
@@ -56,7 +66,6 @@
 4. A new line 'result += y' was added, incorporating the new parameter.
 5. The return statement remained unchanged.
 
-
 ## Review Focus:
 1. Removals (-1:[-]): Identify if removal causes problems in remaining code. Remember any line having -1:[-] is removed line from the new code.
 2. Additions (+1:[+]): Provide detailed feedback and suggest improvements. Remember any line having +1:[+] is added line.
@@ -65,10 +74,10 @@
 5. For 'fixed_code' -> always suggest changes for Additions. 
 
 ## Field Guidelines:
-- "fixed_code": Corrected code for additions only, between start_line and end_line. make sure start_line you suggest does not has `0:[.]`.
-- "actual_code": Current Code line which you think has error. make sure it always done on `+1:[+]` lines. If not, keep it empty ''.
+- "suggested_code": Corrected code for additions only, between start_line and end_line. make sure start_line you suggest does not has `0:[.]`.
+- "current_code": Current Code line which you think has error. make sure it always done on `+1:[+]` lines. If not, keep it empty ''.
 - "start_line" and "end_line": Actual line numbers in the additions.
-- "severity_level": 1 (least severe) to 10 (most critical).
+- "severity": 1 (least severe) to 10 (most critical).
 Prioritize issues based on their potential impact on code quality, functionality, and maintainability. Provide concrete examples or code snippets when suggesting improvements.
 
 ## PATCH DATA:
@@ -97,6 +106,15 @@
   ]
 }}
 
+## Code Quality Percentage Guidelines:
+- Consider the following factors when calculating the final percentage:
+  - Code readability and maintainability
+  - Adherence to best practices and coding standards
+  - Presence of comments and documentation
+  - Efficient use of resources
+  - Proper error handling and edge case consideration
+- Adjust the final percentage based on overall code quality assessment
+- Ensure the final percentage is between 0 and 100
 
 ## Guidelines:
 - Provide specific feedback with file paths and line numbers
@@ -129,20 +147,21 @@
 3. The line 'result = x * 2' remained unchanged.
 4. A new line 'result += y' was added, incorporating the new parameter.
 5. The return statement remained unchanged.
+6. First number is line number of old file. Second number is line number of new filee
 
 
 ## Review Focus:
 1. Removals (-1:[-]): Identify if removal causes problems in remaining code. Remember any line having -1:[-] is removed line from the new code.
 2. Additions (+1:[+]): Provide detailed feedback and suggest improvements. Remember any line having +1:[+] is added line.
 3. Consider impact of changes on overall code structure and functionality.
 4. Note unchanged lines (0:[.]) for context.
-5. For 'fixed_code' -> always suggest changes for Additions. 
+5. For 'suggested_code' -> always suggest changes for Additions. 
 
 ## Field Guidelines:
-- "fixed_code": Corrected code for additions only, between start_line and end_line. make sure start_line you suggest does not has `0:[.]`.
-- "actual_code": Current Code line which you think has error. make sure it always done on `+1:[+]` lines. If not, keep it empty ''.
+- "suggested_code": Corrected code for additions only, between start_line and end_line. make sure start_line you suggest does not has `0:[.]`.
+- "current_code": Current Code line which you think has error. make sure it always done on `+1:[+]` lines. If not, keep it empty ''.
 - "start_line" and "end_line": Actual line numbers in the additions.
-- "severity_level": 1 (least severe) to 10 (most critical).
+- "severity": 1 (least severe) to 10 (most critical).
 
 Prioritize issues based on their potential impact on code quality, functionality, and maintainability. Provide concrete examples or code snippets when suggesting improvements.
 
@@ -213,10 +232,10 @@
 2. Additions (+1:[+]): Provide detailed feedback and suggest improvements. Remember any line having +1:[+] is added line.
 3. Consider impact of changes on overall code structure and functionality.
 4. Note unchanged lines (0:[.]) for context.
-5. For 'fixed_code' -> always suggest changes for Additions. 
+5. For 'suggested_code' -> always suggest changes for Additions. 
 
 ## Field Guidelines:
-- "fixed_code": Corrected code for additions only, between start_line and end_line. make sure start_line you suggest does not has `0:[.]`.
+- "suggested_code": Corrected code for additions only, between start_line and end_line. make sure start_line you suggest does not has `0:[.]`.
 - "actual_code": Current Code line which you think has error. make sure it always done on `+1:[+]` lines. If not, keep it empty ''.
 - "start_line" and "end_line": Actual line numbers in the additions.
 - "severity_level": 1 (least severe) to 10 (most critical).

kaizen/tests/helpers/test_patch_parser.py

@@ -527,4 +527,4 @@ def _process_file_chunk(
 +"""
 '''
 
-print(patch_to_combined_chunks(patch_text=patch_data2))
+print(patch_to_combined_chunks(patch_text=patch_data2, ignore_deletions=True))