probes: 13

The exact question asked of a corpus, kept verbatim (code column): a Python regex on one document, a SPARQL query on a public endpoint, a shell request. result_count/result_of is the re-run figure on result_date; matches_excerpt says whether that figure is the one quoted in the observation (0 = differs, both kept).

This data as json

id	form_id	form_name	file	kind	corpus_id	written	run	result_count	result_of	result_date	matches_excerpt	note	code
13	form-043	Tables with multi-level headers	probes/2026-08/line-043.py	python	corpus-europe-pmc-jats-2023-2026	2026-08-08	python3 tools/run_probe.py probes/2026-08/line-043.py <dir with the 272 .xml>	73	272	2026-09-09	1		"""Fault Atlas probe — catalogue line 43 (form-043). Extracted verbatim from the August 2026 observation campaign (Loxyn SAS, Lyon). A probe answers one question on one JATS XML file: "is this fault here, and where?" It repairs nothing and decides nothing. Returns a list of excerpts, empty if absent. Run on the reproducible corpus (see corpora/europe-pmc-jats-2023-2026-272.json): python3 tools/run_probe.py probes/2026-08/line-043.py <dir with the 272 .xml> """ import re, collections, unicodedata SONDES = {} def sonde(ligne, nom): def deco(f): SONDES[ligne] = (nom, f); return f return deco # ── 43 · tableaux à en-têtes multiples ───────────────────────────────────── @sonde(43, "tableau à en-têtes sur plusieurs niveaux") def d43(x): span = len(re.findall(r'<th[^>](?:col\|row)span="([2-9]\|\d\d)"', x)) lignes_th = len(re.findall(r'(?s)<thead\b.?</thead>', x)) multi = 0 for th in re.findall(r'(?s)<thead\b.*?</thead>', x): if len(re.findall(r'<tr\b', th)) > 1: multi += 1 if span or multi: return [f"th colspan/rowspan ×{span}", f"thead à >1 ligne ×{multi}/{lignes_th}"] return []