"""The Inventory Planning header must quote COSMOS's own figures, not ours re-derived.

The weekly cells always matched, because they come straight from the `invp-insight` dateMap.
The HEADER did not: it was recomputed from our sales-insight history, so one SKU read
56 / 62 / 65 here against 64 / 130 / 67 in COSMOS Inventory Planning -- same product, same
screen position, two different numbers.

Fixed by carrying COSMOS's fields through verbatim:

    daily sale   ours 7-day mean 56    -> COSMOS `averageSale` 64
    middle cell  ours 30-day mean 62   -> COSMOS `potentialDailySale` 130 (a DIFFERENT
                                          quantity: a demand estimate, not a trailing mean)
    third cell   ours 90-day mean 65   -> COSMOS `averageSale6Months` 67
    on hand      first projected week 5,029 -> COSMOS `inventory` 3,999
    inbound      every arrival 1,770        -> the IMMEDIATE arrival 1,030
"""
from __future__ import annotations

import pathlib

SRC = pathlib.Path(__file__).resolve().parents[1] / "app.py"
TEXT = SRC.read_text(encoding="utf-8")


def _table_body() -> str:
    return TEXT.split("def inventory_table_html")[1].split("\ndef ")[0]


def test_header_reads_cosmos_fields_not_history_means():
    body = _table_body()
    assert "invp_stats" in body, "header no longer reads COSMOS's own figures"
    assert '_st.get("avg_sale")' in body
    assert '_st.get("potential_daily_sale")' in body
    assert '_st.get("avg_6m")' in body


def test_the_old_history_derived_cells_are_gone():
    body = _table_body()
    assert "{d30:,.0f}" not in body, "30-day history mean is back in the header"
    assert "{d90:,.0f}" not in body, "90-day history mean is back in the header"


def test_on_hand_is_cosmos_inventory_not_the_first_projected_week():
    """5,029 is week one WITH its arrival folded in; 3,999 is the shelf."""
    body = _table_body()
    assert 'get("invp_stats") or {}).get("inventory")' in body
    assert "on hand" in body


def test_inbound_is_the_immediate_arrival_with_later_ones_reported_separately():
    body = _table_body()
    assert 'total_inbound = float(proj[0]["warehouse_arrival"] or 0)' in body
    assert "later_inbound" in body, "later arrivals must still be reported, not dropped"


# ---------------------------------------------------------------- encoding hygiene
# app.py and dashboard/live_data.py were once written back through a cp1252 round trip, so
# every em dash, middot and emoji in the UI rendered as garbage. It showed in the BROWSER, not
# just an editor, which makes it a user-visible defect rather than a tidiness one.
#
# The signatures are built with chr() rather than spelled out, so this file cannot match its
# own check.
_ROOT = pathlib.Path(__file__).resolve().parents[1]
_MOJIBAKE = (
    chr(0xE2) + chr(0x20AC),   # em/en dash, curly quotes, ellipsis
    chr(0xC2) + chr(0xB7),     # middot
    chr(0xF0) + chr(0x178),    # any 4-byte emoji
    chr(0xC3) + chr(0xA2),     # doubly encoded
)
_SKIP = {".venv", "__pycache__", "new_archtetcure_and frontend", ".mojibake_backup"}


def _py_sources():
    for f in _ROOT.rglob("*.py"):
        if not _SKIP.intersection(f.parts):
            yield f


def test_no_source_file_is_mojibake():
    bad = {}
    for f in _py_sources():
        text = f.read_bytes().decode("utf-8-sig")
        hits = sum(text.count(m) for m in _MOJIBAKE)
        if hits:
            bad[str(f.relative_to(_ROOT))] = hits
    assert not bad, f"mojibake in: {bad}"


def test_no_source_file_has_a_bom():
    """A BOM breaks `ast.parse` on a utf-8 read and shows as a stray char in some editors."""
    bom = bytes([0xEF, 0xBB, 0xBF])
    bad = [str(f.relative_to(_ROOT)) for f in _py_sources() if f.read_bytes()[:3] == bom]
    assert not bad, f"UTF-8 BOM in: {bad}"


def test_every_source_file_is_valid_utf8():
    for f in _py_sources():
        f.read_bytes().decode("utf-8")