Metric changes

Author: gabeorlanski

src/slop_code/metrics/checkpoint/extractors.py

@@ -202,6 +202,12 @@ def _build_metrics_from_snapshot(
         "branches_concentration": functions["branches_concentration"],
         "control_mean": functions["control_mean"],
         "control_concentration": functions["control_concentration"],
+        # Size concentration (Gini for LOC and statements across functions)
+        "lines_concentration": functions.get("lines_concentration", 0.0),
+        "statements_mean": functions.get("statements_mean", 0.0),
+        "statements_concentration": functions.get(
+            "statements_concentration", 0.0
+        ),
         # AST-grep
         "ast_grep_violations": ast_grep["violations"],
         # Source file tracking

src/slop_code/metrics/driver.py

@@ -209,6 +209,7 @@ def _compute_function_stats(
     cc_values: list[int],
     depth_values: list[int],
     lines_values: list[int],
+    statements_values: list[int],
     control_blocks_values: list[int],
     branches_values: list[int],
     comparisons_values: list[int],
@@ -241,9 +242,17 @@ def _compute_function_stats(
     # Depth stats
     depth_max = max(depth_values) if depth_values else 0
 
-    # Lines stats
+    # Lines stats (LOC per function)
     lines_sum = sum(lines_values)
     lines_mean = statistics.mean(lines_values) if lines_values else 0.0
+    lines_concentration = _concentration_score(lines_values)
+
+    # Statements stats (statements per function)
+    statements_sum = sum(statements_values)
+    statements_mean = (
+        statistics.mean(statements_values) if statements_values else 0.0
+    )
+    statements_concentration = _concentration_score(statements_values)
 
     # Control blocks
     control_blocks_sum = sum(control_blocks_values)
@@ -280,6 +289,10 @@ def _dist_stats(
         depth_max=depth_max,
         lines_sum=lines_sum,
         lines_mean=lines_mean,
+        lines_concentration=lines_concentration,
+        statements_sum=statements_sum,
+        statements_mean=statements_mean,
+        statements_concentration=statements_concentration,
         control_blocks_sum=control_blocks_sum,
         # Distribution stats
         nesting_mean=nesting_mean,
@@ -392,6 +405,7 @@ def compute_aggregates(
     func_cc: list[int] = []
     func_depth: list[int] = []
     func_lines: list[int] = []
+    func_statements: list[int] = []
     func_control_blocks: list[int] = []
     func_branches: list[int] = []
     func_comparisons: list[int] = []
@@ -412,6 +426,7 @@ def compute_aggregates(
                 func_cc.append(sym.complexity)
                 func_depth.append(sym.max_nesting_depth)
                 func_lines.append(sym.lines)
+                func_statements.append(sym.statements)
                 func_control_blocks.append(sym.control_blocks)
                 func_branches.append(sym.branches)
                 func_comparisons.append(sym.comparisons)
@@ -426,6 +441,7 @@ def compute_aggregates(
         func_cc,
         func_depth,
         func_lines,
+        func_statements,
         func_control_blocks,
         func_branches,
         func_comparisons,

src/slop_code/metrics/models.py

@@ -342,9 +342,14 @@ class FunctionStats(BaseModel):
     cc_concentration: float = 0.0
     # Depth stats
     depth_max: int = 0
-    # Lines stats
+    # Lines stats (LOC per function)
     lines_sum: int = 0
     lines_mean: float = 0.0
+    lines_concentration: float = 0.0
+    # Statements stats (statements per function)
+    statements_sum: int = 0
+    statements_mean: float = 0.0
+    statements_concentration: float = 0.0
     # Control blocks
     control_blocks_sum: int = 0
 

src/slop_code/visualization/chart_builders.py

@@ -288,13 +288,15 @@ class ProgressLineChartConfig:
         title: Chart title
         width: Figure width
         height: Figure height
+        num_bins: Number of progress bins (default 5 = 20% bins)
     """
 
     metric_col: str
     y_title: str
     title: str
     width: int = GRAPH_WIDTH
     height: int = GRAPH_HEIGHT
+    num_bins: int = 5
 
 
 class ProgressLineChartBuilder:
@@ -325,7 +327,9 @@ def build(
             color_map = MODEL_COLORS
 
         high = filter_high_thinking_checkpoints(checkpoints)
-        progress_data = compute_progress_metric(high, self.config.metric_col)
+        progress_data = compute_progress_metric(
+            high, self.config.metric_col, num_bins=self.config.num_bins
+        )
 
         fig = go.Figure()
 

src/slop_code/visualization/data_transforms.py

@@ -146,16 +146,19 @@ def compute_progress_bins(df: pd.DataFrame) -> pd.DataFrame:
     return df
 
 
-def compute_progress_metric(df: pd.DataFrame, metric_col: str) -> pd.DataFrame:
+def compute_progress_metric(
+    df: pd.DataFrame, metric_col: str, num_bins: int = 5
+) -> pd.DataFrame:
     """Compute progress % and aggregate metric by model.
 
     Progress = checkpoint_idx / total_checkpoints for that problem.
-    Bins into 5 buckets (0.2, 0.4, 0.6, 0.8, 1.0) and averages.
+    Bins into buckets and averages.
     If multiple checkpoints from same problem fall in same bin, take the later one.
 
     Args:
         df: Filtered checkpoints DataFrame with 'cp_num' column
         metric_col: Column name to aggregate
+        num_bins: Number of progress bins (default 5 = 20% bins, use 10 for 10% bins)
 
     Returns:
         DataFrame with columns: model, progress_bin, {metric_col}
@@ -164,8 +167,11 @@ def compute_progress_metric(df: pd.DataFrame, metric_col: str) -> pd.DataFrame:
     df = df.copy()
     df["max_cp"] = df["problem"].map(max_cp)
     df["progress"] = df["cp_num"] / df["max_cp"]
-    df["progress_bin"] = np.ceil(df["progress"] * 5) / 5  # 20% bins
-    df.loc[df["progress_bin"] == 0, "progress_bin"] = 0.2  # Move 0 to first bin
+    bin_size = 1.0 / num_bins
+    df["progress_bin"] = np.ceil(df["progress"] * num_bins) / num_bins
+    df.loc[df["progress_bin"] == 0, "progress_bin"] = (
+        bin_size  # Move 0 to first bin
+    )
 
     # If multiple checkpoints from same problem in same bin, take the later one
     df = df.sort_values("cp_num", ascending=False)