geonames_fetch.py

#!/usr/bin/env python3
"""
geonames_fetch.py — Fetch countries, provinces, and cities from GeoNames.

Outputs JSON files usable by csv_repair.py's real city mode:
- report/countries.json                      (list of country names)
- report/provinces_by_country.json          (country -> [admin1 names])
- report/cities_by_country.json             (country -> [top city names by population])

Sources (preferring official GeoNames dumps for reliability):
- Countries: https://www.geonames.org/countries/ (HTML) or
             http://download.geonames.org/export/dump/countryInfo.txt
- Provinces: http://download.geonames.org/export/dump/admin1CodesASCII.txt
- Cities:    http://download.geonames.org/export/dump/cities500.txt

Usage examples:
  python geonames_fetch.py --outdir report
  python geonames_fetch.py --outdir report --max-cities 30
  python geonames_fetch.py --prefer-html-countries
"""

from __future__ import annotations

import argparse
import json
import os
import re
from typing import Dict, List, Tuple

import urllib.request
import io
import zipfile
import unicodedata
from html.parser import HTMLParser
from typing import Optional


COUNTRIES_HTML_URL = "https://www.geonames.org/countries/"
COUNTRYINFO_URL = "http://download.geonames.org/export/dump/countryInfo.txt"
ADMIN1_URL = "http://download.geonames.org/export/dump/admin1CodesASCII.txt"
CITIES500_URL = "http://download.geonames.org/export/dump/cities500.txt"
CITIES500_ZIP_URL = "http://download.geonames.org/export/dump/cities500.zip"


def fetch_url(url: str, timeout: int = 30) -> bytes:
    with urllib.request.urlopen(url, timeout=timeout) as resp:
        return resp.read()


def parse_countries_from_html(html: str) -> List[str]:
    # Extract the main countries table rows; anchor text typically is country name
    # Look for links to country pages like "/countries/<code>/" or "/<ISO>/"
    # Fallback to table cells containing country names.
    countries: List[str] = []
    # Attempt to capture anchor text inside the countries table
    # This regex finds table rows with a link where the text is the country name.
    for m in re.finditer(r"<a[^>]+href=\"/\w{2}/?[^\"]*\"[^>]*>([^<]+)</a>", html, re.IGNORECASE):
        name = m.group(1).strip()
        if name and name not in countries:
            countries.append(name)
    # If too few extracted, try a broader pattern within the countries page
    if len(countries) < 50:
        for m in re.finditer(r"<a[^>]+href=\"/countries/[^\"]+\"[^>]*>([^<]+)</a>", html, re.IGNORECASE):
            name = m.group(1).strip()
            if name and name not in countries:
                countries.append(name)
    # Deduplicate and sort case-insensitively
    countries = sorted(set(countries), key=lambda x: x.casefold())
    return countries


def parse_countryinfo(text: str) -> Tuple[List[str], Dict[str, str]]:
    # Returns (country_names, iso2_to_name)
    names: List[str] = []
    iso2_to_name: Dict[str, str] = {}
    for line in text.splitlines():
        if not line or line.startswith("#"):
            continue
        parts = line.split("\t")
        # Format ref: https://download.geonames.org/export/dump/readme.txt
        # columns: ISO, ISO3, ISO-Numeric, fips, Country, Capital, Area(in sq km), Population, Continent,
        #          tld, CurrencyCode, CurrencyName, Phone, PostalCodeFormat, PostalCodeRegex, Languages, geonameId, neighbours, EquivalentFipsCode
        if len(parts) >= 5:
            iso2 = parts[0].strip()
            name = parts[4].strip()
            if iso2 and name:
                iso2_to_name[iso2] = name
                names.append(name)
    names = sorted(set(names), key=lambda x: x.casefold())
    return names, iso2_to_name


def parse_admin1(text: str, iso2_to_name: Dict[str, str]) -> Dict[str, List[str]]:
    # admin1CodesASCII.txt: code<TAB>name<TAB>asciiname<TAB>geonameid
    by_country: Dict[str, List[str]] = {}
    for line in text.splitlines():
        if not line:
            continue
        parts = line.split("\t")
        if len(parts) >= 4:
            code = parts[0].strip()  # e.g., US.NY or IN.KA
            name = parts[1].strip()
            # Extract ISO2 from code prefix
            if "." in code:
                iso2 = code.split(".", 1)[0]
                country = iso2_to_name.get(iso2)
                if country:
                    by_country.setdefault(country, []).append(name)
    # Deduplicate and sort entries for each country
    for c, lst in by_country.items():
        by_country[c] = sorted(set(lst), key=lambda x: x.casefold())
    return by_country


def _slugify_country(name: str) -> str:
    s = unicodedata.normalize("NFKD", name)
    s = "".join(c for c in s if not unicodedata.combining(c))
    s = s.lower()
    s = re.sub(r"[^a-z0-9\s-]", "", s)
    s = re.sub(r"\s+", "-", s).strip("-")
    return s


class _SubdivTableParser(HTMLParser):
    def __init__(self) -> None:
        super().__init__()
        self.in_target_table = False
        self.target_table_depth = 0
        self.in_tr = False
        self.in_cell = False
        self.current_cell_text: List[str] = []
        self.current_row: List[str] = []
        self.rows: List[List[str]] = []

    def handle_starttag(self, tag: str, attrs: List[Tuple[str, Optional[str]]]):
        if tag.lower() == "table":
            attrs_dict = {k.lower(): (v or "") for k, v in attrs}
            if attrs_dict.get("id", "").lower() == "subdivtable1":
                self.in_target_table = True
                self.target_table_depth = 1
            elif self.in_target_table:
                self.target_table_depth += 1
        elif self.in_target_table and tag.lower() == "tr":
            self.in_tr = True
            self.current_row = []
        elif self.in_target_table and tag.lower() in ("td", "th"):
            self.in_cell = True
            self.current_cell_text = []

    def handle_endtag(self, tag: str):
        if self.in_target_table and tag.lower() in ("td", "th") and self.in_cell:
            text = "".join(self.current_cell_text).strip()
            self.current_row.append(re.sub(r"\s+", " ", text))
            self.in_cell = False
            self.current_cell_text = []
        elif self.in_target_table and tag.lower() == "tr" and self.in_tr:
            # Only keep non-empty rows
            if any(cell.strip() for cell in self.current_row):
                self.rows.append(self.current_row)
            self.in_tr = False
            self.current_row = []
        elif tag.lower() == "table":
            if self.in_target_table:
                self.target_table_depth -= 1
                if self.target_table_depth <= 0:
                    self.in_target_table = False

    def handle_data(self, data: str):
        if self.in_target_table and self.in_cell:
            self.current_cell_text.append(data)


def parse_admin_divisions_html(html: str) -> List[str]:
    parser = _SubdivTableParser()
    parser.feed(html)
    rows = parser.rows
    if not rows:
        return []
    # Find header row: the first row that contains "Name of Subdivision"
    header_idx = None
    for i, row in enumerate(rows[:5]):
        joined = " ".join(row).lower()
        if "name of subdivision" in joined:
            header_idx = i
            break
    if header_idx is None:
        # Sometimes header is the first row with th; fall back to first row
        header_idx = 0
    headers = [c.strip() for c in rows[header_idx]]
    # Determine the column index for "Name of Subdivision"
    col_idx = None
    for j, h in enumerate(headers):
        if h.strip().lower() == "name of subdivision" or "name of subdivision" in h.strip().lower():
            col_idx = j
            break
    if col_idx is None:
        # Could not find the exact header; bail out
        return []
    names: List[str] = []
    for r in rows[header_idx + 1:]:
        if col_idx < len(r):
            val = r[col_idx].strip()
            if val and not re.fullmatch(r"[-\s]*", val):
                names.append(val)
    names = sorted(set(names), key=lambda x: x.casefold())
    return names


def fetch_provinces_from_html(iso2_to_name: Dict[str, str]) -> Dict[str, List[str]]:
    out: Dict[str, List[str]] = {}
    for iso2, country in iso2_to_name.items():
        slug = _slugify_country(country)
        url = f"https://www.geonames.org/{iso2}/administrative-division-{slug}.html"
        names: List[str] = []
        try:
            html_bytes = fetch_url(url)
            html_text = html_bytes.decode("utf-8", errors="replace")
            names = parse_admin_divisions_html(html_text)
        except Exception:
            names = []
        out[country] = names
    return out


def parse_cities500(text: str, iso2_to_name: Dict[str, str], max_cities: int) -> Dict[str, List[str]]:
    # cities500.txt columns: tab-separated; population at index 14, country code at index 8, name at index 1
    by_country: Dict[str, List[Tuple[int, str]]] = {}
    for line in text.splitlines():
        if not line or line.startswith("#"):
            continue
        parts = line.split("\t")
        if len(parts) >= 15:
            name = parts[1].strip()
            iso2 = parts[8].strip()
            pop_str = parts[14].strip()
            try:
                pop = int(pop_str) if pop_str else 0
            except ValueError:
                pop = 0
            country = iso2_to_name.get(iso2)
            if country and name:
                by_country.setdefault(country, []).append((pop, name))
    # For each country, sort by population desc and keep top N unique names
    out: Dict[str, List[str]] = {}
    for country, entries in by_country.items():
        entries.sort(key=lambda x: x[0], reverse=True)
        seen = set()
        names: List[str] = []
        for pop, nm in entries:
            if nm not in seen:
                names.append(nm)
                seen.add(nm)
            if len(names) >= max_cities:
                break
        out[country] = names
    return out


def safe_makedirs(path: str) -> None:
    if path and not os.path.exists(path):
        os.makedirs(path, exist_ok=True)


def main() -> int:
    ap = argparse.ArgumentParser(description="Fetch GeoNames countries, provinces, and cities into JSON files")
    ap.add_argument("--outdir", default="report", help="Directory to write JSON outputs")
    ap.add_argument("--max-cities", type=int, default=25, help="Top N cities per country by population")
    ap.add_argument("--prefer-html-countries", action="store_true", help="Prefer parsing countries from HTML page")
    ap.add_argument("--provinces-from-html", action="store_true", help="Fetch provinces per country from GeoNames HTML pages instead of admin1 dump")
    args = ap.parse_args()

    safe_makedirs(args.outdir)

    # Fetch country mappings
    countries: List[str] = []
    iso2_to_name: Dict[str, str] = {}

    countryinfo_bytes = fetch_url(COUNTRYINFO_URL)
    countryinfo_text = countryinfo_bytes.decode("utf-8", errors="replace")
    countries, iso2_to_name = parse_countryinfo(countryinfo_text)

    if args.prefer_html_countries:
        try:
            html_bytes = fetch_url(COUNTRIES_HTML_URL)
            html_text = html_bytes.decode("utf-8", errors="replace")
            html_countries = parse_countries_from_html(html_text)
            # If HTML yielded enough, prefer it; otherwise keep countryInfo list
            if len(html_countries) >= len(countries) - 5:
                countries = html_countries
        except Exception:
            pass

    # Fetch provinces (admin1)
    admin1_bytes = fetch_url(ADMIN1_URL)
    admin1_text = admin1_bytes.decode("utf-8", errors="replace")
    if args.provinces_from_html:
        provinces_by_country = fetch_provinces_from_html(iso2_to_name)
        # If some countries yielded empty from HTML, fill from admin1 dump
        admin1_fallback = parse_admin1(admin1_text, iso2_to_name)
        for c in admin1_fallback:
            if not provinces_by_country.get(c):
                provinces_by_country[c] = admin1_fallback[c]
    else:
        provinces_by_country = parse_admin1(admin1_text, iso2_to_name)

    # Fetch cities (top by population) — try .txt first, then .zip fallback
    try:
        cities_bytes = fetch_url(CITIES500_URL)
        cities_text = cities_bytes.decode("utf-8", errors="replace")
    except Exception:
        zip_bytes = fetch_url(CITIES500_ZIP_URL)
        with zipfile.ZipFile(io.BytesIO(zip_bytes)) as zf:
            member_name = None
            for n in zf.namelist():
                if n.lower().endswith(".txt") and "cities500" in n.lower():
                    member_name = n
                    break
            if member_name is None and zf.namelist():
                member_name = zf.namelist()[0]
            with zf.open(member_name) as f:
                cities_text = f.read().decode("utf-8", errors="replace")
    cities_by_country = parse_cities500(cities_text, iso2_to_name, args.max_cities)

    # Write JSON outputs
    with open(os.path.join(args.outdir, "countries.json"), "w", encoding="utf-8") as f:
        json.dump(countries, f, indent=2, ensure_ascii=False)
    with open(os.path.join(args.outdir, "provinces_by_country.json"), "w", encoding="utf-8") as f:
        json.dump(provinces_by_country, f, indent=2, ensure_ascii=False)
    with open(os.path.join(args.outdir, "cities_by_country.json"), "w", encoding="utf-8") as f:
        json.dump(cities_by_country, f, indent=2, ensure_ascii=False)

    print(f"Wrote: {os.path.join(args.outdir, 'countries.json')}")
    print(f"Wrote: {os.path.join(args.outdir, 'provinces_by_country.json')}")
    print(f"Wrote: {os.path.join(args.outdir, 'cities_by_country.json')}")
    return 0


if __name__ == "__main__":
    raise SystemExit(main())