#!/usr/bin/env python3
"""Create author notes from paper-note frontmatter and keep their paper
lists fresh.

One note per unique author name in papers/influences frontmatter. Each note
carries a GENERATED paper-links section between markers — real wikilinks, so
they register in Obsidian's graph and render outside Obsidian (export,
GitHub/HF views), unlike dataview blocks. The generated section is refreshed
on every run; anything the user writes outside the markers is preserved.

    python3 _scripts/make_authors.py

Also reports author notes that no paper references anymore (possible
spelling drift) — it never deletes them.
"""
import glob
import os
import re
import unicodedata

VAULT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
AUTHORS_DIR = os.path.join(VAULT, "authors")

STUB = """---
tags: [author]
---

# {name}

<!-- Enrich freely above/below; the section between markers is regenerated. -->

{papers_section}"""

MARK_A = "<!-- GENERATED:papers -->"
MARK_B = "<!-- /GENERATED:papers -->"


def papers_section(entries):
    lines = "\n".join(f"- [[{slug}]] ({year}, {venue})" if venue else f"- [[{slug}]] ({year})"
                       for slug, year, venue in entries) or "- (none currently)"
    return f"{MARK_A}\n## Papers in this vault\n\n{lines}\n{MARK_B}\n"



def main() -> None:
    os.makedirs(AUTHORS_DIR, exist_ok=True)
    by_author = {}
    for path in glob.glob(os.path.join(VAULT, "papers", "*.md")) + glob.glob(os.path.join(VAULT, "influences", "*.md")):
        txt = open(path, encoding="utf-8").read()
        fm = txt.split("---")[1]
        m = re.search(r"^authors: \[(.*)\]$", fm, flags=re.M)
        if not m:
            continue
        slug = os.path.splitext(os.path.basename(path))[0]
        ym = re.search(r"^year:\s*(\d+)", fm, flags=re.M)
        vm = re.search(r'^venue:\s*"?(.*?)"?\s*$', fm, flags=re.M)
        year = ym.group(1) if ym else "?"
        venue = (vm.group(1) if vm else "")[:40]
        for name in re.findall(r"\[\[(.+?)\]\]", m.group(1)):
            by_author.setdefault(name, []).append((slug, year, venue))
    names = set(by_author)

    created, refreshed = 0, 0
    for name in sorted(names):
        entries = sorted(by_author[name], key=lambda e: (e[1], e[0]))  # year, then slug: stable across machines
        section = papers_section(entries)
        out = os.path.join(AUTHORS_DIR, f"{name}.md")
        if not os.path.exists(out):
            with open(out, "w", encoding="utf-8") as f:
                f.write(STUB.format(name=name, papers_section=section))
            created += 1
            continue
        txt = open(out, encoding="utf-8").read()
        if MARK_A in txt and MARK_B in txt:
            new = re.sub(re.escape(MARK_A) + r".*?" + re.escape(MARK_B) + r"\n?",
                         section, txt, flags=re.S)
        else:
            # legacy stub: replace the dataview papers section wholesale
            new = re.sub(r"## Papers in this vault\n+```dataview.*?```\n?",
                         section, txt, flags=re.S)
            if new == txt:  # enriched/unknown layout: append the generated section
                new = txt.rstrip() + "\n\n" + section
        if new != txt:
            open(out, "w", encoding="utf-8").write(new)
            refreshed += 1

    # macOS may list filenames decomposed (NFD) while notes spell names composed
    # (NFC): compare in NFC so accented authors are not reported as stale.
    nfc = lambda s: unicodedata.normalize("NFC", s)
    existing = {nfc(os.path.splitext(f)[0]) for f in os.listdir(AUTHORS_DIR) if f.endswith(".md")}
    stale = existing - {nfc(n) for n in names}
    print(f"authors referenced: {len(names)}, notes created: {created}, link-sections refreshed: {refreshed}")
    if stale:
        print(f"notes no paper references (check spelling drift, not deleted): {sorted(stale)}")


if __name__ == "__main__":
    main()