mirror of
https://github.com/harvard-edge/cs249r_book.git
synced 2026-07-16 06:07:17 -05:00
efde57313e
Three-phase standalone audit tooling under book/tools/audit/: - emphasis_extract.py: emit body-prose bold/italic spans, skipping code, math, callouts, footnotes, headings, tables, captions (READ-ONLY). - emphasis_ap2map.py: volume-wide bold occurrence map; flags terms bolded in >=2 chapters (AP2 duplicate first-definition candidates), canonical chapter order from the PDF config. - emphasis_rollup.py: roll per-chapter agent ledgers under .claude/_reviews/emphasis/ into a volume summary + actionable fix-list. Ad-hoc audit tooling for now; candidate to fold into Binder as a 'check emphasis' group once it becomes a commit gate.
72 lines
3.0 KiB
Python
72 lines
3.0 KiB
Python
#!/usr/bin/env python3
|
|
"""
|
|
Volume-wide body-prose BOLD occurrence map -> AP2 candidates.
|
|
|
|
Phase 2 helper for the emphasis audit. Runs emphasis_extract.py over every chapter
|
|
of a volume IN CANONICAL ORDER (derived from the PDF config YAML), collects every
|
|
body-prose bold span, and reports any term bolded in >=2 distinct chapters -- the
|
|
only way an AP2 duplicate first-definition can occur. Earlier chapter = owner.
|
|
|
|
Usage: python3 book/tools/audit/emphasis_ap2map.py vol1
|
|
python3 book/tools/audit/emphasis_ap2map.py vol2 --watch inference,arithmetic\\ intensity
|
|
|
|
READ-ONLY. Flags candidates; ruling (P1 owner vs P3-ext/P4 vs real AP2) is a human read.
|
|
"""
|
|
import sys, re, json, subprocess, os
|
|
from collections import defaultdict
|
|
|
|
HERE = os.path.dirname(os.path.abspath(__file__))
|
|
ROOT = os.path.abspath(os.path.join(HERE, '..', '..', '..')) # repo root
|
|
VOL = sys.argv[1] if len(sys.argv) > 1 else 'vol1'
|
|
WATCH = []
|
|
if '--watch' in sys.argv:
|
|
WATCH = [w.strip().lower() for w in sys.argv[sys.argv.index('--watch') + 1].split(',')]
|
|
|
|
def chapter_order(vol):
|
|
"""Canonical chapter order from the volume's PDF config (handles commented lines)."""
|
|
cfg = os.path.join(ROOT, 'book', 'quarto', 'config', f'_quarto-pdf-{vol}.yml')
|
|
pat = re.compile(rf'contents/{vol}/([a-z0-9_]+)/\1\.qmd') # dir==stem -> a chapter (skips parts/, index)
|
|
order, seen = [], set()
|
|
for line in open(cfg, encoding='utf-8'):
|
|
m = pat.search(line)
|
|
if m and m.group(1) not in seen:
|
|
seen.add(m.group(1)); order.append(m.group(1))
|
|
return order
|
|
|
|
def norm(span):
|
|
t = span.strip().strip('*').strip()
|
|
t = re.sub(r'\s+', ' ', t).lower()
|
|
t = re.sub(r'^(the|a|an) ', '', t)
|
|
return t.strip(' .,:;()')
|
|
|
|
order = chapter_order(VOL)
|
|
if not order:
|
|
sys.exit(f"No chapters found for {VOL} in PDF config.")
|
|
|
|
occ = defaultdict(list) # norm_term -> [(chap_idx, chap, line)]
|
|
extractor = os.path.join(HERE, 'emphasis_extract.py')
|
|
for ci, ch in enumerate(order):
|
|
path = os.path.join(ROOT, 'book', 'quarto', 'contents', VOL, ch, f'{ch}.qmd')
|
|
out = subprocess.run(['python3', extractor, path], capture_output=True, text=True).stdout
|
|
for ln in out.splitlines():
|
|
s = json.loads(ln)
|
|
if s['type'] == 'bold':
|
|
occ[norm(s['span'])].append((ci, ch, s['line']))
|
|
|
|
multi = {n: v for n, v in occ.items() if len({x[0] for x in v}) >= 2}
|
|
print(f"{VOL}: {len(order)} chapters, {len(occ)} distinct body-bold terms, "
|
|
f"{len(multi)} bolded in >=2 chapters (AP2 candidates)\n")
|
|
for n in sorted(multi, key=lambda k: (multi[k][0][0], k)):
|
|
rows = sorted(multi[n], key=lambda x: (x[0], x[2]))
|
|
print(f"• {rows[0][2]:>6} {rows[0][1]:<22} {n!r}")
|
|
for _, ch, ln in rows[1:]:
|
|
print(f" {ln:>6} {ch:<22} (later -> check vs owner)")
|
|
|
|
if WATCH:
|
|
print("\n=== WATCHLIST ===")
|
|
for w in WATCH:
|
|
hits = occ.get(w, [])
|
|
chs = ', '.join(f"{c}:L{l}" for _, c, l in hits)
|
|
flag = " <-- MULTI-CHAPTER" if len({h[0] for h in hits}) >= 2 else ""
|
|
print(f" {w:28} {len(hits)}x {chs or '(not body-bolded)'}{flag}")
|