"""Pronunciation fixes for the simulator's speech.

The built-in OS voices are clear and instant, but they mispronounce brand names
and initialisms - "AGIBOT" is the obvious one. Rather than abandoning a voice
that otherwise sounds good, we respell the problem words phonetically just
before they are spoken.

The text the *robot* receives is never touched. This rewrite happens at the very
last step, inside the PC audio layer, so:

  * the dashboard, history and API all keep the operator's original wording;
  * the real AGIBOT A3 always gets the real text - it has its own TTS and its
    own idea of how its name sounds.

Rules live in `pronunciation.json` at the project root so they can be corrected
without touching code:

    {
      "AGIBOT": "Ah-jee-bot",
      "A3":     "A three"
    }

Matching is case-insensitive and respects word boundaries, so "AGIBOT" will not
corrupt a longer word that happens to contain it.
"""

from __future__ import annotations

import json
import logging
import re
from pathlib import Path
from typing import Dict, List, Optional, Tuple

logger = logging.getLogger(__name__)

RULES_FILE = "pronunciation.json"

#: Shipped defaults. Deliberately small - guessing at pronunciations nobody
#: asked for is worse than leaving a word alone. Extend via pronunciation.json.
DEFAULT_RULES: Dict[str, str] = {
    "AGIBOT": "Ah-jee-bot",
    "AgiBot": "Ah-jee-bot",
}


class Pronouncer:
    """Applies respelling rules to text just before synthesis."""

    def __init__(self, rules: Optional[Dict[str, str]] = None) -> None:
        self._compiled: List[Tuple[re.Pattern, str]] = []
        self.rules: Dict[str, str] = {}
        self.load(rules if rules is not None else DEFAULT_RULES)

    def load(self, rules: Dict[str, str]) -> None:
        self.rules = dict(rules or {})
        self._compiled = []
        # Longest first, so a specific phrase wins over a word inside it.
        for phrase in sorted(self.rules, key=len, reverse=True):
            replacement = self.rules[phrase]
            if not phrase:
                continue
            # \b does not anchor against digits/symbols the way we need for
            # names like "A3", so use explicit look-around on word characters.
            pattern = re.compile(
                r"(?<![0-9A-Za-z]){0}(?![0-9A-Za-z])".format(re.escape(phrase)),
                re.IGNORECASE,
            )
            self._compiled.append((pattern, replacement))

    def apply(self, text: str) -> str:
        if not text or not self._compiled:
            return text
        result = text
        for pattern, replacement in self._compiled:
            result = pattern.sub(replacement, result)
        return result

    def __len__(self) -> int:
        return len(self.rules)


def load_rules(project_root: Path) -> Dict[str, str]:
    """Read pronunciation.json, falling back to the shipped defaults."""
    path = Path(project_root) / RULES_FILE
    if not path.exists():
        return dict(DEFAULT_RULES)
    try:
        data = json.loads(path.read_text(encoding="utf-8"))
    except (json.JSONDecodeError, OSError) as exc:
        logger.warning("%s is not valid JSON (%s); using default pronunciations", path, exc)
        return dict(DEFAULT_RULES)

    if not isinstance(data, dict):
        logger.warning("%s must contain a JSON object of {word: respelling}", path)
        return dict(DEFAULT_RULES)

    # Keys starting with "_" are comments, so the file can document itself.
    rules = {
        str(k): str(v)
        for k, v in data.items()
        if str(k).strip() and not str(k).startswith("_")
    }
    logger.info("loaded %d pronunciation rule(s) from %s", len(rules), path.name)
    return rules


def build(project_root: Path, enabled: bool = True) -> Optional[Pronouncer]:
    """A Pronouncer, or None when the feature is switched off."""
    if not enabled:
        return None
    return Pronouncer(load_rules(project_root))