| """ |
| Plotly figure builders using the "standard" PDF-uncertainty style the |
| user prefers: a central curve + uncertainty band computed via LHAPDF's |
| own `PDFSet.uncertainty()`, and a Hessian-style correlation contour |
| between two flavors. |
| |
| IMPORTANT: `PDFSet.uncertainty()` is only mathematically valid for |
| error types "hessian", "symmhessian", and "replicas". Every function |
| here assumes the caller already validated that with |
| `services.pdf_utils.load_validated_pdf_set` — these functions do not |
| re-check it themselves, to keep the validation in exactly one place. |
| |
| Nothing in this module talks to Hugging Face or FastAPI directly; each |
| function takes an already-loaded `pdf_set` and returns a |
| `plotly.graph_objects.Figure`. |
| """ |
|
|
| import logging |
| import time |
| from typing import Optional |
|
|
| import numpy as np |
| import plotly.graph_objects as go |
|
|
| from services.colors import get_main_plot_label, get_parton_color, get_parton_name, get_parton_symbol |
|
|
| logger = logging.getLogger(__name__) |
|
|
|
|
| def build_standard_plot( |
| pdf_set, |
| pdfs, |
| grid_name: str, |
| q_scale: float, |
| parton_id: int, |
| plot_color: Optional[str] = None, |
| x_vals=None, |
| custom_title: Optional[str] = None, |
| ) -> go.Figure: |
| """ |
| Central value + uncertainty band for x*f(x, Q) of a single flavor. |
| |
| `pdfs` is `pdf_set.mkPDFs()`, passed in already-built (see |
| services.pdf_utils.get_pdf_members) so repeat requests for the same |
| PDF set don't reload every member's grid file from disk each time. |
| """ |
| plot_color = plot_color or get_parton_color(parton_id) |
|
|
| if x_vals is None: |
| x_vals = np.logspace(-5, -0.01, 150) |
|
|
| logger.info( |
| "build_standard_plot: grid=%s parton_id=%s q_scale=%s n_points=%d members=%d", |
| grid_name, parton_id, q_scale, len(x_vals), len(pdfs), |
| ) |
| t0 = time.monotonic() |
| central_values, upper_band, lower_band = [], [], [] |
| for x in x_vals: |
| vals = [pdf.xfxQ(parton_id, x, q_scale) for pdf in pdfs] |
| uncertainty = pdf_set.uncertainty(vals) |
|
|
| central = uncertainty.central |
| central_values.append(round(central, 6)) |
| upper_band.append(round(central + uncertainty.errplus, 6)) |
| lower_band.append(round(central - uncertainty.errminus, 6)) |
| logger.info("build_standard_plot: evaluated %d points in %.2fs", len(x_vals), time.monotonic() - t0) |
|
|
| fig = go.Figure() |
|
|
| x_vals_rounded = np.round(x_vals, 6) |
| x_band = np.concatenate([x_vals_rounded, x_vals_rounded[::-1]]) |
| y_band = np.concatenate([upper_band, lower_band[::-1]]) |
|
|
| fig.add_trace(go.Scatter( |
| x=x_band, |
| y=y_band, |
| fill="toself", |
| fillcolor=plot_color, |
| opacity=0.3, |
| line=dict(color="rgba(255,255,255,0)"), |
| name="Uncertainty", |
| hoverinfo="skip", |
| )) |
|
|
| hover_custom_data = np.stack((upper_band, lower_band), axis=-1) |
| fig.add_trace(go.Scatter( |
| x=x_vals, |
| y=central_values, |
| mode="lines", |
| line=dict(color=plot_color, width=2), |
| name=f"{grid_name} Central <br>Q={q_scale} GeV", |
| customdata=hover_custom_data, |
| hovertemplate=( |
| "<b>x:</b> %{x:.4e}<br>" |
| "<b>Central:</b> %{y:.4f}<br>" |
| "<b>Max (Upper):</b> %{customdata[0]:.4f}<br>" |
| "<b>Min (Lower):</b> %{customdata[1]:.4f}" |
| "<extra></extra>" |
| ), |
| )) |
|
|
| default_title = " " |
| fig.update_layout( |
| title=custom_title if custom_title else default_title, |
| xaxis_title="$x$", |
| yaxis_title=f"$x {get_parton_symbol(parton_id)}(x, Q)$", |
| template="plotly_white", |
| paper_bgcolor="rgba(0,0,0,0)", |
| plot_bgcolor="rgba(0,0,0,0)", |
| autosize=True, |
| legend=dict(x=0.02, y=0.98, xanchor="left", yanchor="top", bgcolor="rgba(255,255,255,0.8)"), |
| margin=dict(l=60, r=40, t=60, b=60), |
| ) |
| fig.update_xaxes( |
| type="log", |
| exponentformat="power", |
| showline=True, |
| linewidth=1.5, |
| linecolor="black", |
| mirror="allticks", |
| ticks="inside", |
| tickwidth=1.5, |
| tickcolor="black", |
| showgrid=False, |
| ) |
|
|
| y_upper_limit = max(upper_band) * 1.30 |
| y_lower_limit = min(lower_band) * (1 - 0.20) |
| fig.update_yaxes( |
| range=[y_lower_limit, y_upper_limit], |
| showline=True, |
| zeroline=False, |
| linewidth=1.5, |
| linecolor="black", |
| mirror="allticks", |
| ticks="inside", |
| tickwidth=1.5, |
| tickcolor="black", |
| showgrid=False, |
| ) |
| return fig |
|
|
|
|
| def build_standard_ratio_plot( |
| pdf_set, |
| pdfs, |
| grid_name: str, |
| q_scale: float, |
| parton_id: int, |
| plot_color: Optional[str] = None, |
| x_vals=None, |
| custom_title: Optional[str] = None, |
| ) -> go.Figure: |
| """ |
| Same as build_standard_plot, but normalized to the central value at |
| each x (i.e. central line sits at 1.0, band shows relative uncertainty). |
| |
| `pdfs` is `pdf_set.mkPDFs()`, passed in already-built — see |
| services.pdf_utils.get_pdf_members. |
| """ |
| plot_color = plot_color or get_parton_color(parton_id) |
|
|
| if x_vals is None: |
| x_vals = np.logspace(-5, -0.01, 150) |
|
|
| logger.info( |
| "build_standard_ratio_plot: grid=%s parton_id=%s q_scale=%s n_points=%d members=%d", |
| grid_name, parton_id, q_scale, len(x_vals), len(pdfs), |
| ) |
| t0 = time.monotonic() |
| central_values, upper_band, lower_band = [], [], [] |
| for x in x_vals: |
| vals = [pdf.xfxQ(parton_id, x, q_scale) for pdf in pdfs] |
| uncertainty = pdf_set.uncertainty(vals) |
| central1 = uncertainty.central |
|
|
| if abs(central1) < 1e-10: |
| |
| central_values.append(1.0) |
| upper_band.append(1.0) |
| lower_band.append(1.0) |
| else: |
| central_values.append(round(central1 / central1, 6)) |
| upper_band.append(round(1.0 + uncertainty.errplus / central1, 6)) |
| lower_band.append(round(1.0 - uncertainty.errminus / central1, 6)) |
| logger.info("build_standard_ratio_plot: evaluated %d points in %.2fs", len(x_vals), time.monotonic() - t0) |
|
|
| fig = go.Figure() |
|
|
| x_vals_rounded = np.round(x_vals, 6) |
| x_band = np.concatenate([x_vals_rounded, x_vals_rounded[::-1]]) |
| y_band = np.concatenate([upper_band, lower_band[::-1]]) |
|
|
| fig.add_trace(go.Scatter( |
| x=x_band, |
| y=y_band, |
| fill="toself", |
| fillcolor=plot_color, |
| opacity=0.3, |
| line=dict(color="rgba(255,255,255,0)"), |
| name="Uncertainty", |
| hoverinfo="skip", |
| )) |
|
|
| hover_custom_data = np.stack((upper_band, lower_band), axis=-1) |
| fig.add_trace(go.Scatter( |
| x=x_vals, |
| y=central_values, |
| mode="lines", |
| line=dict(color=plot_color, width=2), |
| name=f"{grid_name} Central <br>Q={q_scale} GeV", |
| customdata=hover_custom_data, |
| hovertemplate=( |
| "<b>x:</b> %{x:.4e}<br>" |
| "<b>Central:</b> %{y:.4f}<br>" |
| "<b>Max (Upper):</b> %{customdata[0]:.4f}<br>" |
| "<b>Min (Lower):</b> %{customdata[1]:.4f}" |
| "<extra></extra>" |
| ), |
| )) |
|
|
| default_title = " " |
| fig.update_layout( |
| title=custom_title if custom_title else default_title, |
| xaxis_title="$x$", |
| yaxis_title=f"$x {get_parton_symbol(parton_id)}(x, Q)/central$", |
| template="plotly_white", |
| paper_bgcolor="rgba(0,0,0,0)", |
| plot_bgcolor="rgba(0,0,0,0)", |
| autosize=True, |
| legend=dict(x=0.02, y=0.98, xanchor="left", yanchor="top", bgcolor="rgba(255,255,255,0.8)"), |
| margin=dict(l=60, r=40, t=60, b=60), |
| ) |
| fig.update_xaxes( |
| type="log", |
| exponentformat="power", |
| showline=True, |
| linewidth=1.5, |
| linecolor="black", |
| mirror="allticks", |
| ticks="inside", |
| tickwidth=1.5, |
| tickcolor="black", |
| showgrid=False, |
| ) |
|
|
| y_upper_limit = max(upper_band) * 1.30 |
| ymin_data = min(lower_band) |
| y_lower_limit = ymin_data * (1 - 0.30) if ymin_data >= 0 else ymin_data * (1 + 0.30) |
|
|
| fig.update_yaxes( |
| range=[y_lower_limit, y_upper_limit], |
| showline=True, |
| zeroline=False, |
| linewidth=1.5, |
| linecolor="black", |
| mirror="allticks", |
| ticks="inside", |
| tickwidth=1.5, |
| tickcolor="black", |
| showgrid=False, |
| ) |
| return fig |
|
|
|
|
| def build_correlation_plot( |
| pdfs, |
| q_scale: float, |
| parton_id_1: int, |
| parton_id_2: int, |
| x_vals=None, |
| color_correlated: Optional[str] = None, |
| color_anti_correlated: str = "black", |
| ) -> go.Figure: |
| """ |
| Hessian-style correlation contour between two flavors across an |
| x-grid, at a fixed Q scale. |
| |
| `pdfs` is `pdf_set.mkPDFs()`, passed in already-built — see |
| services.pdf_utils.get_pdf_members. |
| """ |
| color_correlated = color_correlated or get_parton_color(parton_id_1) |
|
|
| num_members = len(pdfs) |
|
|
| if x_vals is None: |
| x_vals = np.logspace(-4, -0.01, 35) |
|
|
| x_vals_rounded = np.round(x_vals, 6) |
| num_x = len(x_vals_rounded) |
|
|
| logger.info( |
| "build_correlation_plot: parton_id_1=%s parton_id_2=%s q_scale=%s n_points=%d members=%d", |
| parton_id_1, parton_id_2, q_scale, num_x, num_members, |
| ) |
| t0 = time.monotonic() |
| matrix_p1 = np.zeros((num_members, num_x)) |
| matrix_p2 = np.zeros((num_members, num_x)) |
| for m_idx, pdf in enumerate(pdfs): |
| for x_idx, x in enumerate(x_vals_rounded): |
| matrix_p1[m_idx, x_idx] = pdf.xfxQ(parton_id_1, x, q_scale) |
| matrix_p2[m_idx, x_idx] = pdf.xfxQ(parton_id_2, x, q_scale) |
| logger.info("build_correlation_plot: evaluated %d members x %d points in %.2fs", num_members, num_x, time.monotonic() - t0) |
|
|
| correlation_matrix = np.zeros((num_x, num_x)) |
| p1_central = matrix_p1[0, :] |
| p2_central = matrix_p2[0, :] |
| p1_dev = matrix_p1[1:, :] - p1_central |
| p2_dev = matrix_p2[1:, :] - p2_central |
|
|
| for i in range(num_x): |
| for j in range(num_x): |
| delta_x = p1_dev[:, i] |
| delta_y = p2_dev[:, j] |
|
|
| numerator = np.sum(delta_x * delta_y) |
| sum_x_sq = np.sum(delta_x ** 2) |
| sum_y_sq = np.sum(delta_y ** 2) |
| denominator = np.sqrt(sum_x_sq * sum_y_sq) |
|
|
| if denominator > 1e-10: |
| corr_value = np.clip(numerator / denominator, -1.0, 1.0) |
| correlation_matrix[j, i] = round(corr_value, 4) |
| else: |
| correlation_matrix[j, i] = 0.0 |
|
|
| name_1 = get_parton_symbol(parton_id_1) |
| name_2 = get_parton_symbol(parton_id_2) |
|
|
| fig = go.Figure(data=go.Contour( |
| z=correlation_matrix, |
| x=x_vals_rounded, |
| y=x_vals_rounded, |
| colorscale=[ |
| [0.0, color_anti_correlated], |
| [0.5, "rgb(245, 245, 245)"], |
| [1.0, color_correlated], |
| ], |
| zmin=-1.0, |
| zmax=1.0, |
| line=dict(width=1, color="rgba(0, 0, 0, 0.2)"), |
| contours=dict(coloring="heatmap", showlines=False), |
| ncontours=15, |
| colorbar=dict( |
| title=dict(text="Corr. Coefficient", side="right"), |
| thickness=15, |
| len=1.0, |
| ), |
| hovertemplate=( |
| f"<b>x₁ ({name_1}):</b> %{{x:.4e}}<br>" |
| f"<b>x₂ ({name_2}):</b> %{{y:.4e}}<br>" |
| "<b>Correlation:</b> %{z:.4f}<extra></extra>" |
| ), |
| )) |
|
|
| fig.update_layout( |
| title=" ", |
| template="plotly_white", |
| paper_bgcolor="rgba(0,0,0,0)", |
| plot_bgcolor="rgba(0,0,0,0)", |
| margin=dict(l=70, r=20, t=70, b=70), |
| autosize=True, |
| ) |
| fig.update_xaxes( |
| title_text=f"$x_1 \\text{{ in }} {name_1}(x_1, Q)$", |
| type="log", |
| exponentformat="power", |
| showline=True, |
| linewidth=1.5, |
| linecolor="black", |
| mirror="allticks", |
| ticks="inside", |
| tickwidth=1.5, |
| tickcolor="black", |
| showgrid=False, |
| ) |
| fig.update_yaxes( |
| title_text=f"$x_2 \\text{{ in }} {name_2}(x_2, Q)$", |
| type="log", |
| exponentformat="power", |
| showline=True, |
| linewidth=1.5, |
| linecolor="black", |
| mirror="allticks", |
| ticks="inside", |
| tickwidth=1.5, |
| tickcolor="black", |
| showgrid=False, |
| ) |
| return fig |
|
|
|
|
| def build_main_plot( |
| pdf_set, |
| pdfs, |
| grid_name: str, |
| q_scale: float, |
| parton_ids: list, |
| x_vals=None, |
| custom_title: Optional[str] = None, |
| ) -> go.Figure: |
| """ |
| The classic multi-flavor "overview" plot (CT18 Fig. 2 style, see |
| arXiv:1912.10053): every flavor in parton_ids overlaid on one figure as |
| central value + 90% C.L. uncertainty band, log-x / linear-y. The gluon |
| (parton_id 21) is always scaled down by a fixed factor of 5 so it fits |
| on the same y-scale as the quark flavors, following that same |
| convention -- its legend entry is labelled "g/5" accordingly (see |
| services.colors.get_main_plot_label). |
| |
| `pdfs` is `pdf_set.mkPDFs()`, passed in already-built (see |
| services.pdf_utils.get_pdf_members). |
| """ |
| if x_vals is None: |
| x_vals = np.logspace(-6, np.log10(0.9), 500) |
|
|
| logger.info( |
| "build_main_plot: grid=%s parton_ids=%s q_scale=%s n_points=%d members=%d", |
| grid_name, parton_ids, q_scale, len(x_vals), len(pdfs), |
| ) |
| t0 = time.monotonic() |
|
|
| fig = go.Figure() |
| x_vals_rounded = np.round(x_vals, 6) |
| x_band = np.concatenate([x_vals_rounded, x_vals_rounded[::-1]]) |
|
|
| all_upper_bands = [] |
|
|
| for parton_id in parton_ids: |
| plot_color = get_parton_color(parton_id) |
| label = get_main_plot_label(parton_id) |
| scale = 1.0 / 5.0 if parton_id == 21 else 1.0 |
|
|
| central_values, upper_band, lower_band = [], [], [] |
| for x in x_vals: |
| vals = [pdf.xfxQ(parton_id, x, q_scale) for pdf in pdfs] |
| uncertainty = pdf_set.uncertainty(vals) |
|
|
| central = uncertainty.central * scale |
| central_values.append(round(central, 6)) |
| upper_band.append(round(central + uncertainty.errplus * scale, 6)) |
| lower_band.append(round(central - uncertainty.errminus * scale, 6)) |
|
|
| all_upper_bands.extend(upper_band) |
|
|
| y_band = np.concatenate([upper_band, lower_band[::-1]]) |
| fig.add_trace(go.Scatter( |
| x=x_band, |
| y=y_band, |
| fill="toself", |
| fillcolor=plot_color, |
| opacity=0.3, |
| line=dict(color="rgba(255,255,255,0)"), |
| name=label + " (unc.)", |
| legendgroup=label, |
| showlegend=False, |
| hoverinfo="skip", |
| )) |
|
|
| hover_custom_data = np.stack((upper_band, lower_band), axis=-1) |
| fig.add_trace(go.Scatter( |
| x=x_vals, |
| y=central_values, |
| mode="lines", |
| line=dict(color=plot_color, width=2), |
| name=f"${label}$", |
| legendgroup=label, |
| customdata=hover_custom_data, |
| hovertemplate=( |
| f"<b>Flavor:</b> {label}<br>" |
| "<b>x:</b> %{x:.4e}<br>" |
| "<b>Central:</b> %{y:.4f}<br>" |
| "<b>Max (Upper):</b> %{customdata[0]:.4f}<br>" |
| "<b>Min (Lower):</b> %{customdata[1]:.4f}" |
| "<extra></extra>" |
| ), |
| )) |
|
|
| logger.info( |
| "build_main_plot: evaluated %d flavors x %d points in %.2fs", |
| len(parton_ids), len(x_vals), time.monotonic() - t0, |
| ) |
|
|
| default_title = " " |
| fig.update_layout( |
| title=custom_title if custom_title else default_title, |
| xaxis_title="$x$", |
| yaxis_title="$x f(x, Q)$", |
| template="plotly_white", |
| paper_bgcolor="rgba(0,0,0,0)", |
| plot_bgcolor="rgba(0,0,0,0)", |
| autosize=True, |
| legend=dict(x=0.98, y=0.98, xanchor="right", yanchor="top", bgcolor="rgba(255,255,255,0.8)"), |
| margin=dict(l=60, r=40, t=60, b=60), |
| ) |
| fig.update_xaxes( |
| type="log", |
| exponentformat="power", |
| showline=True, |
| linewidth=1.5, |
| linecolor="black", |
| mirror="allticks", |
| ticks="inside", |
| tickwidth=1.5, |
| tickcolor="black", |
| showgrid=False, |
| ) |
|
|
| y_upper_limit = max(all_upper_bands) * 1.15 if all_upper_bands else 1.0 |
| fig.update_yaxes( |
| range=[0, y_upper_limit], |
| showline=True, |
| zeroline=False, |
| linewidth=1.5, |
| linecolor="black", |
| mirror="allticks", |
| ticks="inside", |
| tickwidth=1.5, |
| tickcolor="black", |
| showgrid=False, |
| ) |
| return fig |
|
|
