"""Text OCR pipeline: screenshot -> cleaned-up text -> parsed board info.

Only the text sub-pipeline is implemented. There's no image/icon
recognition sub-pipeline yet (spotting markers, ship icons, etc.), that's
a separate future pipeline, out of scope here.

Preprocessing matters more than the regexes: the typewriter photo has an
uneven vignette (in-game light falloff) that a single global threshold
can't handle, it either loses faint corners or blobs-out dark ones. We
flatten that by dividing by a heavily blurred copy of itself (crude local
background normalization) before thresholding, which recovers text in the
darkened areas reliably.

Parsing is deliberately fuzzy: OCR on an in-game screenshot will misread
the odd character (keywords slightly garbled, '0'/'O', '1'/'I'/'l',
'5'/'S', '8'/'B' confused, stray specks turning ':' into ';' or '.'). We
match keywords by similarity rather than exact spelling, and normalize
digit-shaped letters before parsing numbers, so a couple of misrecognized
characters don't drop an otherwise-good line.
"""

from __future__ import annotations

import difflib
import html
import re
from dataclasses import dataclass, field

import numpy as np
import pytesseract
from PIL import Image, ImageFilter

from . import solver
from .models import Board, Clue, Coord, Location, TargetType
from .shells import Shell

# --- preprocessing -----------------------------------------------------------

_BLUR_RADIUS = 35
_THRESHOLD = 200


def preprocess(image: Image.Image) -> Image.Image:
    gray = image.convert("L")
    arr = np.asarray(gray, dtype=np.float32)
    background = np.asarray(gray.filter(ImageFilter.GaussianBlur(_BLUR_RADIUS)), dtype=np.float32)
    normalized = np.clip(arr / (background + 1e-3) * 255.0, 0, 255).astype(np.uint8)
    normalized_img = Image.fromarray(normalized)
    return normalized_img.point(lambda p: 255 if p > _THRESHOLD else 0)


def ocr_text(image: Image.Image) -> str:
    return pytesseract.image_to_string(preprocess(image), config="--psm 6")


# --- fuzzy keyword matching ---------------------------------------------------

NEST_KEYWORD = "IRON NEST"
SPOTTER_KEYWORD = "SPOTTER"
TARGET_IS_AT_KEYWORD = "Target is at"
_FUZZY_THRESHOLD = 0.65


def _fuzzy_locate(line: str, keyword: str, threshold: float = _FUZZY_THRESHOLD) -> tuple[int, int] | None:
    """Span of the best approximate match of `keyword` in `line`, or None."""
    line_u = line.upper()
    kw = keyword.upper()
    n = len(kw)
    best_ratio = 0.0
    best_span = None
    for size in range(max(n - 2, 1), n + 3):
        for start in range(0, max(len(line_u) - size, 0) + 1):
            end = start + size
            ratio = difflib.SequenceMatcher(None, line_u[start:end], kw).ratio()
            if ratio > best_ratio:
                best_ratio = ratio
                best_span = (start, end)
    return best_span if best_ratio >= threshold else None


def _fuzzy_contains(line: str, keyword: str, threshold: float = _FUZZY_THRESHOLD) -> bool:
    """True if some run of characters in `line` approximately matches `keyword`."""
    return _fuzzy_locate(line, keyword, threshold) is not None


# --- coordinate / id extraction, tolerant of digit<->letter OCR mixups --------

_DIGIT_CLASS = r"[0-9OoIiLlSsBbZzGg]"
_DIGIT_FIX = str.maketrans({
    "O": "0", "o": "0",
    "I": "1", "i": "1", "L": "1", "l": "1",
    "S": "5", "s": "5",
    "B": "8", "b": "8",
    "Z": "2", "z": "2",
    "G": "6", "g": "6",
})
# The full-precision coordinate shape ('H3 5:5', 'Q4 4:2'), one big-grid
# letter, one big-grid number, then the sub-grid x:y pair. Every coord
# regex in this module that captures a full coordinate (not just the
# large-cell letter+number, see _LARGE_GRID_ONLY_RE) embeds this exact
# fragment rather than restating it, so its four capture groups are
# always (letter, big_y, x, y) in that order, matching what
# _coord_from_groups() below expects.
_COORD_FRAGMENT = rf"([A-T])\s*({_DIGIT_CLASS}{{1,2}})\s+({_DIGIT_CLASS})\s*[:;.,]\s*({_DIGIT_CLASS})"
_COORD_RE = re.compile(_COORD_FRAGMENT)


def _coord_from_groups(letter: str, y: str, x: str, yy: str) -> Coord | None:
    """Build a Coord from a _COORD_FRAGMENT match's four groups, digit-
    fixing each numeric one, None on an out-of-range result rather than
    raising (OCR garbage can still parse as *a* number, just not a valid
    grid position)."""
    try:
        return Coord(X=letter.upper(), Y=int(_fix_digits(y)), x=int(_fix_digits(x)), y=int(_fix_digits(yy)))
    except ValueError:
        return None
# No literal '#' required, it's just as OCR-corruptible as anything else
# (missing entirely, or misread as e.g. 'H'). We instead anchor to *where*
# the fuzzy keyword match ended and take the first run of digit-shaped
# characters after that, skipping over whatever separator survived.
_ID_RE = re.compile(rf"({_DIGIT_CLASS}+)")

# Not seen in a real screenshot yet, so no extraction for these, add a
# keyword + extractor here (plus a field below and a case in parse_text)
# once we know the format:
#   - reference points
#   - targets


def _fix_digits(s: str) -> str:
    return s.translate(_DIGIT_FIX)


def _fix_id_digits(raw: str) -> str:
    """Like _fix_digits, but for id-length runs specifically: also
    collapses a 2-character run where one char is a genuine digit and the
    other a look-alike letter that maps to the *same* digit, e.g. '1l' or
    'S5' both fix to '11'/'55', but a real 2-digit id wouldn't plausibly
    render as one numeral plus one letter of the identical value; that
    shape is the signature of OCR ghosting a single thin glyph twice
    (seen repeatedly: 'AmmoCache#l1' -> #1, 'HostileTank#S5' -> #5).
    A genuine two-digit id (both chars already real digits, e.g. '11')
    is left alone."""
    fixed = _fix_digits(raw)
    if len(raw) == 2 and len(set(fixed)) == 1:
        has_digit = any(c.isdigit() for c in raw)
        has_letter = any(c.isalpha() for c in raw)
        if has_digit and has_letter:
            return fixed[0]
    return fixed


# A target can also be spotted with an absolute grid ref directly
# ("Target#10 Spotted. Grid Q3 9:0") instead of/alongside bearing/distance
# clues, same coordinate shape as _COORD_RE, just anchored after "Grid".
_GRID_COORD_RE = re.compile(rf"Grid\s+{_COORD_FRAGMENT}", re.IGNORECASE)


def _extract_grid_coord(text: str) -> Coord | None:
    m = _GRID_COORD_RE.search(text)
    if not m:
        return None
    return _coord_from_groups(*m.groups())


# A fire-support request gives its coord differently again: "<Shell>
# Shells requested on <coord>", no "Grid" keyword. Same shape otherwise.
_REQUESTED_ON_COORD_RE = re.compile(rf"requested\s+on\s+{_COORD_FRAGMENT}", re.IGNORECASE)
# "SMK Shells requested on ..." names the shell by its short code, one of
# the Shell enum's own member names.
_SHELL_REQUEST_RE = re.compile(r"([A-Za-z]+)\s+Shells?\s+requested", re.IGNORECASE)
# "Requested before - T10:31:41 -": an in-game clock deadline, kept as
# the raw string, this app doesn't track a game clock to compare it
# against.
_REQUESTED_BEFORE_RE = re.compile(r"Requested\s+before\s*-\s*(T?\d{1,2}:\d{2}:\d{2})\s*-", re.IGNORECASE)


def _extract_requested_on_coord(text: str) -> Coord | None:
    m = _REQUESTED_ON_COORD_RE.search(text)
    if not m:
        return None
    return _coord_from_groups(*m.groups())


def _extract_shell_request(text: str) -> Shell | None:
    m = _SHELL_REQUEST_RE.search(text)
    if not m:
        return None
    try:
        return Shell[m.group(1).upper()]
    except KeyError:
        return None


def _extract_requested_time(text: str) -> str | None:
    m = _REQUESTED_BEFORE_RE.search(text)
    return m.group(1) if m else None


# A second, unrelated fire-support-request grammar, seen from Infantry
# under attack ("taking fire") rather than a pinned Marine Garrison:
#   Infantry#1 taking fire from id1! Requesting SMK Shell on our
#       position at J6 2:7 before 10:38:57!
#   Infantry#3 taking fire! Requesting HE Shell at bearing 239°,
#       distance 10.76km from our position, J6 2:5, by 10:38:18 or we
#       will be overrun!
# Differs from the Marine Garrison shape in every particular: the shell
# word order is reversed ("Requesting X Shell", not "X Shells requested"),
# the deadline has no "Requested"/dashes, just a bare "before"/"by <time>",
# and the target position is either given directly ("on our position at
# <coord>") or as a bearing/distance offset from that same inline
# position (never a *named* reference -- "our position" isn't a board
# entity, so this resolves the offset directly rather than going through
# a Clue).
# \s* (not \s+) between the shell code and 'Shell(s)': a rich-text paste's
# '<b>SMK Shell</b>' span gets squashed into one no-space token 'SMKShell'
# by squash_bold_spans() before this ever runs (same as any other
# multi-word bold span, see _squash_span_content), same reasoning as
# _extract_requesting_shell()'s docstring.
_REQUESTING_SHELL_RE = re.compile(r"Requesting\s+([A-Za-z]+?)\s*Shells?\b", re.IGNORECASE)
_TAKING_FIRE_TIME_RE = re.compile(r"\b(?:before|by)\s+(T?\d{1,2}:\d{2}:\d{2})\b", re.IGNORECASE)
_ON_OUR_POSITION_COORD_RE = re.compile(
    rf"on\s+our\s+position\s+at\s+{_COORD_FRAGMENT}", re.IGNORECASE
)
_BEARING_DISTANCE_FROM_POSITION_RE = re.compile(
    rf"bearing\s*({_DIGIT_CLASS}{{1,3}})\s*°?\s*,?\s*distance\s*([\d.]+)\s*k?m\s+from\s+our\s+"
    rf"position,?\s*{_COORD_FRAGMENT}",
    re.IGNORECASE,
)


def _extract_requesting_shell(text: str) -> Shell | None:
    """'Requesting SMK Shell' -- the taking-fire grammar's shell mention,
    word order reversed from _extract_shell_request()'s Marine Garrison
    one ('SMK Shells requested')."""
    m = _REQUESTING_SHELL_RE.search(text)
    if not m:
        return None
    try:
        return Shell[m.group(1).upper()]
    except KeyError:
        return None


def _extract_taking_fire_time(text: str) -> str | None:
    m = _TAKING_FIRE_TIME_RE.search(text)
    return m.group(1) if m else None


def _extract_on_our_position_coord(text: str) -> Coord | None:
    m = _ON_OUR_POSITION_COORD_RE.search(text)
    if not m:
        return None
    return _coord_from_groups(*m.groups())


def _extract_bearing_distance_from_position_coord(text: str) -> Coord | None:
    """The bearing/distance variant of a taking-fire request: the shell is
    wanted somewhere OFF the reporting unit's own position, given as a
    bearing/distance from it, with that position itself given inline
    right there ('...from our position, J6 2:5, by ...'). "our position"
    isn't a named board entity to hang a Clue off of, so this resolves
    the offset directly via the same polar-projection math solve_location()
    uses for an ordinary single bearing+distance clue."""
    m = _BEARING_DISTANCE_FROM_POSITION_RE.search(text)
    if not m:
        return None
    bearing, distance, letter, y, x, yy = m.groups()
    origin = _coord_from_groups(letter, y, x, yy)
    if origin is None:
        return None
    point = solver.point_from_bearing_distance(
        origin.as_fraction(), float(_fix_digits(bearing)), float(distance))
    return solver.point_to_coord(point)


def _extract_our_position_coord(text: str) -> Coord | None:
    """The bearing/distance variant's OWN inline position ('...from our
    position, J6 2:5, by ...'), as opposed to
    _extract_bearing_distance_from_position_coord's computed offset from
    it. Reported unit and requested fire point are two different places
    for this variant (unlike the direct "on our position at <coord>" one,
    a real danger-close call), so parse_intel_blocks's flush() uses this
    for the reporting unit's own entry and the offset for a second,
    separate Strike entry -- see its comment."""
    m = _BEARING_DISTANCE_FROM_POSITION_RE.search(text)
    if not m:
        return None
    _bearing, _distance, letter, y, x, yy = m.groups()
    return _coord_from_groups(letter, y, x, yy)


# "Reported active in grid D10": only the large-grid cell, no sub-grid
# x:y at all, unlike every other coord shape in this file. Tried last
# (after _extract_grid_coord, which requires the full x:y and so is
# strictly more precise when both would otherwise match). No sub-
# position is given, so it defaults to the cell's rough middle (5:5,
# there's no exact center on a 0-9 grid) rather than leaving it unset.
_LARGE_GRID_ONLY_RE = re.compile(rf"grid\s+([A-T])\s*({_DIGIT_CLASS}{{1,2}})\b", re.IGNORECASE)


def _extract_large_grid_only_coord(text: str) -> Coord | None:
    m = _LARGE_GRID_ONLY_RE.search(text)
    if not m:
        return None
    letter, y = m.groups()
    try:
        return Coord(X=letter.upper(), Y=int(_fix_digits(y)), x=5, y=5)
    except ValueError:
        return None


def _extract_coord(line: str) -> Coord | None:
    m = _COORD_RE.search(line)
    if not m:
        return None
    return _coord_from_groups(*m.groups())


def _extract_leading_id(remainder: str) -> int | None:
    """First digit-shaped run in `remainder` (text after the keyword match)."""
    m = _ID_RE.search(remainder)
    if not m:
        return None
    try:
        return int(_fix_id_digits(m.group(1)))
    except ValueError:
        return None


# --- "field intelligence" blocks: relative Bearing/Distance descriptions ------
#
#   Target#5  Spotted. 088, 12.10km from Spotter#1
#   Reference Point Alpha:
#       Bearing 094 from Spotter#1
#       Distance 13.26km from Spotter#2
#   .
#   AmmoCache#3:
#       Bearing 217 & Distance 10.48km from AmmoCache#2
#
# Each named entity is a block of one or more clue lines, terminated by a
# blank line or a lone '.'. Degree signs, colons, and the 'km' unit are all
# treated as optional/lossy, OCR drops them unpredictably.

_RP_HEADER_RE = re.compile(r"Reference\s+Point\s+([A-Za-z][\w-]*)\s*:?", re.IGNORECASE)
# '#' isn't required literally, same reasoning as the spotter-id fix: OCR
# drops it or renders it as noise (seen: '€'). Up to 2 junk characters
# between the type word and its digits is enough slack without risking a
# false match elsewhere.
#   ^ the junk-class run is REQUIRED (1-2 chars, not 0-2): the digit class
#   deliberately overlaps the alphabet (g/s/i/l/o/... look like digits), so
#   with a 0-width separator allowed, "Bearing" backtracks into itself,
#   word="Bearin", "digit"=its own trailing 'g', and falsely matches as a
#   header. Requiring real punctuation between word and digits (true of
#   every observed header: '#', a misread substitute, ...) rules that out,
#   and also stops a bare "Word 094" clue line (space only) from matching.
#
# This '<word><junk><digits>' shape ('AmmoCache#3', 'HostileTank#3') is
# what a "Type#id" reference/header looks like anywhere in this module,
# _NAMED_HEADER_RE, _REF_NAMED_RE, and _DESTROYED_RE below all embed this
# same fragment rather than restating it.
_TYPE_ID_FRAGMENT = rf"([A-Za-z]+)[^A-Za-z0-9\s]{{1,2}}({_DIGIT_CLASS}+)"
_NAMED_HEADER_RE = re.compile(rf"^{_TYPE_ID_FRAGMENT}\s*:?\s*(.*)$")

# Last-resort header: a name with no keyword and no id at all, just
# '<Name>:' after squashing ('HMS Rockingham:' -> 'HMSRockingham:'). The
# colon is REQUIRED here (unlike every other header regex's optional
# one), nothing else anchors this match, so an optional colon would
# false-positive on an ordinary clue-continuation line's leading word.
_BARE_NAME_HEADER_RE = re.compile(r"^([A-Za-z][A-Za-z0-9]*)\s*:\s*(.*)$")

# Ad-hoc enemy installations are named in plain English rather than given a
# Type#N id ("Enemy Signal Station:", "Bearing 034 from Enemy Signal
# Station"), which breaks two assumptions everywhere else in this module:
# headers/references are always a single whitespace-free token, and an id
# is always digit-shaped. squash_enemy_names() fixes both up front by
# collapsing 'Enemy' + however many Title Case words follow it into one
# token in our own 'Type#id' shape ('Enemy#SignalStation'), before
# anything else tries to parse the text, so every from-<ref> clue pattern
# (which just captures \S+) and _NAMED_HEADER_RE both work unmodified
# except NAMED_HEADER_RE's id group is digit-only, so a second header
# regex below handles the now-squashed, letter-only id.
_ENEMY_NAME_RE = re.compile(r"\bEnemy(?:[ \t]+[A-Z][a-zA-Z]*){1,4}\b")
_ENEMY_HEADER_RE = re.compile(r"^(Enemy)#([A-Za-z]+)\s*:?\s*(.*)$", re.IGNORECASE)


def squash_enemy_names(text: str) -> str:
    """'Enemy Signal Station' -> 'Enemy#SignalStation', anywhere it
    appears, header or reference alike. The word-separator inside a name
    is deliberately [ \\t]+, not \\s+: it must not cross a newline, or a
    header right at the end of a line ('...Enemy Signal Station') would
    swallow the next line's 'Bearing ...' clue into the same 'name'."""
    return _ENEMY_NAME_RE.sub(lambda m: "Enemy#" + "".join(m.group(0).split()[1:]), text)


# Some named things have a multi-word type ("Coastal Battery#2",
# "Listening Post#1") but, unlike the Enemy case above, already come with
# a real digit id attached, nothing to invent. Only the embedded space
# needs collapsing so _NAMED_HEADER_RE and every from-<ref> clue pattern
# see the single token they expect ('CoastalBattery#2'). Requires 2+
# words specifically, a single-word id like 'AmmoCache#3' already works
# and shouldn't be touched here. Each word must be Title Case, same
# reasoning as _ENEMY_NAME_RE: without it, ordinary lowercase prose right
# before some unrelated '#N' ("...South-East from Listening Post#1" ->
# matching backward from "Post#1" through "from") gets swallowed into the
# 'name' too, an all-lowercase connector word is never actually part of
# one of these names.
_MULTIWORD_ID_RE = re.compile(rf"\b([A-Z][a-zA-Z]*(?:[ \t]+[A-Z][a-zA-Z]*){{1,3}})\s*#\s*({_DIGIT_CLASS}+)")


def squash_multiword_ids(text: str) -> str:
    """'Coastal Battery#2' -> 'CoastalBattery#2', run before
    squash_enemy_names() (order doesn't actually matter, the two patterns
    can't overlap: this one requires a literal '#'+digits, the Enemy one
    requires no digits at all)."""
    return _MULTIWORD_ID_RE.sub(lambda m: "".join(m.group(1).split()) + "#" + m.group(2), text)


# A named anchor can also be given inline, all on one line, rather than as
# its own block: "Listening Post#1 at K6 7:8 audio reports on:". RP-shaped
# same as everything else that resolves to a name + coord, searched for
# anywhere rather than tied to a section keyword, this shape is generic
# enough (not specific to "Listening Post") to catch whatever else turns
# up named this way.
_NAMED_AT_COORD_RE = re.compile(rf"([A-Za-z]+)#({_DIGIT_CLASS}+)\s+at\s+{_COORD_FRAGMENT}", re.IGNORECASE)


def parse_named_at_coord(text: str) -> list[dict]:
    """Parse every '<Name>#<id> at <Coord>' anchor into an RP-shaped
    entry. Call after squash_multiword_ids(), so a multi-word name here
    is already a single 'Name#id' token by the time this regex sees it."""
    entries = []
    for m in _NAMED_AT_COORD_RE.finditer(text):
        word, num, letter, y, x, yy = m.groups()
        coord = _coord_from_groups(letter, y, x, yy)
        if coord is None:
            continue
        entries.append({
            "kind": "rp", "name": f"{word}#{_fix_id_digits(num)}", "type_word": None, "id": None,
            "raw": m.group(0).strip(), "clues": [], "coord": coord,
        })
    return entries


# Forward-observer reports: several standalone one-liners, each an
# independent FO position (a literal one-off coordinate right there in
# the line, not a name referencing some other known entity) giving one
# bearing/distance reading on a named target:
#   FO#5 Audio report on HMS Rockingham: 2.24km From I8 6:9 . . .
#   FO#4 Eyes on HMS Rockingham: 087° From G7 6:7 . . .
#   FO Eyes on HMS Rockingham: 099° From C9 1:2 . . .
# ('FO' alone, no '#N', for exactly one report is fine, still its own
# distinct position.) Distance tried before bearing, same reasoning as
# _CLUE_PATTERNS elsewhere: an ambiguous run of digits could otherwise
# let the bearing pattern's optional '°' swallow part of a distance
# reading if bearing ran first.
_FO_TARGET_NAME = r"([A-Za-z][A-Za-z0-9'\s]*?)"
_FO_DISTANCE_RE = re.compile(
    rf"FO(?:#({_DIGIT_CLASS}+))?\s+(?:Audio\s+report\s+on|Eyes\s+on)\s+{_FO_TARGET_NAME}\s*:\s*"
    rf"([\d.]+)\s*k?m\s+From\s+{_COORD_FRAGMENT}",
    re.IGNORECASE,
)
_FO_BEARING_RE = re.compile(
    rf"FO(?:#({_DIGIT_CLASS}+))?\s+(?:Audio\s+report\s+on|Eyes\s+on)\s+{_FO_TARGET_NAME}\s*:\s*"
    rf"({_DIGIT_CLASS}{{1,3}})\s*°?\s+From\s+{_COORD_FRAGMENT}",
    re.IGNORECASE,
)


def parse_fo_reports(text: str) -> dict[str, tuple[str, Coord | None]]:
    """An FO position isn't tracked as a real board entity, nothing to
    re-reference later, no ongoing observation post, just a one-off
    fix, so this triangulates immediately: for each named target, build
    a throwaway scratch Board with just that target's FO positions as
    (never-persisted) reference points, run the exact same
    solve_location() geometry/priority the real board uses, keep only
    the resulting Coord (or None, if the readings under- or
    over-determine it in a way solve_location() can't resolve), then
    discard the scratch board and every ephemeral FO entity in it.
    Returns target_name -> (raw report text, resolved coord or None)."""
    consumed: list[tuple[int, int]] = []

    def overlaps(span: tuple[int, int]) -> bool:
        return any(span[0] < e and span[1] > s for s, e in consumed)

    # target_key -> list of (fo_name, fo_coord, bearing_deg, distance_km, raw_line)
    groups: dict[str, list[tuple[str, Coord, float | None, float | None, str]]] = {}

    def collect(m, is_distance: bool) -> None:
        span = m.span()
        if overlaps(span):
            return
        consumed.append(span)
        fo_num, target_name, value, letter, y, x, yy = m.groups()
        fo_name = f"FO#{_fix_id_digits(fo_num)}" if fo_num else "FO"
        coord = _coord_from_groups(letter, y, x, yy)
        if coord is None:
            return
        target_key = "".join(re.sub(r"[^A-Za-z0-9]", "", w) for w in target_name.split())
        bearing = None if is_distance else float(_fix_digits(value))
        distance = float(_fix_digits(value)) if is_distance else None
        groups.setdefault(target_key, []).append((fo_name, coord, bearing, distance, m.group(0).strip()))

    for m in _FO_DISTANCE_RE.finditer(text):
        collect(m, is_distance=True)
    for m in _FO_BEARING_RE.finditer(text):
        collect(m, is_distance=False)

    results: dict[str, tuple[str, Coord | None]] = {}
    for target_key, readings in groups.items():
        scratch = Board()
        clues = []
        placed: set[str] = set()
        for fo_name, coord, bearing, distance, _raw in readings:
            if fo_name not in placed:
                placed.add(fo_name)
                scratch.add_reference_point(coord, name=fo_name)
            clues.append(Clue(reference=fo_name, bearing_deg=bearing, distance_km=distance))
        raw_text = "\n".join(r[4] for r in readings)
        result = solver.solve_location(Location.from_desc(raw_text, clues), scratch)
        results[target_key] = (raw_text, result.coord)
    return results


_BLOCK_END_RE = re.compile(r"^[.\s]{1,6}$")
_LEADING_NOISE_RE = re.compile(r"^[^A-Za-z]{1,3}(?=[A-Za-z])")

# Reference capture is (\S+), not (.+): references are always a single
# token with no spaces, and being non-greedy this way is what lets
# finditer() find more than one clue per line/block, "Bearing 118 from
# Spotter#2 & Bearing 125 from Spotter#1" needs two separate matches, and
# a greedy (.+) would let the first one swallow the rest of the string.
#
# _GAP sits right before "from": plain whitespace normally, but an OCR
# line-wrap can drop a stray junk token right at the break ("Bearing 125°"
# / "P; from Spotter#1"), and that junk can itself contain a letter (the
# 'P' above), so this isn't just non-alnum noise like the header-bullet
# case; tolerate any single short token, not just punctuation.
_GAP = r"[\s]*(?:\S{1,3}\s*)?"
_CLUE_COMBINED_RE = re.compile(
    rf"Bearing\s*(\d{{1,3}})\s*°?\s*&\s*Distance\s*([\d.]+)\s*k?m?{_GAP}from\s+(\S+)", re.IGNORECASE
)
_CLUE_INLINE_RE = re.compile(
    rf"(\d{{1,3}})\s*°?\s*,\s*([\d.]+)\s*k?m{_GAP}from\s+(\S+)", re.IGNORECASE
)
_CLUE_BEARING_RE = re.compile(rf"Bearing\s*(\d{{1,3}})\s*°?{_GAP}from\s+(\S+)", re.IGNORECASE)

# A listening post gives distance readings by ear, an approximate 8-point
# compass direction instead of a precise degree bearing: "Distance 6.28km
# South-East from Listening Post#1". Word order differs from every other
# clue shape too (the direction sits between the distance and 'from',
# there's no separate 'Bearing' keyword at all), so this needs its own
# pattern rather than reusing _CLUE_COMBINED_RE with a looser bearing group.
_COMPASS_BEARINGS = {
    "NORTH": 0.0, "NORTHEAST": 45.0, "EAST": 90.0, "SOUTHEAST": 135.0,
    "SOUTH": 180.0, "SOUTHWEST": 225.0, "WEST": 270.0, "NORTHWEST": 315.0,
    # 16-point compass, only ever seen written with a space rather than a
    # hyphen ('North Northwest', not 'North-Northwest'), unlike the
    # 8-point diagonals above.
    "NORTHNORTHEAST": 22.5, "EASTNORTHEAST": 67.5, "EASTSOUTHEAST": 112.5, "SOUTHSOUTHEAST": 157.5,
    "SOUTHSOUTHWEST": 202.5, "WESTSOUTHWEST": 247.5, "WESTNORTHWEST": 292.5, "NORTHNORTHWEST": 337.5,
}
# Longest/most specific alternatives first: alternation tries each in
# order and stops at the first that matches, so 'North Northwest' must
# reach the 16-point alternative before the bare 'North' one, or the
# latter would win and strand ' Northwest' unmatched.
_COMPASS_WORD_RE = (
    r"(North[-\s]?North[-\s]?East|North[-\s]?East|East[-\s]?North[-\s]?East|"
    r"South[-\s]?South[-\s]?East|South[-\s]?East|East[-\s]?South[-\s]?East|"
    r"South[-\s]?South[-\s]?West|South[-\s]?West|West[-\s]?South[-\s]?West|"
    r"North[-\s]?North[-\s]?West|North[-\s]?West|West[-\s]?North[-\s]?West|"
    r"North|South|East|West)"
)
# Half the width of one 16-point compass sector (360/16 = 22.5 deg each),
# a compass WORD names a whole sector, not a single ray.
_COMPASS_TOLERANCE_DEG = 11.25


def _compass_to_bearing(word: str) -> float:
    return _COMPASS_BEARINGS[re.sub(r"[-\s]", "", word).upper()]


_CLUE_DISTANCE_COMPASS_RE = re.compile(
    rf"Distance\s*([\d.]+)\s*k?m?\s*{_COMPASS_WORD_RE}{_GAP}from\s+(\S+)", re.IGNORECASE
)
_CLUE_DISTANCE_RE = re.compile(rf"Distance\s*([\d.]+)\s*k?m?{_GAP}from\s+(\S+)", re.IGNORECASE)

# A different clue grammar entirely, reference FIRST then a colon then
# the bare reading, no 'Bearing'/'Distance' keyword and no 'from' at all:
#   Spotter#2: 4.04km
#   Spotter#3: 298°
#   Spotter#1: West
# Distance tried before bearing/compass, same reasoning as _CLUE_PATTERNS
# generally: an ambiguous run of digits could otherwise let the bearing
# pattern's optional '°' swallow part of a distance reading.
_CLUE_REF_DISTANCE_RE = re.compile(rf"(\S+)\s*:\s*([\d.]+)\s*k?m", re.IGNORECASE)
_CLUE_REF_BEARING_RE = re.compile(rf"(\S+)\s*:\s*({_DIGIT_CLASS}{{1,3}})\s*°", re.IGNORECASE)
_CLUE_REF_COMPASS_RE = re.compile(rf"(\S+)\s*:\s*{_COMPASS_WORD_RE}\b", re.IGNORECASE)
# 'Spotter#2: 4.04km' has the exact same 'Word#digits:' shape as a real
# block header ('AmmoCache#3:'), genuinely indistinguishable from one by
# shape alone. The deciding signal is what follows the colon: a real
# header's trailing content is never JUST a bare clue reading with
# nothing else, so _NAMED_HEADER_RE's own inline-content group gets
# checked against this before deciding it's actually a new header.
_BARE_CLUE_VALUE_RE = re.compile(
    rf"^(?:[\d.]+\s*k?m|{_DIGIT_CLASS}{{1,3}}\s*°|{_COMPASS_WORD_RE})\s*$", re.IGNORECASE
)

_TYPE_BY_SHORT = {t.short: t for t in TargetType}
# The game's typewriter has used "AmmoCache" for what's now modeled as
# SupplyCache, and "CoastalBattery" for what's just a HostileArtillery
# under a different name, treat both as the same type rather than
# dropping the target or inventing a redundant enum member for it.
_TYPE_WORD_ALIASES = {"AmmoCache": "SupplyCache", "CoastalBattery": "HostileArtillery"}


_REF_NAMED_RE = re.compile(rf"^{_TYPE_ID_FRAGMENT}")


def _clean_reference(raw: str) -> str:
    """Leading name-shaped token, dropping trailing OCR noise (stray dots,
    double spaces, etc.), reference names never contain spaces. Named
    references ('Spotter#1', 'AmmoCache#2') get their digit part fixed up
    and their separator normalized to '#'; plain word references ('Alpha')
    are left untouched, don't run digit-fixing over them or real letters
    like the 'l' in 'Alpha' get corrupted into '1'. 'Enemy#SignalStation'
    (squash_enemy_names() already normalized the separator, and its id is
    letters, not digits) is passed through as-is rather than falling into
    the digit-fixing path below, which would otherwise stop at the first
    letter outside _DIGIT_CLASS and truncate it (e.g. down to 'Enemy#Sig')."""
    token = re.match(r"\S+", raw.strip())
    token = token.group(0) if token else raw.strip()

    enemy = _ENEMY_HEADER_RE.match(token)
    if enemy:
        return f"{enemy.group(1)}#{enemy.group(2)}"

    named = _REF_NAMED_RE.match(token)
    if named:
        word, num = named.groups()
        word = _TYPE_WORD_ALIASES.get(word, word)
        return f"{word}#{_fix_id_digits(num)}"

    return re.sub(r"[^\w-]+$", "", token)  # trim trailing punctuation off a plain name


# Priority order matters: try the "both bearing and distance" shape before
# the single-value shapes, so e.g. "Bearing 217 & Distance 10.48km from X"
# becomes one clue, not a spurious extra bearing-only one from the same
# text. `\s` already matches a literal newline, so this bridges an OCR
# line-wrap ("...Bearing 125°" / "from Spotter#1" split across two lines)
# without needing the caller to rejoin anything. Extract functions return
# (bearing_deg, distance_km, reference, bearing_tolerance_deg), the last
# one None except for the compass-word shapes.
_CLUE_PATTERNS = (
    (_CLUE_COMBINED_RE, lambda m: (float(m.group(1)), float(m.group(2)), m.group(3), None)),
    (_CLUE_INLINE_RE, lambda m: (float(m.group(1)), float(m.group(2)), m.group(3), None)),
    (_CLUE_BEARING_RE, lambda m: (float(m.group(1)), None, m.group(2), None)),
    (_CLUE_DISTANCE_COMPASS_RE, lambda m: (
        _compass_to_bearing(m.group(2)), float(m.group(1)), m.group(3), None
    )),
    (_CLUE_DISTANCE_RE, lambda m: (None, float(m.group(1)), m.group(2), None)),
    # reference-first shape: '<ref>: <value>', no keyword, no 'from'
    (_CLUE_REF_DISTANCE_RE, lambda m: (None, float(m.group(2)), m.group(1), None)),
    (_CLUE_REF_BEARING_RE, lambda m: (float(_fix_digits(m.group(2))), None, m.group(1), None)),
    (_CLUE_REF_COMPASS_RE, lambda m: (
        _compass_to_bearing(m.group(2)), None, m.group(1), _COMPASS_TOLERANCE_DEG
    )),
)


def _parse_all_clues(text: str) -> list[Clue]:
    """Every Bearing/Distance clue found anywhere in `text`, a block can
    have several (one per clue line, or more than one on a single line
    joined with '&')."""
    clues: list[Clue] = []
    consumed: list[tuple[int, int]] = []

    def overlaps(span: tuple[int, int]) -> bool:
        return any(span[0] < e and span[1] > s for s, e in consumed)

    for pattern, extract in _CLUE_PATTERNS:
        for m in pattern.finditer(text):
            span = m.span()
            if overlaps(span):
                continue
            consumed.append(span)
            bearing, distance, ref, tolerance = extract(m)
            clues.append(Clue(reference=_clean_reference(ref), bearing_deg=bearing, distance_km=distance,
                               bearing_tolerance_deg=tolerance))
    return clues


def parse_clues_from_text(text: str) -> list[Clue]:
    """Parse every Bearing/Distance clue found in free-form text, used
    for manually-typed descriptions in the coord dialog, sharing the
    exact same clue grammar as the OCR'd intel blocks."""
    text = strip_html(squash_bold_spans(text))
    return _parse_all_clues(squash_enemy_names(squash_multiword_ids(text)))


_ALLY_PREFIX_RE = re.compile(r"^(Friendly|Hostile)", re.IGNORECASE)


def _resolve_target_type(type_word: str) -> tuple[TargetType | None, bool]:
    """(TargetType, is_ally). A leading 'Friendly'/'Hostile' word is
    stripped off the type word first ('FriendlyTank' -> ally, TANK;
    'HostileTank' or bare 'Tank' -> not ally, TANK, an explicit
    'Hostile' and no prefix at all mean the same thing, not-ally is the
    default). What's left is matched exactly against the type word
    (after aliasing), falling back to fuzzy (OCR can garble the type
    word itself, e.g. 'AmmoCoche')."""
    is_ally = False
    prefix_m = _ALLY_PREFIX_RE.match(type_word)
    if prefix_m:
        is_ally = prefix_m.group(1).lower() == "friendly"
        type_word = type_word[prefix_m.end():]

    type_word = _TYPE_WORD_ALIASES.get(type_word, type_word)
    if type_word in _TYPE_BY_SHORT:
        return _TYPE_BY_SHORT[type_word], is_ally
    best, best_ratio = None, 0.0
    for short, target_type in _TYPE_BY_SHORT.items():
        ratio = difflib.SequenceMatcher(None, type_word.upper(), short.upper()).ratio()
        if ratio > best_ratio:
            best, best_ratio = target_type, ratio
    return (best if best_ratio >= _FUZZY_THRESHOLD else None), is_ally


def parse_intel_blocks(text: str) -> list[dict]:
    """Parse 'field intelligence' blocks into a list of dicts with keys
    kind ('rp' | 'named'), name, type_word, id, raw, clues, coord, shell,
    requested_time. Entries with none of clue/coord/shell/requested_time
    are dropped (nothing to store).

    Clues are extracted once per block, from the whole joined block text,
    at flush time, not accumulated line-by-line while scanning. That's
    what lets a clue split across an OCR line-wrap ("...Bearing 125°" /
    "from Spotter#1" on separate lines) or two clues on one line
    ("Bearing X from A & Bearing Y from B") both resolve correctly. A
    target can also carry an absolute grid ref directly ("Grid Q3 9:0")
    instead of/alongside clues, or a fire-support request's own coord
    shape ("SMK Shells requested on J6 8:3"), plus that request's shell
    and deadline if given."""
    entries: list[dict] = []
    current: dict | None = None

    def flush():
        nonlocal current
        if current is not None:
            joined = "\n".join(current["raw"])
            current["clues"] = _parse_all_clues(joined)
            # The bearing/distance taking-fire variant names TWO different
            # places (see _extract_our_position_coord's docstring): the
            # reporting unit's own position, and a separate fire point
            # offset from it. Everything else in this module is "one block
            # -> one entry", so that offset gets split into a second,
            # synthetic StrikeRequest entry below rather than folded into
            # this one -- otherwise the fire point either overwrites the
            # unit's real position (wrong place) or gets silently dropped.
            offset_coord = _extract_bearing_distance_from_position_coord(joined)
            current["coord"] = (
                _extract_grid_coord(joined) or _extract_requested_on_coord(joined)
                or _extract_on_our_position_coord(joined)
                or _extract_our_position_coord(joined)
                or _extract_large_grid_only_coord(joined)
            )
            shell = _extract_shell_request(joined) or _extract_requesting_shell(joined)
            requested_time = _extract_requested_time(joined) or _extract_taking_fire_time(joined)
            # The offset variant's shell/deadline describe the FIRE POINT,
            # not the reporting unit itself -- they move to the synthetic
            # StrikeRequest entry below, not kept here too.
            current["shell"] = None if offset_coord is not None else shell
            current["requested_time"] = None if offset_coord is not None else requested_time
            if (current["clues"] or current["coord"] is not None
                    or current["shell"] is not None or current["requested_time"] is not None):
                current["raw"] = joined
                entries.append(current)
                if offset_coord is not None:
                    # TargetType.STRIKE_REQUEST, not STRIKE: this is a
                    # friendly unit calling in a strike over the radio, not
                    # one the player placed themselves (see that type's own
                    # comment in models.py). type_word must match its
                    # TargetType.short exactly ("StrikeRequest", no space),
                    # same as every other type_word this module produces.
                    strike_id = f"{current['type_word']}{current['id']}"
                    entries.append({
                        "kind": "named", "name": f"StrikeRequest#{strike_id}",
                        "type_word": "StrikeRequest",
                        "id": strike_id, "raw": joined, "clues": [], "coord": offset_coord,
                        "shell": shell, "requested_time": requested_time,
                    })
        current = None

    for raw_line in text.splitlines():
        line = raw_line.strip()
        if not line:
            continue  # blank lines are just visual spacing here, not a section boundary
        if _BLOCK_END_RE.match(line):
            flush()
            continue

        # The '.' block-separator bullet often survives OCR as 1-3 stray
        # junk characters glued onto the *next* line ('i AmmoCache#1:',
        # '* AmmoCache#2:') instead of its own line, which would otherwise
        # defeat the column-0-anchored header regexes below. Try the line
        # as-is first, and only if that fails, retry with its first
        # whitespace-delimited token stripped (covers symbol junk *and*
        # a misread bullet that happened to OCR as a stray letter), but
        # only when that leading token is bullet-length (<=3 chars), else
        # a genuine wrapped clue continuation like "from Spotter#2" gets
        # its "from" stripped and "Spotter#2" misread as a new header.
        first_token_m = re.match(r"\S+", line)
        strip_first_token = (
            re.sub(r"^\S+\s+", "", line, count=1)
            if first_token_m and len(first_token_m.group(0)) <= 3
            else line
        )
        candidates = (line, _LEADING_NOISE_RE.sub("", line), strip_first_token)

        rp_m = next((m for c in candidates if (m := _RP_HEADER_RE.search(c))), None)
        if rp_m:
            flush()
            current = {"kind": "rp", "name": rp_m.group(1), "type_word": None, "id": None,
                       "raw": [line], "clues": []}
            continue

        # Tried before _NAMED_HEADER_RE: its digit class overlaps plain
        # letters (S/B/Z/G/O/I/L), so on a squashed 'Enemy#SignalStation'
        # header it would otherwise partial-match ('Sig' -> digit-fixed
        # into the bogus id '516') before this more specific check ever
        # gets a look.
        enemy_m = next((m for c in candidates if (m := _ENEMY_HEADER_RE.match(c))), None)
        if enemy_m:
            flush()
            type_word, letter_id, inline = enemy_m.groups()
            current = {"kind": "named", "name": f"{type_word}#{letter_id}", "type_word": type_word,
                       "id": letter_id, "raw": [line], "clues": []}
            continue

        named_m = next((m for c in candidates if (m := _NAMED_HEADER_RE.match(c))), None)
        if named_m:
            type_word, num, inline = named_m.groups()
            if _BARE_CLUE_VALUE_RE.match(inline.strip()):
                # Not actually a header, a '<ref>: <value>' clue line
                # for whatever block is already open (see
                # _BARE_CLUE_VALUE_RE), append rather than start a new
                # block over it.
                if current is not None:
                    current["raw"].append(line)
                continue
            flush()
            num = _fix_id_digits(num)
            current = {"kind": "named", "name": f"{type_word}#{num}", "type_word": type_word,
                       "id": num, "raw": [line], "clues": []}
            continue

        # Last resort: a bare name with none of the above (no 'Reference
        # Point'/'Enemy' keyword, no digit id), just '<Name>:' -
        # ('HMS Rockingham:'). Requiring the colon (not optional, unlike
        # the other header regexes) matters here specifically: without
        # any keyword or id anchoring the match, an optional colon would
        # false-positive on an ordinary clue-continuation line's first
        # word ('Distance 6.14km...' -> 'Distance' read as a header).
        # Treated as a Target (TargetType.UNKNOWN, whose short form IS
        # 'Target'), not a Reference Point, a named thing giving its own
        # clues is being spotted, not a fixed landmark spotters aim off
        # of, same reasoning as the "Target is at-" calibration line.
        bare_m = next((m for c in candidates if (m := _BARE_NAME_HEADER_RE.match(c))), None)
        if bare_m:
            flush()
            name = bare_m.group(1)
            current = {"kind": "named", "name": f"Target#{name}", "type_word": "Target",
                       "id": name, "raw": [line], "clues": []}
            continue

        if current is not None:
            current["raw"].append(line)

    flush()
    return entries


# Train-arrival intel is a different shape entirely, a station's absolute
# grid ref, the rail's bearing from it, and a handful of waypoints given
# only as a distance along that same bearing, plus a T=HH:MM:SS timestamp
# on every line that this app makes no attempt to use, there's no game
# clock to compare it against, so it's simply ignored:
#   ARRIVAL STATION:
#       Valle de Mula MainStation: J6 0:4
#   Estimated arrival: T=10:16:50
#   TRACK ALIGNMENT:
#       Rail line runs straight. Bearing 090° from MainStation.
#   FINAL APPROACH:
#       Waypoint A - 6.00km from station: T=10:06:50
#       Waypoint B - 4.00km from station: T=10:10:10
# The station and each waypoint all come out RP-shaped (kind='rp'), same
# as parse_intel_blocks()'s entries, so app.py can merge them the exact
# same way, no separate code path needed on that side.
_ARRIVAL_STATION_KEYWORD = "ARRIVAL STATION"
_TRACK_ALIGNMENT_KEYWORD = "TRACK ALIGNMENT"
_FINAL_APPROACH_KEYWORD = "FINAL APPROACH"

_STATION_LINE_RE = re.compile(rf"(\S+):\s*{_COORD_FRAGMENT}")
_WAYPOINT_RE = re.compile(r"Waypoint\s+(\S+)\s*-\s*([\d.]+)\s*k?m\s+from\s+(\S+)", re.IGNORECASE)


def _section(text: str, start_keyword: str, end_keywords: tuple[str, ...]) -> str | None:
    """Text between a fuzzy match of `start_keyword` and whichever comes
    first: a block-end '.' line, one of `end_keywords`, or the end of the
    text. None if `start_keyword` isn't found at all. Scopes each
    train-intel sub-parser to just its own section, so e.g. a
    coincidental ':' + coord shape elsewhere in the text can't be
    mistaken for the arrival station line."""
    lines = text.splitlines()
    start = next((i + 1 for i, line in enumerate(lines) if _fuzzy_contains(line.strip(), start_keyword)), None)
    if start is None:
        return None
    collected = []
    for line in lines[start:]:
        stripped = line.strip()
        if _BLOCK_END_RE.match(stripped) or any(_fuzzy_contains(stripped, k) for k in end_keywords):
            break
        collected.append(line)
    return "\n".join(collected)


def parse_train_intel(text: str) -> list[dict]:
    """Parse a train-arrival intel block into RP-shaped entries: the
    named station itself (absolute grid ref) plus one entry per waypoint,
    each resolved via bearing (the rail's alignment) + distance (that
    waypoint's distance along it) from the station, solve_location()'s
    simplest case, one clue with both. Returns [] if this doesn't look
    like a train intel block at all (no ARRIVAL STATION section)."""
    station_block = _section(text, _ARRIVAL_STATION_KEYWORD, (_TRACK_ALIGNMENT_KEYWORD, _FINAL_APPROACH_KEYWORD))
    if station_block is None:
        return []
    station_m = _STATION_LINE_RE.search(station_block)
    if station_m is None:
        return []
    station_name, letter, y, x, yy = station_m.groups()
    station_coord = _coord_from_groups(letter, y, x, yy)
    if station_coord is None:
        return []

    entries = [{
        "kind": "rp", "name": station_name, "type_word": None, "id": None,
        "raw": station_block.strip(), "clues": [], "coord": station_coord,
    }]

    track_block = _section(text, _TRACK_ALIGNMENT_KEYWORD, (_FINAL_APPROACH_KEYWORD,))
    track_clues = _parse_all_clues(track_block) if track_block else []
    bearing = next((c.bearing_deg for c in track_clues if c.bearing_deg is not None), None)
    if bearing is None:
        return entries  # the station alone is still useful even without the rail's bearing

    approach_block = _section(text, _FINAL_APPROACH_KEYWORD, ())
    if approach_block is None:
        return entries

    for m in _WAYPOINT_RE.finditer(approach_block):
        letter_id, distance, ref_word = m.groups()
        ref = station_name if ref_word.strip(".:").lower() == "station" else _clean_reference(ref_word)
        entries.append({
            "kind": "rp", "name": f"Waypoint {letter_id}", "type_word": None, "id": None,
            "raw": m.group(0).strip(),
            "clues": [Clue(reference=ref, bearing_deg=bearing, distance_km=float(_fix_digits(distance)))],
            "coord": None,
        })
    return entries


# Destruction reports are standalone one-liners, not tied to a block:
#   "SupplyCache#2 Destroyed. Additional Requisition Granted."
#   "Direct Hit! HostileTank#3 Destroyed."
# Just need "<Type>#<id>" immediately followed by "Destroyed", the
# "Direct Hit!" prefix (or its absence) doesn't matter, search() finds
# the name+Destroyed pair anywhere in the line either way.
_DESTROYED_RE = re.compile(rf"{_TYPE_ID_FRAGMENT}\s*Destroyed", re.IGNORECASE)
# Same digit-vs-letter-id split as the header regexes: an ad-hoc "Enemy
# X Y" installation's id (after squash_enemy_names()) is letters, not
# digits, so it needs its own pattern, _DESTROYED_RE's digit class won't
# match it at all ("Enemy#SignalStation Destroyed" was silently dropped).
_ENEMY_DESTROYED_RE = re.compile(r"(Enemy)#([A-Za-z]+)\s*Destroyed", re.IGNORECASE)


def parse_destroyed(text: str) -> set[tuple[TargetType, str]]:
    destroyed = set()
    for m in _DESTROYED_RE.finditer(text):
        type_word, num = m.groups()
        target_type, _is_ally = _resolve_target_type(type_word)
        if target_type is None:
            continue
        destroyed.add((target_type, _fix_id_digits(num)))
    for m in _ENEMY_DESTROYED_RE.finditer(text):
        type_word, letter_id = m.groups()
        target_type, _is_ally = _resolve_target_type(type_word)
        if target_type is None:
            continue
        destroyed.add((target_type, letter_id))
    return destroyed


_BOLD_SPAN_RE = re.compile(r"<b>(.*?)</b>", re.IGNORECASE | re.DOTALL)
# Same two shapes squash_enemy_names()/squash_multiword_ids() look for,
# but anchored to fullmatch just one bold span's own content rather than
# searched for across the whole document. That's strictly safer: those
# two functions have to *guess* where a multi-word name ends using
# capitalization (and got that wrong once already, swallowing 'from' into
# a name it bordered), where here the closing </b> tag itself is the
# actual boundary, nothing to guess.
_ENEMY_SPAN_RE = re.compile(r"Enemy(?:\s+[A-Za-z]+){1,4}$", re.IGNORECASE)
_MULTIWORD_ID_SPAN_RE = re.compile(r"([A-Za-z]+(?:\s+[A-Za-z]+){1,3})\s*#\s*(\d+)$")


def _squash_span_content(content: str) -> str:
    if _ENEMY_SPAN_RE.match(content):
        words = content.split()
        return "Enemy#" + "".join(words[1:])
    multiword_id = _MULTIWORD_ID_SPAN_RE.match(content)
    if multiword_id:
        return "".join(multiword_id.group(1).split()) + "#" + multiword_id.group(2)
    words = content.split()
    if len(words) > 1 and not any(re.search(r"\d", w) for w in words):
        # Neither shape above, but still a multi-word span with no digits
        # in it, a plain multi-word name ('The Mole', "Dockmaster's
        # House"): still one logical name, collapse it into one
        # alphanumeric token the same way every reference elsewhere is
        # matched (_RP_HEADER_RE, the (\S+) clue-reference capture),
        # stripping punctuation too (apostrophes etc.), those aren't in
        # _RP_HEADER_RE's word-character class and would otherwise
        # truncate the match. The no-digits guard matters: a multi-word
        # bold span can just as easily be a coordinate ('C9 7:9'), and
        # squashing THAT the same way corrupts it instead of a name.
        return "".join(re.sub(r"[^A-Za-z0-9]", "", w) for w in words)
    return content


def squash_bold_spans(text: str) -> str:
    """Rich-text pastes wrap each atomic value/name in its own
    '<b>...</b>' ('<b>Enemy Signal Station</b>', '<b>Listening
    Post#1</b>'), collapse a multi-word one down to our single-token
    'Type#id' shape using that tag boundary as ground truth, before the
    tags themselves get stripped. Must run before strip_html(), it needs
    the tags still there to know a span's extent. A no-op on plain OCR
    text, which never has '<b>' in it to begin with, squash_enemy_names()/
    squash_multiword_ids() (run afterward regardless) are what handle
    that case, guessing from capitalization since there's no markup left
    to lean on."""
    return _BOLD_SPAN_RE.sub(lambda m: "<b>" + _squash_span_content(m.group(1)) + "</b>", text)


_HTML_BREAK_RE = re.compile(r"(?i)<br\s*/?>|</(?:p|div|li|tr)>")
_HTML_TAG_RE = re.compile(r"<[^>]+>")


def strip_html(text: str) -> str:
    """Drop markup from rich-text clipboard pastes (e.g. '<b>145</b>°'),
    leaving plain text the rest of the pipeline can parse. Block-ish
    closing tags become newlines first so pastes that rely on markup
    rather than real line breaks don't get run together; harmless no-op
    on plain OCR text with no '<' in it."""
    text = _HTML_BREAK_RE.sub("\n", text)
    text = _HTML_TAG_RE.sub("", text)
    return html.unescape(text)


@dataclass
class ParsedInfo:
    nest_coord: Coord | None = None
    spotters: dict[int, Coord] = field(default_factory=dict)
    # name -> (raw description, clues, absolute coord if given directly)
    reference_points: dict[str, tuple[str, list[Clue], Coord | None]] = field(default_factory=dict)
    # (type, id) -> (raw description, clues, absolute coord if given
    # directly, requested shell if a fire-support request named one,
    # requested-before deadline string if given)
    targets: dict[
        tuple[TargetType, str], tuple[str, list[Clue], Coord | None, Shell | None, str | None]
    ] = field(default_factory=dict)
    # Same shape as targets, for a 'Friendly'-prefixed type word (see
    # _resolve_target_type()), a separate collection entirely, not a
    # flag on a target, an ally's id doesn't share a namespace with a
    # same-typed hostile target and it's never fired on.
    allies: dict[
        tuple[TargetType, str], tuple[str, list[Clue], Coord | None]
    ] = field(default_factory=dict)
    # (type, id) of targets reported destroyed
    destroyed: set[tuple[TargetType, str]] = field(default_factory=set)

    def is_empty(self) -> bool:
        return not (self.nest_coord or self.spotters or self.reference_points
                    or self.targets or self.allies or self.destroyed)


def parse_text(text: str) -> ParsedInfo:
    text = strip_html(squash_bold_spans(text))
    text = squash_enemy_names(squash_multiword_ids(text))
    info = ParsedInfo()

    for raw_line in text.splitlines():
        line = raw_line.strip()
        if not line:
            continue

        if info.nest_coord is None and _fuzzy_contains(line, NEST_KEYWORD):
            coord = _extract_coord(line)
            if coord is not None:
                info.nest_coord = coord
                continue

        spotter_span = _fuzzy_locate(line, SPOTTER_KEYWORD)
        if spotter_span is not None:
            spotter_id = _extract_leading_id(line[spotter_span[1]:])
            coord = _extract_coord(line)
            if spotter_id is not None and coord is not None:
                info.spotters[spotter_id] = coord

        # The game's opening calibration order gives the target as plain
        # prose ("Target is at- Q4 4:2") rather than the usual "Target#N:"
        # block shape, there's only ever one of these, so it's stored under
        # a fixed id ("1") rather than one parsed from the text.
        if (TargetType.UNKNOWN, "1") not in info.targets and _fuzzy_contains(line, TARGET_IS_AT_KEYWORD):
            coord = _extract_coord(line)
            if coord is not None:
                info.targets[(TargetType.UNKNOWN, "1")] = (line, [], coord, None, None)
                continue

    for entry in parse_intel_blocks(text):
        if entry["kind"] == "rp":
            info.reference_points[entry["name"]] = (entry["raw"], entry["clues"], entry["coord"])
            continue
        target_type, is_ally = _resolve_target_type(entry["type_word"])
        if target_type is None:
            continue
        if is_ally:
            info.allies[(target_type, entry["id"])] = (entry["raw"], entry["clues"], entry["coord"])
        else:
            info.targets[(target_type, entry["id"])] = (
                entry["raw"], entry["clues"], entry["coord"], entry["shell"], entry["requested_time"]
            )

    for entry in parse_train_intel(text):
        info.reference_points[entry["name"]] = (entry["raw"], entry["clues"], entry["coord"])

    for entry in parse_named_at_coord(text):
        info.reference_points[entry["name"]] = (entry["raw"], entry["clues"], entry["coord"])

    for target_key, (raw, coord) in parse_fo_reports(text).items():
        key = (TargetType.UNKNOWN, target_key)
        if key in info.targets:
            old_raw, old_clues, old_coord, shell, requested_time = info.targets[key]
            info.targets[key] = (
                f"{old_raw}\n{raw}", old_clues, old_coord if old_coord is not None else coord,
                shell, requested_time,
            )
        else:
            info.targets[key] = (raw, [], coord, None, None)

    info.destroyed = parse_destroyed(text)
    return info


def run(image: Image.Image) -> ParsedInfo:
    return parse_text(ocr_text(image))