"""WP38 — build literature/references.bib from literature/references_verified.csv (verified entries ONLY).

Entry types follow the Crossref `type` (journal-article -> @article, book-chapter / handbook chapters -> @incollection,
book / monograph / edited-book -> @book, report (NBER) -> @techreport, posted-content -> @unpublished) and, for
URL-verified grey literature, the `type` given in manual_references.yaml (article/book/report).  Titles are escaped
for LaTeX (no bare '&', no '$'); grey literature carries `url` + `note` with the verification date.
"""
from __future__ import annotations
import re, sys, datetime as dt
from pathlib import Path
import pandas as pd

ROOT = Path(__file__).resolve().parents[2]; LIT = ROOT / "literature"

def tex(s):
    if not isinstance(s, str): return ""
    s = s.replace("&", "&").replace("\\&", "&").replace("&", "\\&").replace("$", "\\$").replace("%", "\\%").replace("#", "\\#")
    s = s.replace("‐", "-").replace("—", "---").replace("–", "--")
    return s.strip()

def protect_caps(title):
    """Keep acronyms / proper nouns capitalised under title-casing styles: wrap tokens with inner capitals or all-caps."""
    out = []
    for tok in title.split(" "):
        core = re.sub(r"[^A-Za-z]", "", tok)
        if len(core) > 1 and (core.isupper() or (core[0].isupper() and any(c.isupper() for c in core[1:])) or core in ("Quebec", "Québec", "Canada", "China", "Ireland", "eBay", "MercadoLibre", "Amazon", "Airbnb", "Google", "Pareto", "Balassa", "Internet", "Hollywood")):
            m = re.match(r"^(.*?)([.,:;?!)]*)$", tok)
            tok = "{" + m.group(1) + "}" + m.group(2)
        out.append(tok)
    return " ".join(out)

def num(x):
    if pd.isna(x) or str(x) in ("nan", "None", ""): return None
    s = str(x); return s[:-2] if s.endswith(".0") else s

def authors_of(r):
    a = r.authors if isinstance(r.authors, str) and r.authors.strip() else r.author
    a = str(a)
    if ";" in a: return " and ".join(x.strip() for x in a.split(";") if x.strip())
    return a

def main():
    df = pd.read_csv(LIT / "references_verified.csv"); df = df[df.verified == True].copy()
    out = []
    for _, r in df.sort_values("key").iterrows():
        title = r.crossref_title if isinstance(r.crossref_title, str) and r.crossref_title else r.title
        title = re.sub(r"^Chapter \d+\.\s*", "", title)
        title = protect_caps(tex(title))
        au = authors_of(r); yr = num(r.pub_year) or num(r.year) or "n.d."
        typ = str(r.type) if isinstance(r.type, str) else "article"
        f = [f"  author = {{{tex(au)}}}", f"  title = {{{title}}}", f"  year = {{{yr}}}"]
        doi = r.doi if isinstance(r.doi, str) and r.doi else None
        url = r.url if isinstance(r.url, str) and r.url else None
        cont = tex(r.container) if isinstance(r.container, str) else ""
        if typ in ("journal-article", "article"):
            kind = "article"; f.append(f"  journal = {{{cont}}}")
            if num(r.volume): f.append(f"  volume = {{{num(r.volume)}}}")
            if num(r.issue): f.append(f"  number = {{{num(r.issue)}}}")
            if num(r.pages): f.append(f"  pages = {{{num(r.pages).replace('--', '-').replace('-', '--')}}}")
        elif typ in ("book-chapter",) or (typ in ("book", "edited-book") and cont):
            kind = "incollection"; f.append(f"  booktitle = {{{cont}}}")
            if isinstance(r.publisher, str) and r.publisher: f.append(f"  publisher = {{{tex(r.publisher)}}}")
            if num(r.volume): f.append(f"  volume = {{{num(r.volume)}}}")
            if num(r.pages): f.append(f"  pages = {{{num(r.pages).replace('--', '-').replace('-', '--')}}}")
        elif typ in ("book", "monograph", "edited-book"):
            kind = "book"
            if isinstance(r.publisher, str) and r.publisher: f.append(f"  publisher = {{{tex(r.publisher)}}}")
        elif typ == "report":
            kind = "techreport"
            inst = tex(r.publisher) if isinstance(r.publisher, str) and r.publisher else ("National Bureau of Economic Research" if doi and doi.startswith("10.3386/") else "")
            if inst: f.append(f"  institution = {{{inst}}}")
            if doi and doi.startswith("10.3386/"): f.append(f"  type = {{NBER Working Paper}}"); f.append(f"  number = {{{doi.split('/w')[-1]}}}")
        elif typ == "posted-content":
            kind = "unpublished"; f.append("  note = {Working paper / preprint}")
        else:
            kind = "misc"
        if doi: f.append(f"  doi = {{{doi}}}")
        if r.verification_method == "url:fetch" and url:
            f.append(f"  url = {{{url}}}")
            f.append(f"  note = {{Accessed {r.verified_date}}}")
        elif url and not doi:
            f.append(f"  url = {{{url}}}")
        out.append(f"@{kind}{{{r.key},\n" + ",\n".join(f) + "\n}\n")
    head = (f"% WP38 bibliography — generated {dt.date.today().isoformat()} by src/literature/make_bib.py from literature/references_verified.csv.\n"
            "% Every DOI comes from a Crossref response cached in literature/crossref_cache/; URL entries were fetched on the stated date.\n"
            "% Unverified seeds (literature/unverified.csv) are deliberately absent and must not be cited.\n\n")
    (LIT / "references.bib").write_text(head + "\n".join(out))
    print(f"wrote {len(out)} entries to literature/references.bib")
    # sanity: parse back
    import bibtexparser
    txt = (LIT / "references.bib").read_text()
    if hasattr(bibtexparser, "parse_string"):   # v2
        db = bibtexparser.parse_string(txt); entries = [(e.key, e.fields_dict["title"].value) for e in db.entries]; failed = len(db.failed_blocks)
    else:                                       # v1
        db = bibtexparser.loads(txt); entries = [(e["ID"], e["title"]) for e in db.entries]; failed = 0
    bad = [k for k, t in entries if "$" in t or re.search(r"(?<!\\)&", t)]
    print(f"bibtexparser: {len(entries)} entries parsed, {failed} failed blocks, bad titles: {bad}")

if __name__ == "__main__":
    main()