#!/usr/bin/env python3
"""Copies analysis/*.md into the site's studies collection, normalising front matter.

The analyses are authored as standalone research documents; some carry rich front matter
(subtitle, abstract, status), some carry none at all. The site's content schema needs a
title, a description and a sort order on every one. Rather than maintain two divergent
copies, this regenerates the collection from analysis/ on every build.

Documents listed in INTERNAL are working notes or build inputs and are not published.
"""
import os
import re
import sys

ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
SRC = os.path.join(ROOT, "analysis")
DST = os.path.join(ROOT, "site", "src", "content", "studies")

# Not published: build inputs and unedited working notes. raw-* are the per-book profiling
# notes the lens reports were built from — they are kept in the repo as evidence trail but
# are not written for a reader.
INTERNAL = {"design-research.md", "raw-profiles-social-studies.md",
            "raw-religion-social-studies.md", "historical-quotations.md"}

# Reading order on /studies. Lower sorts first. Anything unlisted falls to the end
# alphabetically, which is the right default for material added later.
ORDER = {
    "synthesis.md": 1,
    "rubric.md": 2,
    "edition-audit.md": 3,
    "article-assessment-demand.md": 4,
    "psychological-implications.md": 5,
    "thread-mother-tongue.md":
        "Every board names its regional poets. Only Sindh teaches one as a set text — and no "
        "Punjabi poet appears in any of Punjab's five Urdu literary readers.",
    "thread-iconoclasm.md": 6,
    "thread-moral-rescue.md": 7,
    "thread-mother-tongue.md": 8,
    "lens-religion.md": 11, "lens-civics.md": 12, "lens-history.md": 13,
    "lens-patriotism.md": 14, "lens-othering.md": 15, "lens-gender.md": 16,
    "track-pakistan-studies.md": 21, "track-social-studies.md": 22,
    "track-islamiyat.md": 23, "track-urdu.md": 24, "track-english.md": 25,
    "track-sciences.md": 26,
}

# A negative result is a result, and the guideline requires it be published at equal
# prominence. Equal prominence is not the same as identical presentation: a study that
# went looking and found nothing needs to say so before a reader infers a positive.
NEGATIVE = {
    "thread-mother-tongue.md":
        "Every board names its regional poets. Only Sindh teaches one as a set text — and no "
        "Punjabi poet appears in any of Punjab's five Urdu literary readers.",
    "thread-iconoclasm.md":
        "Tested and NOT established. No board reaches the threshold for an explicit link "
        "between Makkah's idols and Muslim conquest in India.",
    "thread-moral-rescue.md":
        "Tested and NOT established for the Indian side. Sati does not appear in any of the "
        "112 books read; the Arabian-side framing is examined, the Indian-side equivalent is not.",
}

DESCRIPTIONS = {
    "synthesis.md":
        "The headline comparison: how the five boards score across seven axes, and why two "
        "defensible composite indices rank them differently.",
    "rubric.md":
        "The seven-axis instrument, fixed before any book was read, with the scoring rules "
        "and the evidence standards every claim on this site must meet.",
    "edition-audit.md":
        "Proof that the books analysed here are the current ones — and the finding that "
        "curriculum vintage, not province, is the sharper divide.",
    "article-assessment-demand.md":
        "A journal-length article arguing that what pupils are required to write in "
        "examinations is a stronger instrument for reading a curriculum than what its "
        "textbooks say.",
    "psychological-implications.md":
        "Twenty-six falsifiable predictions about what these four curricula should produce in the "
        "pupils formed by them — attitudes, and then personality structure — each with the "
        "test that would refute it.",
    "thread-moral-rescue.md":
        "Testing whether pre-Islamic Arabia and pre-Islamic India are constructed as a "
        "shared barbarism from which Muslim rule delivered rescue.",
    "thread-mother-tongue.md":
        "Every board names its regional poets. Only Sindh teaches one as a set text — and no "
        "Punjabi poet appears in any of Punjab's five Urdu literary readers.",
    "thread-iconoclasm.md":
        "Does any board link the smashing of idols at Makkah to Muslim conquest in India? "
        "A four-tier evidence test, and what each province actually does.",
    "lens-religion.md":
        "How much religious content appears outside Islamiyat — in science, English, "
        "geography and history — and how far each board separates faith from subject matter.",
    "lens-civics.md":
        "What a pupil is taught about rights, democracy, law and the citizen's relationship "
        "to the state — and which province teaches them to judge rather than recall.",
    "lens-history.md":
        "Whose history is told, where each province starts the story, and what is left out "
        "of the account of Partition, 1971 and military rule.",
    "lens-patriotism.md":
        "Patriotism is taught everywhere in roughly equal volume. What differs is the kind: "
        "civic belonging, religious destiny, or military sacrifice.",
    "lens-othering.md":
        "How Hindus, India, the West and religious minorities are portrayed, with the "
        "quoted passages and the counter-examples each board also prints.",
    "lens-gender.md":
        "Women as agents, as subjects, or as absent — counted across the corpus, with the "
        "distinction between what is taught and who is depicted.",
    "track-pakistan-studies.md":
        "The subject that carries the national narrative, compared board by board across "
        "classes 9 and 10.",
    "track-social-studies.md":
        "The boards do not even agree what this subject is: one integrated volume in some "
        "provinces, split History and Geography in others.",
    "track-islamiyat.md":
        "Compulsory religious instruction compared across the boards that publish it, classes 6 to 10.",
    "track-urdu.md":
        "What the national-language reader teaches beyond language: the selection of "
        "authors, themes and moral lessons in each province.",
    "track-english.md":
        "The English reader as a vehicle for values — what its comprehension passages are "
        "about, and how that differs by province.",
    "track-sciences.md":
        "Biology, chemistry, physics and general science: how much religious framing "
        "enters, and how differently the provinces handle evolution and inquiry.",
}


# Cross-references in the analyses are written as file paths, because that is what they are on
# disk. A reader does not have the disk. Rewrite them into links to the published page, but only
# where the target is actually published — an INTERNAL document has no page to link to.
_DOC_TITLES = {
    "synthesis": "the synthesis", "rubric": "the rubric", "edition-audit": "the edition audit",
    "lens-religion": "the religion lens", "lens-civics": "the civics lens",
    "lens-history": "the history lens", "lens-patriotism": "the patriotism lens",
    "lens-othering": "the othering lens", "lens-gender": "the gender lens",
    "track-pakistan-studies": "the Pakistan Studies track",
    "track-social-studies": "the Social Studies track", "track-islamiyat": "the Islamiyat track",
    "track-urdu": "the Urdu track", "track-english": "the English track",
    "track-sciences": "the sciences track",
    "thread-iconoclasm": "the iconoclasm thread", "thread-moral-rescue": "the moral-rescue thread",
    "article-assessment-demand": "What the Pupil Must Write",
    "psychological-implications": "What the Treatment Predicts",
}


def linkify_refs(text):
    """`analysis/lens-civics.md` and bare `lens-civics.md` -> a link to /studies/lens-civics/.
    A trailing section reference (§2.4) is carried into the link text so the pointer survives."""
    def repl(m):
        stem = m.group("stem").lower()
        if stem not in _DOC_TITLES:
            return m.group(0)
        sect = (m.group("sect") or "").strip()
        label = _DOC_TITLES[stem] + (" " + sect if sect else "")
        return "[%s](/studies/%s/)" % (label, stem)

    pattern = re.compile(
        r"`?(?:analysis/)?(?P<stem>[A-Za-z0-9_-]+)\.md`?(?P<sect>\s*§\s*[\d.]+)?")

    out, last = [], 0
    # Never rewrite inside a fenced code block or an existing markdown link target.
    protected = [(m.start(), m.end()) for m in
                 re.finditer(r"```.*?```|\]\([^)]*\)", text, re.S)]
    for m in pattern.finditer(text):
        if any(a <= m.start() < b for a, b in protected):
            continue
        out.append(text[last:m.start()])
        out.append(repl(m))
        last = m.end()
    out.append(text[last:])
    return "".join(out)


def split_front_matter(text):
    """Returns (dict_of_scalar_keys, body). Block scalars (>-, |) are folded to one line."""
    if not text.startswith("---"):
        return {}, text
    end = text.find("\n---", 3)
    if end == -1:
        return {}, text
    raw = text[3:end]
    body = text[end + 4:].lstrip("\n")
    meta, key = {}, None
    for line in raw.splitlines():
        m = re.match(r"^(\w+):\s*(.*)$", line)
        if m:
            key, val = m.group(1), m.group(2).strip()
            meta[key] = "" if val in (">", "|", ">-", "|-") else val.strip('"\'')
        elif key and line.strip():
            meta[key] = (meta[key] + " " + line.strip()).strip()
    return meta, body


def first_sentence(s, limit=240):
    s = " ".join(s.split())
    m = re.search(r"^(.{40,%d}?\.)(\s|$)" % limit, s)
    out = m.group(1) if m else s[:limit].rstrip()
    return out


def yaml_quote(s):
    return '"%s"' % " ".join(s.split()).replace('\\', '\\\\').replace('"', '\\"')


def main():
    os.makedirs(DST, exist_ok=True)
    for f in os.listdir(DST):
        if f.endswith(".md"):
            os.remove(os.path.join(DST, f))

    written = 0
    for name in sorted(os.listdir(SRC)):
        if not name.endswith(".md") or name.lower() in INTERNAL:
            continue
        slug = name.lower()
        text = open(os.path.join(SRC, name), encoding="utf-8").read()
        meta, body = split_front_matter(text)

        title = meta.get("title")
        if not title:
            m = re.search(r"^#\s+(.+)$", body, re.M)
            title = m.group(1).strip() if m else name[:-3].replace("-", " ").capitalize()
        # The H1 duplicates the title the layout already renders. Drop the first one only.
        body = re.sub(r"^#\s+.+\n+", "", body, count=1)

        desc = (DESCRIPTIONS.get(slug) or meta.get("description")
                or meta.get("abstract") or meta.get("subtitle"))
        if not desc:
            para = next((p for p in body.split("\n\n")
                         if p.strip() and not p.lstrip().startswith(("#", ">", "|", "-", "*"))), "")
            desc = re.sub(r"[*_`\[\]]|\(([^)]*)\)", "", para)
        desc = first_sentence(desc)

        # Real revision date, taken from the source document's mtime. A reference work with no
        # date on it cannot be cited properly — the reader cannot say which version they read.
        import datetime as _dt
        revised = _dt.date.fromtimestamp(
            os.path.getmtime(os.path.join(SRC, name))).isoformat()
        fm = ["---", "title: %s" % yaml_quote(title),
              "description: %s" % yaml_quote(desc),
              "revised: %s" % yaml_quote(revised),   # quoted: bare ISO dates parse as YAML dates, not strings
              "order: %d" % ORDER.get(slug, 90)]
        if slug in NEGATIVE:
            fm.append("result: %s" % yaml_quote("negative"))
            fm.append("resultNote: %s" % yaml_quote(NEGATIVE[slug]))
        for k in ("subtitle", "status"):
            if meta.get(k):
                fm.append("%s: %s" % (k, yaml_quote(meta[k])))
        fm.append("---\n")

        body = linkify_refs(body)
        open(os.path.join(DST, slug), "w", encoding="utf-8").write("\n".join(fm) + "\n" + body)
        written += 1
        print("  %-38s order %-3s %s" % (slug, ORDER.get(slug, 90), desc[:60]))

    print("studies synced: %d" % written)
    if written == 0:
        sys.exit(1)


if __name__ == "__main__":
    main()