Spaces:
Sleeping
Sleeping
File size: 6,374 Bytes
a68b00d 979f3c3 a68b00d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 | """Rendu HTML « Throughput effectif » — Sprint 91 (A.II.6).
Suite directe ``picarones/core/throughput.py``. Pattern
identique aux autres rendus : server-side, pas de JS, anti-
injection systématique.
Vue
---
Tableau résumé moteur × {pages/h brut, pages/h **utilisable**,
% temps de correction (drag), n_pages, n_errors}. La cellule
**pages/h utilisable** est colorée en gradient rouge (faible)
→ vert (élevé), normalisé sur le maximum observé.
Adaptive : ``""`` si ``aggregate_effective_throughput`` retourne
``None`` (aucun moteur exploitable).
Note d'intégration
------------------
Cette vue est un **module pur** — l'utilisateur compose :
.. code-block:: python
from picarones.measurements.throughput import (
aggregate_effective_throughput,
)
from picarones.report.throughput_render import (
build_throughput_html,
)
per_engine = []
for report in benchmark.engine_reports:
n_errors = sum(
int(round(dr.metrics.wer * dr.metrics.reference_length / 5))
for dr in report.document_results
)
per_engine.append({
"engine_name": report.engine_name,
"n_pages": len(report.document_results),
"duration_seconds": sum(
dr.duration_seconds for dr in report.document_results
),
"n_errors": n_errors,
})
agg = aggregate_effective_throughput(per_engine)
html = build_throughput_html(agg, labels)
"""
from __future__ import annotations
from html import escape as _e
from typing import Optional
def _color_for_pages_per_hour(value: float, max_value: float) -> str:
"""Rouge (faible) → jaune → vert (= max)."""
if max_value <= 0:
return "#e0e0e0"
f = max(0.0, min(1.0, value / max_value))
if f < 0.5:
t = f / 0.5
r = 235
g = int(70 + (200 - 70) * t)
b = 70
else:
t = (f - 0.5) / 0.5
r = int(235 + (60 - 235) * t)
g = int(200 + (160 - 200) * t)
b = int(70 + (90 - 70) * t)
return f"#{r:02x}{g:02x}{b:02x}"
def _color_for_drag(drag: float) -> str:
"""Vert (faible drag) → orange → rouge (drag élevé)."""
f = max(0.0, min(1.0, drag))
if f < 0.5:
t = f / 0.5
r = int(167 + (235 - 167) * t)
g = int(240 + (180 - 240) * t)
b = int(167 + (60 - 167) * t)
else:
t = (f - 0.5) / 0.5
r = int(235 + (220 - 235) * t)
g = int(180 + (50 - 180) * t)
b = int(60 + (50 - 60) * t)
return f"#{r:02x}{g:02x}{b:02x}"
def build_throughput_html(
aggregated: Optional[dict],
labels: Optional[dict[str, str]] = None,
) -> str:
"""Construit la vue HTML throughput effectif.
Parameters
----------
aggregated:
Sortie de ``aggregate_effective_throughput``. Si
``None`` ou liste vide, retourne ``""``.
labels:
Dict i18n. Clés sous le préfixe ``throughput_*``.
"""
if not aggregated:
return ""
rows = aggregated.get("engines") or []
if not rows:
return ""
labels = labels or {}
title = labels.get("throughput_title", "Throughput effectif")
note = labels.get(
"throughput_note",
"Pages traitables par heure en intégrant le temps de "
"correction humaine post-OCR. Discrimine entre un cloud "
"rapide mais imprécis et un local lent mais fiable. "
"Constante de correction : {time_per_error}s par erreur "
"(défaut HTR-United, surchargeable).",
)
time_per_error = aggregated.get("time_per_error_seconds", 5.0)
note = note.replace("{time_per_error}", f"{time_per_error:.0f}")
h_engine = labels.get("throughput_engine", "Moteur")
h_raw = labels.get("throughput_raw", "Pages/h brut")
h_effective = labels.get(
"throughput_effective", "Pages/h utilisable",
)
h_drag = labels.get("throughput_drag", "% correction")
h_pages = labels.get("throughput_pages", "Pages")
h_errors = labels.get("throughput_errors", "Erreurs")
max_eff = max(
(r.get("pages_per_hour_effective") or 0.0) for r in rows
)
parts = [
'<section class="throughput-section" style="margin:1rem 0">',
f'<h3 style="margin:0 0 .3rem 0">{_e(title)}</h3>',
f'<div style="font-size:.85rem;opacity:.75;margin-bottom:.6rem">'
f'{_e(note)}</div>',
'<table style="border-collapse:collapse;width:100%;'
'font-size:.9rem">',
'<thead><tr>',
]
for col in (h_engine, h_raw, h_effective, h_drag, h_pages, h_errors):
parts.append(
f'<th style="padding:.4rem .6rem;text-align:left;'
f'border-bottom:1px solid #ccc;font-weight:600">'
f'{_e(col)}</th>'
)
parts.append("</tr></thead><tbody>")
for row in sorted(
rows,
key=lambda r: -(r.get("pages_per_hour_effective") or 0.0),
):
engine = str(row.get("engine_name") or "?")
raw = row.get("pages_per_hour_raw")
eff = row.get("pages_per_hour_effective") or 0.0
drag = row.get("drag_ratio") or 0.0
n_pages = int(row.get("n_pages") or 0)
n_errors = int(row.get("n_errors") or 0)
eff_color = _color_for_pages_per_hour(eff, max_eff)
drag_color = _color_for_drag(drag)
raw_str = (
f"{raw:,.0f}" if isinstance(raw, (int, float)) else "—"
)
parts.append(
f'<tr>'
f'<td style="padding:.4rem .6rem">{_e(engine)}</td>'
f'<td style="padding:.4rem .6rem;text-align:right;'
f'font-family:monospace">{raw_str}</td>'
f'<td style="padding:.4rem .6rem;text-align:right;'
f'background:{eff_color};font-family:monospace;'
f'font-weight:600">{eff:,.0f}</td>'
f'<td style="padding:.4rem .6rem;text-align:right;'
f'background:{drag_color};font-family:monospace">'
f'{drag * 100:.1f}%</td>'
f'<td style="padding:.4rem .6rem;text-align:right;'
f'font-family:monospace">{n_pages}</td>'
f'<td style="padding:.4rem .6rem;text-align:right;'
f'font-family:monospace">{n_errors}</td>'
f'</tr>'
)
parts.append("</tbody></table></section>")
return "".join(parts)
__all__ = ["build_throughput_html"]
|