"""T1. Profile screen: where does Voynichese fall among 103 languages (plain and
vowel-less) on 14 statistics? Outputs percentile ranks per statistic, a nearest-
language ranking on z-scored distance, and the set of statistics on which the
manuscript lies outside the attested range (decision rule: outside 2.5–97.5 pct).
"""
import os, sys
_here = os.path.dirname(os.path.abspath(__file__))  # data/analysis, or a subfolder of it
sys.path.insert(0, _here if os.path.exists(os.path.join(_here, "paths.py")) else os.path.dirname(_here))
from paths import *  # noqa: repository locations
import json, csv, math, statistics as st
import numpy as np
from harness import *

C = corpus(); pages = structured()
V = flatten(pages)
pv = profile(V); pvu = profile(V, units=True)
pA = profile(flatten(structured(lang="A"))); pB = profile(flatten(structured(lang="B")))
langs = languages()
rows = {}
for name, ws in langs.items():
    rows[name] = profile(ws[:len(V)])
keys = list(pv.keys())
plain = {k: v for k, v in rows.items() if not k.endswith(".novowel")}
nov = {k: v for k, v in rows.items() if k.endswith(".novowel")}

out = []
for k in keys:
    vals = np.array([r[k] for r in plain.values()]); nv = np.array([r[k] for r in nov.values()])
    pct = (vals < pv[k]).mean() * 100; pctn = (nv < pv[k]).mean() * 100
    lo, hi = np.percentile(vals, [2.5, 97.5])
    out.append(dict(statistic=k, voynich=pv[k], voynich_units=pvu[k], voynich_A=pA[k], voynich_B=pB[k],
                    lang_min=vals.min(), lang_p2_5=lo, lang_median=np.median(vals), lang_p97_5=hi, lang_max=vals.max(),
                    pct_rank_plain=pct, pct_rank_novowel=pctn, outside_range=(pv[k] < lo or pv[k] > hi)))
with open("out/profile_screen.csv", "w", newline="") as f:
    w = csv.DictWriter(f, fieldnames=out[0].keys()); w.writeheader(); w.writerows(out)
for r in out:
    print(f"{r['statistic']:14s} V={r['voynich']:.3f} (units {r['voynich_units']:.3f})  langs [{r['lang_min']:.3f} .. {r['lang_max']:.3f}] median {r['lang_median']:.3f}  pct {r['pct_rank_plain']:.0f}  {'OUTSIDE' if r['outside_range'] else ''}")

# nearest languages on z-scored distance (plain + novowel pooled), two feature sets:
# (a) all 14; (b) character-level only; (c) word-level only
allv = np.array([[r[k] for k in keys] for r in rows.values()]); names = list(rows.keys())
mu, sd = allv.mean(0), allv.std(0) + 1e-9
def rank(feat, target):
    idx = [keys.index(k) for k in feat]
    z = (allv[:, idx] - mu[idx]) / sd[idx]; t = (np.array([target[k] for k in feat]) - mu[idx]) / sd[idx]
    d = np.sqrt(((z - t) ** 2).sum(1))
    return sorted(zip(d, names))
sets = {"all": keys, "character-level": ["alphabet", "h1", "h2", "mean_len", "sd_len", "short_words"],
        "word-level": ["ttr", "hapax", "zipf", "top1", "top10", "word_h2", "adj_repeat"]}
res = {}
for sname, feat in sets.items():
    for tname, target in [("Voynich", pv), ("Voynich (glyph units)", pvu)]:
        r = rank(feat, target); res[f"{sname}|{tname}"] = [(n, round(float(d), 2)) for d, n in r[:12]]
        print(f"\n{sname} / {tname}: " + ", ".join(f"{n} {d:.2f}" for d, n in r[:10]))
json.dump(res, open("out/profile_nearest.json", "w"), indent=1)
with open("out/profile_languages.csv", "w", newline="") as f:
    w = csv.writer(f); w.writerow(["language"] + keys)
    w.writerow(["Voynich"] + [pv[k] for k in keys]); w.writerow(["Voynich-units"] + [pvu[k] for k in keys])
    for n, r in rows.items(): w.writerow([n] + [r[k] for k in keys])
outside = [r["statistic"] for r in out if r["outside_range"]]
log("H-LANG-ANY", "T1-profile", "statistics outside attested range", ";".join(outside), "103 languages ×2 forms", "see note",
    "Voynich lies outside the 95% range of attested languages on these statistics")