| from __future__ import annotations |
|
|
| from typing import Any |
|
|
| import pandas as pd |
| import plotly.graph_objects as go |
|
|
| from src.i18n import t |
| from src.schemas import Kpis, MethodScore |
|
|
|
|
| def empty_plot(title: str = "—") -> go.Figure: |
| fig = go.Figure() |
| fig.update_layout(template="plotly_white", height=340, title=title) |
| return fig |
|
|
|
|
| def kpi_html(kpis: dict[str, Any] | Kpis, lang: str) -> str: |
| data = kpis.model_dump() if isinstance(kpis, Kpis) else kpis |
|
|
| def card(label: str, value: str, sub: str) -> str: |
| return ( |
| f"<article class='kpi-card'><p class='kpi-label'>{label}</p>" |
| f"<p class='kpi-value'>{value}</p><p class='kpi-sub'>{sub}</p></article>" |
| ) |
|
|
| return ( |
| "<div class='kpi-grid'>" |
| + card(t(lang, "kpi_docs"), str(data.get("n_docs") or 0), data.get("engine_profile") or "policy") |
| + card(t(lang, "kpi_spans"), str(data.get("n_spans") or 0), t(lang, "optimized")) |
| + card( |
| t(lang, "kpi_precision"), |
| f"{100 * float(data.get('policy_precision') or 0):.0f}%", |
| f"baseline {100 * float(data.get('baseline_precision') or 0):.0f}%", |
| ) |
| + card( |
| t(lang, "kpi_recall"), |
| f"{100 * float(data.get('policy_recall') or 0):.0f}%", |
| f"baseline {100 * float(data.get('baseline_recall') or 0):.0f}%", |
| ) |
| + card( |
| t(lang, "kpi_left"), |
| str(data.get("leftover_exports") or 0), |
| f"{data.get('elapsed_ms') or 0} ms", |
| ) |
| + "</div>" |
| ) |
|
|
|
|
| def compare_figure(scores: dict[str, MethodScore] | dict[str, Any], title: str) -> go.Figure: |
| if not scores: |
| return empty_plot(title) |
| names, precision, recall, over = [], [], [], [] |
| for key in ("regex_only", "checksum_policy"): |
| item = None |
| for raw_key, value in scores.items(): |
| name = raw_key.value if hasattr(raw_key, "value") else str(raw_key) |
| if name == key or (hasattr(raw_key, "value") and raw_key.value == key): |
| item = value |
| break |
| if str(raw_key).endswith(key): |
| item = value |
| if item is None: |
| continue |
| names.append(key) |
| precision.append(round(100 * float(item.precision if hasattr(item, "precision") else item.get("precision", 0)), 1)) |
| recall.append(round(100 * float(item.recall if hasattr(item, "recall") else item.get("recall", 0)), 1)) |
| over.append(round(100 * float(item.over_redaction if hasattr(item, "over_redaction") else item.get("over_redaction", 0)), 2)) |
| fig = go.Figure() |
| fig.add_bar(name="precision %", x=names, y=precision, marker_color="#0f766e") |
| fig.add_bar(name="recall %", x=names, y=recall, marker_color="#1d4ed8") |
| fig.add_bar(name="over-redaction %", x=names, y=over, marker_color="#b45309") |
| fig.update_layout(template="plotly_white", height=380, barmode="group", title=title) |
| return fig |
|
|
|
|
| def type_figure(score: MethodScore | dict[str, Any] | None, title: str) -> go.Figure: |
| if score is None: |
| return empty_plot(title) |
| by_type = score.by_type if hasattr(score, "by_type") else (score or {}).get("by_type") or {} |
| if not by_type: |
| return empty_plot(title) |
| names = list(by_type.keys()) |
| precision = [round(100 * float(by_type[name].get("precision", 0)), 1) for name in names] |
| recall = [round(100 * float(by_type[name].get("recall", 0)), 1) for name in names] |
| fig = go.Figure() |
| fig.add_bar(name="precision %", x=names, y=precision, marker_color="#4338ca") |
| fig.add_bar(name="recall %", x=names, y=recall, marker_color="#7c3aed") |
| fig.update_layout(template="plotly_white", height=360, barmode="group", title=title) |
| return fig |
|
|
|
|
| def leftover_figure(doc_rows: list[dict[str, Any]], title: str) -> go.Figure: |
| if not doc_rows: |
| return empty_plot(title) |
| frame = pd.DataFrame(doc_rows) |
| if "leftover" not in frame.columns: |
| return empty_plot(title) |
| fig = go.Figure(go.Bar(x=list(frame["doc_id"].astype(str)), y=list(frame["leftover"]), marker_color="#be123c")) |
| fig.update_layout(template="plotly_white", height=360, title=title) |
| return fig |
|
|
