From ebb1f560de543700d278caf490464abab1daa8da Mon Sep 17 00:00:00 2001 From: Wolf Date: Tue, 29 Sep 2026 21:58:32 +0200 Subject: [PATCH] Add places: folded whole-word place matching, country, region and town tokens, suggestions Co-Authored-By: Claude --- app/places.py | 537 +++++++++++++++++++++++++++++++++++++++++++ tests/test_places.py | 453 ++++++++++++++++++++++++++++++++++++ 2 files changed, 990 insertions(+) create mode 100644 app/places.py create mode 100644 tests/test_places.py diff --git a/app/places.py b/app/places.py new file mode 100644 index 0000000..ae564a6 --- /dev/null +++ b/app/places.py @@ -0,0 +1,537 @@ +"""Places as a filter: what a board writes in `place`, matched by whole words. + +A location line is free text: "København K", "Kgs. Lyngby", "Aarhus N, Region +Central Jutland, Denmark". A substring match finds `Lund` in `Billund` and +`Denmark` in nothing that only names a town. Here a place is folded (casefolded, +marks stripped, `ø` to `o`, `æ` to `ae`, `å` to `a`, punctuation to spaces) and +matched as whole words, so `Lund` is `lund` and no other word. A filter is one +of three tokens, and several tokens are joined by OR: + + country:DK the role's country, or a line that names it while `country` + is empty or `EU` (a board that lists several offices) + region:skane any town or name of a region in `REGIONS` + place:malmo the town's fold and its aliases, so `copenhagen`, + `kobenhavn` and `kopenhamn` find each other + +`match_sql` turns a token into SQL over `roles.place_fold` and `roles.country`; +`suggest` turns typed text into tokens for the autocomplete; `label` reads a +chip's text back from the address. No database access here, the caller hands +over the places it holds. +""" + +from __future__ import annotations + +import re +from collections import Counter +from collections.abc import Iterable +from dataclasses import dataclass + +from app.models import _CITIES, _COUNTRY_NAMES, _REGIONS, _place_fold + +__all__ = ["REGIONS", "Region", "Suggestion", "Token", "fold", "label", "match_sql", + "parse", "suggest"] + + +def fold(text: str) -> str: + """`Skåne län` -> `skane lan`, `København K` -> `kobenhavn k`, `Kgs. Lyngby` + -> `kgs lyngby`. What `roles.place_fold` holds. `å` needs no rule of its own: + NFKD splits it into `a` and a ring, and the ring is a mark.""" + return _place_fold(text) + + +# ------------------------------------------------------------------ countries + + +def _countries() -> dict[str, tuple[str, ...]]: + """Every code to the names `_COUNTRY_NAMES` gives it, English name first.""" + found: dict[str, tuple[str, ...]] = {} + for line in _COUNTRY_NAMES.split("\n"): + if line: + code, rest = line.split(" ", 1) + found[code] = tuple(name.strip() for name in rest.split("|")) + return found + + +_COUNTRY_LABELS = _countries() + +#: A code to the folded names a country goes by. +_COUNTRY_FOLDS: dict[str, tuple[str, ...]] = { + code: tuple(dict.fromkeys(filter(None, map(fold, names)))) + for code, names in _COUNTRY_LABELS.items()} + +#: What a line has to say for a role of that country whose `country` column is +#: empty or `EU`: a name of the country, or a city `country_of` knows in it. A +#: board that lists offices ("Copenhagen; London; Warsaw") names cities only. +_COUNTRY_MENTIONS: dict[str, tuple[str, ...]] = { + code: tuple(dict.fromkeys((*names, *(fold(c) for c, cc in _CITIES.items() if cc == code)))) + for code, names in _COUNTRY_FOLDS.items()} + +#: A folded country name to its code. +_COUNTRY_BY_NAME: dict[str, str] = { + name: code for code, names in _COUNTRY_FOLDS.items() for name in names} + +#: Words that are not a town: a country, a macro region, how the role is worked. +_NOT_A_TOWN = frozenset(_COUNTRY_BY_NAME) | frozenset(map(fold, _REGIONS)) \ + | {"remote", "hybrid"} + + +# -------------------------------------------------------------------- aliases + +#: Towns boards write in more than one language, first spelling shown. Each +#: group is symmetric: any spelling finds the others. The Nordics and the DACH +#: countries in full for their large cities, and the European cities that +#: appear both ways in boards. +_ALIAS_GROUPS: tuple[tuple[str, ...], ...] = ( + # Denmark + ("Copenhagen", "København", "Kobenhavn", "Köpenhamn", "Kopenhamn", "Kopenhagen", + "Copenhague", "Kopenhaga"), + ("Aarhus", "Århus"), + ("Aalborg", "Ålborg"), + ("Kgs. Lyngby", "Kongens Lyngby", "Lyngby"), + ("Taastrup", "Tåstrup"), + ("Elsinore", "Helsingør"), + # Sweden + ("Gothenburg", "Göteborg", "Goteborg"), + ("Stockholm",), + ("Malmö", "Malmo"), + ("Helsingborg", "Hälsingborg"), + ("Västerås", "Vasteras"), + ("Linköping", "Linkoping"), + ("Örebro", "Orebro"), + ("Norrköping", "Norrkoping"), + ("Jönköping", "Jonkoping"), + # Norway, Finland, Iceland + ("Oslo",), + ("Bergen",), + ("Trondheim", "Trondhjem"), + ("Helsinki", "Helsingfors"), + ("Espoo", "Esbo"), + ("Vantaa", "Vanda"), + ("Reykjavik", "Reykjavík"), + # Germany + ("Munich", "München", "Muenchen", "Munchen"), + ("Cologne", "Köln", "Koeln", "Koln"), + ("Nuremberg", "Nürnberg", "Nuernberg", "Nurnberg"), + ("Düsseldorf", "Duesseldorf", "Dusseldorf"), + ("Hanover", "Hannover"), + ("Mainz", "Mayence"), + ("Frankfurt", "Frankfurt am Main"), + ("Göttingen", "Goettingen", "Gottingen"), + ("Würzburg", "Wuerzburg", "Wurzburg"), + ("Tübingen", "Tuebingen", "Tubingen"), + ("Lübeck", "Luebeck", "Lubeck"), + ("Münster", "Muenster", "Munster"), + # Austria + ("Vienna", "Wien", "Vienne"), + ("Salzburg",), + ("Innsbruck",), + # Switzerland + ("Zurich", "Zürich", "Zuerich"), + ("Geneva", "Genève", "Geneve", "Genf", "Ginevra"), + ("Basel", "Basle", "Bâle"), + ("Bern", "Berne"), + ("Lucerne", "Luzern"), + ("St. Gallen", "Sankt Gallen", "Saint Gall"), + # The rest of Europe + ("Prague", "Praha", "Prag"), + ("Warsaw", "Warszawa", "Warschau"), + ("Krakow", "Kraków", "Cracow"), + ("Wroclaw", "Wrocław", "Breslau"), + ("Gdansk", "Gdańsk", "Danzig"), + ("Poznan", "Poznań", "Posen"), + ("Lisbon", "Lisboa", "Lissabon"), + ("Milan", "Milano", "Mailand"), + ("Rome", "Roma", "Rom"), + ("Turin", "Torino"), + ("Florence", "Firenze"), + ("Naples", "Napoli", "Neapel"), + ("Venice", "Venezia", "Venedig"), + ("Genoa", "Genova"), + ("Brussels", "Bruxelles", "Brussel", "Brüssel"), + ("Antwerp", "Antwerpen", "Anvers"), + ("Ghent", "Gent", "Gand"), + ("Bruges", "Brugge"), + ("Liège", "Liege", "Luik", "Lüttich"), + ("The Hague", "Den Haag", "s-Gravenhage"), + ("Luxembourg", "Luxemburg"), + ("Seville", "Sevilla"), + ("Athens", "Athina", "Athen"), + ("Bucharest", "București", "Bucuresti"), + ("Belgrade", "Beograd"), + ("Bratislava", "Pressburg"), + ("Kyiv", "Kiev"), + ("Cluj-Napoca", "Cluj"), + ("Timișoara", "Timisoara"), + ("Marseille", "Marseilles"), + ("Bolzano", "Bozen"), +) + + +def _variants(text: str) -> tuple[str, ...]: + """A fold and its `aa` spelling: Danish writes `å` as `aa` (`Aarhus`, `Taastrup`) + and `Århus` and `Tåstrup` fold to `arhus` and `tastrup`.""" + return tuple(dict.fromkeys((text, text.replace("aa", "a")))) + + +def _aliases() -> tuple[dict[str, tuple[str, ...]], dict[str, str]]: + """Every folded spelling to its whole group (all folded), and to its shown + spelling. The group starts with the first spelling listed.""" + groups: dict[str, tuple[str, ...]] = {} + shown: dict[str, str] = {} + for group in _ALIAS_GROUPS: + members = tuple(dict.fromkeys(v for name in group for v in _variants(fold(name)))) + for name in group: + shown.setdefault(fold(name), name) + for member in members: + groups[member] = members + return groups, shown + + +_GROUPS, _SHOWN = _aliases() + + +def _expand(text: str) -> tuple[str, ...]: + """A folded town and everything it is also written as.""" + return _GROUPS.get(text) or _variants(text) + + +# -------------------------------------------------------------------- regions + + +@dataclass(frozen=True) +class Region: + """A region below a country, or across two. `names` are what a board calls + the region itself, `towns` the places in it; a line is in the region when it + has one of either as whole words. Spellings are shown with their marks and + folded when used.""" + + key: str + label: str + names: tuple[str, ...] + towns: tuple[str, ...] + country: str = "" + + +def _split(text: str) -> tuple[str, ...]: + return tuple(part.strip() for part in text.split(",") if part.strip()) + + +_SKANE = Region( + "skane", "Skåne", + _split("Skåne, Skåne län, Skåne County, Region Skåne, Skåne Region, Scania, Skaane"), + _split("""Malmö, Lund, Helsingborg, Kristianstad, Landskrona, Trelleborg, Ängelholm, + Hässleholm, Ystad, Eslöv, Höganäs, Staffanstorp, Lomma, Bjuv, Burlöv, Kävlinge, + Svedala, Vellinge, Simrishamn, Tomelilla, Sjöbo, Svalöv, Klippan, Åstorp, + Bromölla, Osby, Perstorp, Örkelljunga, Östra Göinge, Hörby, Höör, Båstad, + Skurup, Arlöv, Limhamn, Dalby, Åkarp, Bjärred, Höllviken, Ljungbyhed, + Löddeköpinge, Furulund, Genarp, Veberöd, Skanör, Falsterbo, Mölle, Arild"""), + "SE") + +_HOVEDSTADEN = Region( + "hovedstaden", "Capital Region of Denmark", + _split("Capital Region of Denmark, Capital Region Denmark, Region Hovedstaden, Hovedstaden"), + _split("""København, Copenhagen, Frederiksberg, Albertslund, Allerød, Ballerup, + Bornholm, Brøndby, Dragør, Egedal, Fredensborg, Frederikssund, Furesø, Gentofte, + Gladsaxe, Glostrup, Gribskov, Halsnæs, Helsingør, Herlev, Hillerød, Hvidovre, + Høje-Taastrup, Hørsholm, Ishøj, Lyngby-Taarbæk, Rudersdal, Rødovre, Tårnby, + Vallensbæk, Kgs. Lyngby, Kongens Lyngby, Lyngby, Virum, Holte, Nærum, Birkerød, + Vedbæk, Skodsborg, Klampenborg, Charlottenlund, Hellerup, Ordrup, Bagsværd, + Søborg, Mørkhøj, Måløv, Skovlunde, Smørum, Ølstykke, Stenløse, Værløse, Farum, + Slangerup, Hedehusene, Taastrup, Kastrup, Amager, Valby, Vanløse, Brønshøj, + Nørrebro, Østerbro, Vesterbro, Nordhavn, Ørestad, Sydhavn, Islands Brygge, + Refshaleøen, Humlebæk, Kokkedal, Espergærde, Snekkersten, Rungsted, Rungsted Kyst, + Lillerød, Frederiksværk, Hundested, Helsinge, Gilleleje, Græsted, Vallensbæk Strand, + Skævinge, Rønne, Nexø, Allinge, Svaneke, Aakirkeby, Hasle"""), + "DK") + +# Region Sjælland, not the island: Copenhagen is on Zealand too, but a board that +# writes "Region Zealand" means the region and the Capital Region has its own key. +# The bare word is left out of `names` since "New Zealand" would match it. +_ZEALAND = Region( + "zealand", "Zealand", + _split("Region Zealand, Region Sjælland, Zealand Denmark, Sjælland, Sjaelland"), + _split("""Faxe, Greve, Guldborgsund, Holbæk, Kalundborg, Køge, Lejre, Lolland, Næstved, + Odsherred, Ringsted, Roskilde, Slagelse, Solrød, Sorø, Stevns, Vordingborg, + Nykøbing Falster, Nykøbing F, Nykøbing Sjælland, Nakskov, Maribo, Sakskøbing, + Haslev, Korsør, Skælskør, Præstø, Stege, Fakse, Rødby, Rødbyhavn, Jyllinge, + Viby Sjælland, Solrød Strand, Fuglebjerg, Gørlev, Dianalund, Jyderup, Hvalsø, + Tølløse, Hørve, Fårevejle, Bjæverskov, Nørre Asmindrup"""), + "DK") + +_JUTLAND = Region( + "jutland", "Jutland", + _split("""Jutland, Jylland, Central Jutland, Midtjylland, Region Midtjylland, + North Jutland, Nordjylland, Region Nordjylland, South Jutland, Sønderjylland, + East Jutland, West Jutland, Vestjylland, Østjylland"""), + _split("""Aarhus, Aalborg, Esbjerg, Randers, Kolding, Horsens, Vejle, Silkeborg, Herning, + Viborg, Billund, Fredericia, Holstebro, Skive, Ringkøbing, Videbæk, Lem, Risskov, + Åbyhøj, Brabrand, Viby J, Aabenraa, Haderslev, Sønderborg, Grenaa, Grenå, + Hjørring, Frederikshavn, Skagen, Thisted, Struer, Ikast, Brande, Varde, Ribe, + Tønder, Vojens, Hadsten, Hobro, Skanderborg, Odder, Hedensted, Vejen, Ebeltoft, + Lemvig, Rønde, Egå, Lystrup, Hinnerup, Tilst, Hammel, Galten, Sabro, Harlev, + Løsning, Højbjerg, Juelsminde, Nørresundby, Brønderslev, Aars, Nibe, Sæby, Løgstør, Tarm, + Skjern, Padborg, Rødekro, Toftlund, Bramming, Kruså, Brabrand, Ry, Gjern"""), + "DK") + +_FUNEN = Region( + "funen", "Funen", + _split("Funen, Fyn, Region Fyn"), + _split("""Odense, Svendborg, Nyborg, Faaborg, Fåborg, Middelfart, Assens, Kerteminde, + Bogense, Langeskov, Ringe, Otterup, Munkebo, Vissenbjerg, Årslev, Tommerup, + Ærø, Ærøskøbing, Langeland, Rudkøbing, Tåsinge, Marstal, Nordfyn"""), + "DK") + +# The Øresund region is the two capital regions either side of the sound. +_ORESUND = Region( + "oresund", "Øresund", + _split("Øresund, Öresund, Oresund, Oresund Region, Øresundsregionen, Öresundsregionen") + + _HOVEDSTADEN.names + _SKANE.names, + _HOVEDSTADEN.towns + _SKANE.towns) + +REGIONS: dict[str, Region] = { + r.key: r for r in (_SKANE, _HOVEDSTADEN, _ZEALAND, _ORESUND, _JUTLAND, _FUNEN)} + + +def _phrases(region: Region) -> tuple[str, ...]: + """Every folded phrase that puts a line in the region, with the other + spellings of its towns.""" + found: dict[str, None] = {} + for text in (*region.names, *region.towns): + for form in _expand(fold(text)): + found[form] = None + return tuple(found) + + +_PHRASES: dict[str, tuple[str, ...]] = {key: _phrases(r) for key, r in REGIONS.items()} + +#: A folded key, label or name to the region key; a name in two regions ("Capital +#: Region of Denmark" is also in Øresund) belongs to the one that lists it first. +_REGION_BY_NAME: dict[str, str] = {} +for _key, _region in REGIONS.items(): + for _name in (_region.key, _region.label, *_region.names): + _REGION_BY_NAME.setdefault(fold(_name), _key) + +#: What a place is shown as when a chip is read from the address: every alias +#: spelling, every region and every town of one, as written above. +_TOWN_LABELS: dict[str, str] = dict(_SHOWN) +for _region in REGIONS.values(): + for _town in (_region.label, *_region.towns, *_region.names): + _TOWN_LABELS.setdefault(fold(_town), _town) + +_NOT_A_TOWN |= frozenset(_REGION_BY_NAME) + + +# --------------------------------------------------------------------- tokens + + +@dataclass(frozen=True) +class Token: + """One place filter. `value` is a country code, a region key or a fold.""" + + kind: str + value: str + label: str + + def __str__(self) -> str: + return f"{self.kind}:{self.value}" + + +def _region_key(text: str) -> str: + folded = fold(text) + return _REGION_BY_NAME.get(folded, folded) + + +def parse(token: str) -> Token: + """`country:DK`, `region:skane` or `place:malmo`. Bare text, or a prefix that + is none of these, is a place: `Skåne` is `place:skane`. A country by name + (`country:Denmark`) and a region by label (`region:Skåne`) are read as well.""" + kind, sep, value = (token or "").partition(":") + kind = kind.strip().lower() + if not sep or kind not in ("country", "region", "place"): + kind, value = "place", token or "" + value = value.strip() + if kind == "country": + code = value.upper() + if code not in _COUNTRY_LABELS: + code = _COUNTRY_BY_NAME.get(fold(value), code) + return Token("country", code, _COUNTRY_LABELS.get(code, (code,))[0]) + if kind == "region": + key = _region_key(value) + return Token("region", key, REGIONS[key].label if key in REGIONS else key) + folded = fold(value) + return Token("place", folded, _TOWN_LABELS.get(folded) or folded.title()) + + +def label(token: str | Token) -> str: + """The text of a chip read from the address: `country:DK` is `Denmark`, + `region:skane` is `Skåne`, `place:malmo` is `Malmö`. A town the tables do not + know is its fold in title case.""" + return (token if isinstance(token, Token) else parse(token)).label + + +# ------------------------------------------------------------------------ SQL + +_NEVER = ("0", []) + + +def _escape(word: str) -> str: + """A folded word holds no wildcard, but a caller may hand over raw text.""" + return re.sub(r"([\\%_])", r"\\\1", word) + + +def _words(column: str, phrases: Iterable[str]) -> tuple[str, list]: + """Any of the phrases as whole words of the column. A multi-word phrase is + a phrase: `kgs lyngby` needs both words, side by side.""" + phrases = [p for p in dict.fromkeys(phrases) if p] + if not phrases: + return _NEVER + clause = f"' ' || {column} || ' ' LIKE ? ESCAPE '\\'" + return " OR ".join([clause] * len(phrases)), [f"% {_escape(p)} %" for p in phrases] + + +def _country(code: str, place: str, country: str) -> tuple[str, list]: + """The role's own country, or a line that names the country while `country` + is empty or `EU`: a board listing several offices has no single country.""" + names, params = _words(place, _COUNTRY_MENTIONS.get(code, ())) + if not params: + return f"{country} = ?", [code] + return f"{country} = ? OR ({country} IN ('', 'EU') AND ({names}))", [code, *params] + + +def match_sql(token: str, *, place: str = "place_fold", country: str = "country") \ + -> tuple[str, list]: + """SQL over the two columns for one token, parenthesised, with `?` parameters. + Several tokens are joined with OR by the caller. Bare text is a place. + + A `place:` value that is also a region or a country is matched as that too: + `place:skane` is Skåne the region, `place:denmark` is Denmark, since a gate + written as a bare word means what the owner meant by it. A token that names + nothing known (`region:atlantis`, an empty value) matches no role.""" + tok = parse(token) + if tok.kind == "country": + sql, params = _country(tok.value, place, country) + elif tok.kind == "region": + sql, params = _words(place, _PHRASES.get(tok.value, ())) + else: + region = _REGION_BY_NAME.get(tok.value) + sql, params = _words(place, [*_expand(tok.value), *_PHRASES.get(region, ())]) \ + if tok.value else _NEVER + if tok.value in _COUNTRY_BY_NAME: + more, extra = _country(_COUNTRY_BY_NAME[tok.value], place, country) + sql, params = f"{sql} OR {more}", [*params, *extra] + return f"({sql})", params + + +# ---------------------------------------------------------------- suggestions + + +@dataclass(frozen=True) +class Suggestion: + """`token` is what the chip holds, `kind` is `country`, `region` or `place`, + `count` the open roles it stands for.""" + + token: str + label: str + kind: str + count: int + + +def _prefixes(q: str, candidates: Iterable[str]) -> bool: + """`q` starts a word of a candidate, or a run of words: `kgs ly` finds + `kgs lyngby`, `hague` finds `the hague`.""" + return any(c.startswith(q) or f" {q}" in f" {c}" for c in candidates) + + +#: Where one town ends in a line: `;` between places, and `or` / `eller` / `og` +#: between two towns in one ("København V eller Aalborg"). +_PART = re.compile(r"\s*;\s*|\s+(?:or|eller|og)\s+", re.IGNORECASE) + +#: Danish boards append "and the possibility of working from home". +_AND_HOME = re.compile(r"\s+(?:og|samt) mulighed for hjemmearbejde\s*$", re.IGNORECASE) + +#: A postal district after the town, "København K", "Aarhus C", "Odense SV". +_DISTRICT = re.compile(r"\s+[A-ZÆØÅ]{1,2}$") + +#: A country or "Remote" ahead of the town, "Sweden - Lund", "Remote-New Zealand". +_LEAD = re.compile(r"^(.+?)\s*[-\u2013]\s*(.+)$") + + +def _towns(place: str) -> list[str]: + """The first comma part of each place of a line: `Aarhus N, Region Central + Jutland, Denmark; København, Denmark` is `Aarhus`, `København`. A postal + district is dropped, since `place:kobenhavn` finds `København K` anyway.""" + found = [] + for part in _PART.split(_AND_HOME.sub("", place or "")): + head = part.split(",")[0].strip() + if (lead := _LEAD.match(head)) and fold(lead[1]) in _NOT_A_TOWN: + head = lead[2] + head = _DISTRICT.sub("", head) + if fold(head) and fold(head) not in _NOT_A_TOWN: + found.append(head) + return found + + +def _region_counts(keys: Iterable[str], places: list[tuple[str, int]]) -> dict[str, int]: + """Open roles per region, over folded lines. A line is cut into every run of + words once, and each region asks whether it holds one of its phrases.""" + sets = {key: set(_PHRASES[key]) for key in keys} + longest = max((len(p.split()) for key in sets for p in sets[key]), default=1) + counts: Counter[str] = Counter() + for text, roles in places: + words = text.split() + runs = {" ".join(words[i:i + n]) + for n in range(1, longest + 1) for i in range(len(words) - n + 1)} + for key, phrases in sets.items(): + if not phrases.isdisjoint(runs): + counts[key] += roles + return counts + + +def suggest(q: str, places: Iterable[tuple[str, str, int]], limit: int = 12) \ + -> list[Suggestion]: + """Tokens for the text typed so far. `places` is `(place, country, open + roles)`. Countries first, then regions, then towns, each by open roles. + + A country is suggested when `q` starts a word of one of its names, with the + roles that carry its code. A region likewise on its label and names, with the + roles whose line is in it. A town is the first comma part of each `;` part of + a line, grouped by fold and by alias (`Copenhagen` and `København` are one + town, shown as its commonest spelling). Countries, regions and towns with no + open role are left out: a chip that finds nothing helps nobody.""" + text = fold(q) + if not text: + return [] + rows = [(place or "", country or "", int(roles)) for place, country, roles in places] + + by_country: Counter[str] = Counter() + for _, country, roles in rows: + by_country[country] += roles + countries = [Suggestion(f"country:{code}", names[0], "country", by_country[code]) + for code, names in _COUNTRY_LABELS.items() + if by_country[code] and _prefixes(text, _COUNTRY_FOLDS[code])] + + named = [key for key, r in REGIONS.items() + if _prefixes(text, [fold(r.label), *map(fold, r.names)])] + counts = _region_counts(named, [(fold(place), roles) for place, _, roles in rows]) + regions = [Suggestion(f"region:{key}", REGIONS[key].label, "region", counts[key]) + for key in named if counts[key]] + + spellings: dict[str, Counter[str]] = {} + written: dict[str, set[str]] = {} + for place, _, roles in rows: + for town in _towns(place): + folded = fold(town) + group = _GROUPS[folded][0] if folded in _GROUPS else _variants(folded)[-1] + spellings.setdefault(group, Counter())[town] += roles + written.setdefault(group, set()).add(folded) + towns = [] + for group, seen in spellings.items(): + if not _prefixes(text, [form for folded in written[group] for form in _expand(folded)]): + continue + shown = min(seen, key=lambda s: (-seen[s], s)) + towns.append(Suggestion(f"place:{fold(shown)}", shown, "place", sum(seen.values()))) + + countries.sort(key=lambda s: (-s.count, s.label)) + regions.sort(key=lambda s: (-s.count, s.label)) + towns.sort(key=lambda s: (-s.count, s.label)) + return (countries + regions + towns)[:limit] diff --git a/tests/test_places.py b/tests/test_places.py new file mode 100644 index 0000000..d1f83df --- /dev/null +++ b/tests/test_places.py @@ -0,0 +1,453 @@ +"""`app.places`: folding, the three token kinds run as SQL, and the suggestions. + +The rows are `place` strings as boards write them, taken from the live database. +Each token's SQL runs against an in-memory table shaped like `roles`.""" + +from __future__ import annotations + +import sqlite3 + +import pytest + +from app import places +from app.places import REGIONS, Suggestion, Token, fold, label, match_sql, parse, suggest + +#: (place, country) as a board wrote them and `country_of` read them. +ROWS = [ + ("København", "DK"), ("Copenhagen, Denmark", "DK"), ("Humlebæk, Denmark", "DK"), + ("København K", "DK"), ("København Ø", "DK"), ("Ballerup", "DK"), ("Denmark", "DK"), + ("Aarhus N, Region Central Jutland, Denmark", "DK"), ("Aarhus C, Denmark", "DK"), + ("Billund, Denmark", "DK"), ("Kgs. Lyngby", "DK"), ("Kongens Lyngby", "DK"), + ("Måløv, Capital Region of Denmark, Denmark", "DK"), ("Søborg", "DK"), + ("Kalundborg, Region Zealand, Denmark", "DK"), ("Albertslund", "DK"), + ("Odense, REGION_DNK 1852 South Denmark, Denmark", "DK"), ("Nakskov", "DK"), + ("Nordhavn og mulighed for hjemmearbejde", "DK"), ("Remote OR Copenhagen", "DK"), + ("Malmö", "SE"), ("Lund", "SE"), ("Sweden - Lund", "SE"), ("Helsingborg", "SE"), + ("Malmö, Skåne County, Sweden", "SE"), ("Skåne län, Sweden", "SE"), + ("Höganäs, Sweden", "SE"), ("Stockholm, Sweden", "SE"), ("Gothenburg, Sweden", "SE"), + ("Aarhus, Denmark; København, Denmark; Stockholm, Sweden", "EU"), + ("Copenhagen; London; Warsaw; Milano", "EU"), + ("Copenhagen; Warsaw; New York", ""), + ("Berlin, Germany; Munich, Germany", "DE"), ("Berlin, Germany", "DE"), + ("München", "DE"), ("Munich, Bavaria, Germany", "DE"), ("Köln", "DE"), ("Cologne", "DE"), + ("Zürich", "CH"), ("Zurich, Switzerland", "CH"), ("Genève", "CH"), ("Geneva", "CH"), + ("Wien", "AT"), ("Vienna, Austria", "AT"), + ("Auckland, New Zealand", "NZ"), ("Remote", ""), ("Hybrid", ""), ("Europe", ""), + ("Remote - Denmark", "DK"), ("", ""), +] + + +@pytest.fixture() +def table() -> sqlite3.Connection: + db = sqlite3.connect(":memory:") + db.execute("CREATE TABLE roles (place TEXT, place_fold TEXT, country TEXT)") + db.executemany("INSERT INTO roles VALUES (?, ?, ?)", + [(place, fold(place), country) for place, country in ROWS]) + return db + + +def found(db: sqlite3.Connection, *tokens: str) -> set[str]: + """The places the tokens find, joined with OR as a caller would.""" + parts = [match_sql(token) for token in tokens] + sql = " OR ".join(sql for sql, _ in parts) + params = [p for _, params in parts for p in params] + return {row[0] for row in db.execute(f"SELECT place FROM roles WHERE {sql}", params)} + + +# ------------------------------------------------------------------- folding + + +@pytest.mark.parametrize("text, folded", [ + ("Skåne län", "skane lan"), + ("København K", "kobenhavn k"), + ("Kgs. Lyngby", "kgs lyngby"), + ("Malmö", "malmo"), + ("Åbyhøj", "abyhoj"), + ("Ærø", "aero"), + ("Düsseldorf", "dusseldorf"), + ("Zürich", "zurich"), + ("Wrocław", "wroclaw"), + ("Höje-Taastrup, Denmark", "hoje taastrup denmark"), + (" Aarhus N ", "aarhus n"), + ("", ""), +]) +def test_fold(text: str, folded: str) -> None: + assert fold(text) == folded + + +def test_fold_turns_a_ring_and_umlauts_into_the_plain_letter() -> None: + assert fold("å") == "a" + assert (fold("ä"), fold("ö"), fold("ü")) == ("a", "o", "u") + + +# ------------------------------------------------------------------- parsing + + +def test_parse_reads_the_three_kinds() -> None: + assert parse("country:DK") == Token("country", "DK", "Denmark") + assert parse("region:skane") == Token("region", "skane", "Skåne") + assert parse("place:malmo") == Token("place", "malmo", "Malmö") + + +def test_parse_reads_bare_text_as_a_place() -> None: + assert parse("Skåne") == Token("place", "skane", "Skåne") + assert parse("Kgs. Lyngby").value == "kgs lyngby" + assert parse("Foo: Bar").kind == "place" + + +def test_parse_is_forgiving_about_case_and_names() -> None: + assert parse("Country:dk").value == "DK" + assert parse("country:Denmark").value == "DK" + assert parse("region:Skåne").value == "skane" + assert parse("region:Capital Region of Denmark").value == "hovedstaden" + assert parse("place:København").value == "kobenhavn" + + +def test_a_token_round_trips_through_its_string() -> None: + for text in ("country:DK", "region:oresund", "place:kgs lyngby"): + assert str(parse(text)) == text + + +def test_label_reads_a_chip_from_the_address() -> None: + assert label("country:SE") == "Sweden" + assert label("region:hovedstaden") == "Capital Region of Denmark" + assert label("place:kobenhavn") == "København" + assert label("place:copenhagen") == "Copenhagen" + assert label("place:malmo") == "Malmö" + assert label("place:tromso") == "Tromso" + assert label(parse("place:kgs lyngby")) == "Kgs. Lyngby" + assert label("country:ZZ") == "ZZ" + assert label("region:atlantis") == "atlantis" + + +# ---------------------------------------------------------------------- SQL + + +def test_a_place_is_matched_on_whole_words(table: sqlite3.Connection) -> None: + lund = found(table, "place:lund") + assert lund == {"Lund", "Sweden - Lund"} + assert "Billund, Denmark" not in lund and "Albertslund" not in lund + assert found(table, "place:billund") == {"Billund, Denmark"} + + +def test_a_multi_word_place_is_a_phrase(table: sqlite3.Connection) -> None: + assert found(table, "place:kgs lyngby") == {"Kgs. Lyngby", "Kongens Lyngby"} + assert found(table, "place:capital denmark") == set() + assert found(table, "place:denmark capital") == set() + + +def test_diacritics_and_spellings_meet(table: sqlite3.Connection) -> None: + assert found(table, "place:Skåne") == found(table, "place:skane") + assert "Malmö" in found(table, "place:malmo") and "Malmö" in found(table, "place:Malmö") + assert {"København", "København K"} <= found(table, "place:kobenhavn") + + +COPENHAGENS = {"København", "Copenhagen, Denmark", "København K", "Remote OR Copenhagen"} + + +@pytest.mark.parametrize("spelling", ["copenhagen", "kobenhavn", "kopenhamn", "kopenhagen", + "København", "Köpenhamn"]) +def test_copenhagen_is_every_spelling(table: sqlite3.Connection, spelling: str) -> None: + assert found(table, f"place:{spelling}") >= COPENHAGENS + + +@pytest.mark.parametrize("group", places._ALIAS_GROUPS, ids=lambda g: g[0]) +def test_every_alias_finds_every_other(group: tuple[str, ...]) -> None: + db = sqlite3.connect(":memory:") + db.execute("CREATE TABLE roles (place TEXT, place_fold TEXT, country TEXT)") + db.executemany("INSERT INTO roles VALUES (?, ?, '')", + [(name, fold(name)) for name in group]) + for spelling in group: + sql, params = match_sql(f"place:{spelling}") + hit = {r[0] for r in db.execute(f"SELECT place FROM roles WHERE {sql}", params)} + assert hit == set(group), spelling + + +def test_no_spelling_belongs_to_two_groups() -> None: + seen: dict[str, str] = {} + for group in places._ALIAS_GROUPS: + for name in group: + assert seen.setdefault(fold(name), group[0]) == group[0], name + + +@pytest.mark.parametrize("pair", [("Munich", "München"), ("Cologne", "Köln"), + ("Vienna", "Wien"), ("Zurich", "Zürich"), + ("Geneva", "Genève")]) +def test_a_city_written_two_ways_is_found_either_way( + table: sqlite3.Connection, pair: tuple[str, str]) -> None: + one, other = pair + hit = found(table, f"place:{one}") + assert hit == found(table, f"place:{other}") + words = {w for place in hit for w in fold(place).split()} + assert {fold(one), fold(other)} <= words + + +def test_aa_and_the_ring_are_one_letter(table: sqlite3.Connection) -> None: + hit = found(table, "place:århus") + assert hit == found(table, "place:aarhus") == { + "Aarhus N, Region Central Jutland, Denmark", "Aarhus C, Denmark", + "Aarhus, Denmark; København, Denmark; Stockholm, Sweden"} + + +def test_a_line_with_several_places_is_found_by_each(table: sqlite3.Connection) -> None: + line = "Aarhus, Denmark; København, Denmark; Stockholm, Sweden" + for tok in ("place:aarhus", "place:kobenhavn", "place:stockholm", "place:copenhagen"): + assert line in found(table, tok), tok + + +def test_several_tokens_are_joined_by_or(table: sqlite3.Connection) -> None: + hit = found(table, "country:DK", "region:skane") + assert {"København", "Malmö", "Lund"} <= hit + assert "Stockholm, Sweden" not in hit + + +def test_a_country_is_its_code(table: sqlite3.Connection) -> None: + hit = found(table, "country:DK") + assert {"København", "Denmark", "Ballerup", "Nakskov", "Odense, REGION_DNK 1852 South Denmark, Denmark"} <= hit + assert "Malmö" not in hit and "Berlin, Germany" not in hit + assert "Auckland, New Zealand" not in hit + + +def test_a_country_is_also_a_line_that_names_it_when_country_is_unset( + table: sqlite3.Connection) -> None: + hit = found(table, "country:DK") + # `country` is EU: a board listing offices in several countries. + assert "Aarhus, Denmark; København, Denmark; Stockholm, Sweden" in hit + # Only cities are named, and one of them is Danish. + assert "Copenhagen; London; Warsaw; Milano" in hit + assert "Copenhagen; Warsaw; New York" in hit + # The same words under a country that is set are that country's business. + assert found(table, "country:SE") >= {"Stockholm, Sweden", "Malmö"} + assert "Aarhus, Denmark; København, Denmark; Stockholm, Sweden" in found(table, "country:SE") + assert "Copenhagen, Denmark" in hit # its own country is DK + + +def test_a_country_name_does_not_reach_a_role_of_another_country() -> None: + db = sqlite3.connect(":memory:") + db.execute("CREATE TABLE roles (place TEXT, place_fold TEXT, country TEXT)") + db.executemany("INSERT INTO roles VALUES (?, ?, ?)", [ + ("Remote, Denmark", "remote denmark", "GB"), ("Copenhagen", "copenhagen", "DE"), + ("Copenhagen", "copenhagen", "DK"), ("Nowhere, Denmark", "nowhere denmark", "")]) + sql, params = match_sql("country:DK") + assert db.execute(f"SELECT country FROM roles WHERE {sql} ORDER BY country", + params).fetchall() == [("",), ("DK",)] + + +def test_country_eu_is_the_column_alone(table: sqlite3.Connection) -> None: + assert found(table, "country:EU") == {"Aarhus, Denmark; København, Denmark; Stockholm, Sweden", + "Copenhagen; London; Warsaw; Milano"} + + +def test_a_region_is_its_names_and_its_towns(table: sqlite3.Connection) -> None: + skane = found(table, "region:skane") + assert skane == {"Malmö", "Lund", "Sweden - Lund", "Helsingborg", "Höganäs, Sweden", + "Malmö, Skåne County, Sweden", "Skåne län, Sweden"} + assert found(table, "region:Skåne") == skane + + hovedstaden = found(table, "region:hovedstaden") + assert {"København", "København K", "København Ø", "Copenhagen, Denmark", "Ballerup", + "Humlebæk, Denmark", "Kgs. Lyngby", "Kongens Lyngby", "Søborg", "Albertslund", + "Måløv, Capital Region of Denmark, Denmark", "Nordhavn og mulighed for hjemmearbejde", + } <= hovedstaden + assert not hovedstaden & {"Aarhus C, Denmark", "Billund, Denmark", "Malmö", "Nakskov"} + + +def test_zealand_is_the_danish_region_and_not_new_zealand(table: sqlite3.Connection) -> None: + assert found(table, "region:zealand") == {"Kalundborg, Region Zealand, Denmark", "Nakskov"} + + +def test_jutland_and_funen(table: sqlite3.Connection) -> None: + assert {"Billund, Denmark", "Aarhus C, Denmark"} <= found(table, "region:jutland") + assert found(table, "region:funen") == {"Odense, REGION_DNK 1852 South Denmark, Denmark"} + + +def test_oresund_is_the_two_capital_regions(table: sqlite3.Connection) -> None: + assert found(table, "region:oresund") == ( + found(table, "region:hovedstaden") | found(table, "region:skane")) + + +def test_a_region_is_a_place_written_as_bare_text(table: sqlite3.Connection) -> None: + assert found(table, "Skåne") == found(table, "region:skane") + assert found(table, "place:scania") == found(table, "region:skane") + assert found(table, "Denmark") == found(table, "country:DK") + + +def test_a_token_that_names_nothing_finds_nothing(table: sqlite3.Connection) -> None: + assert found(table, "region:atlantis") == set() + assert found(table, "place:") == set() + assert found(table, "") == set() + assert found(table, "place:nowhere-in-particular") == set() + + +def test_the_columns_are_the_callers(table: sqlite3.Connection) -> None: + sql, params = match_sql("country:DK", place="r.pf", country="r.c") + assert "r.pf" in sql and "r.c " in sql and "place_fold" not in sql + db = sqlite3.connect(":memory:") + db.execute("CREATE TABLE r (pf TEXT, c TEXT)") + db.executemany("INSERT INTO r VALUES (?, ?)", [("lund", "SE"), ("billund", "DK")]) + sql, params = match_sql("place:lund", place="r.pf", country="r.c") + assert db.execute(f"SELECT pf FROM r WHERE {sql}", params).fetchall() == [("lund",)] + + +def test_sql_is_parenthesised_and_has_one_parameter_per_question_mark() -> None: + for token in ("country:DK", "region:oresund", "place:copenhagen", "place:denmark", + "region:atlantis"): + sql, params = match_sql(token) + assert sql.startswith("(") and sql.endswith(")") + assert sql.count("?") == len(params), token + + +def test_a_wildcard_in_a_value_is_a_letter() -> None: + assert places._escape("50%_off\\") == "50\\%\\_off\\\\" + db = sqlite3.connect(":memory:") + db.execute("CREATE TABLE roles (place TEXT, place_fold TEXT, country TEXT)") + db.executemany("INSERT INTO roles VALUES (?, ?, '')", [("a b", "a b"), ("axb", "axb"), + ("a_b", "a_b")]) + for token in ("place:a_b", "place:a%b"): + sql, params = match_sql(token) + assert {r[0] for r in db.execute(f"SELECT place FROM roles WHERE {sql}", params)} \ + == {"a b"}, token + # A column that is not folded still matches a literal underscore, not any character. + sql, params = places._words("place", ["a_b"]) + assert db.execute(f"SELECT place FROM roles WHERE {sql}", params).fetchall() == [("a_b",)] + + +# ----------------------------------------------------------------- the tables + + +def test_every_region_has_names_and_towns() -> None: + assert set(REGIONS) == {"skane", "hovedstaden", "zealand", "oresund", "jutland", "funen"} + for key, region in REGIONS.items(): + assert region.key == key and region.label and region.names and region.towns + + +def test_the_bare_word_zealand_is_no_name_since_new_zealand_has_it() -> None: + assert "zealand" not in places._PHRASES["zealand"] + + +def test_the_capital_and_scanian_towns_are_in_their_regions() -> None: + skane = {fold(t) for t in REGIONS["skane"].towns} + assert {"malmo", "lund", "helsingborg", "kristianstad", "landskrona", "trelleborg", + "angelholm", "hassleholm", "ystad", "eslov", "hoganas", "staffanstorp", "lomma", + "bjuv", "burlov", "kavlinge", "svedala", "vellinge"} <= skane + hoved = {fold(t) for t in REGIONS["hovedstaden"].towns} + assert {"kobenhavn", "kgs lyngby", "soborg", "ballerup", "malov", "bagsvaerd", "humlebaek", + "smorum", "nordhavn", "frederiksberg", "hillerod", "helsingor"} <= hoved + + +# ----------------------------------------------------------------- suggestions + +PLACES = [ + ("København", "DK", 69), ("Copenhagen, Denmark", "DK", 56), ("København K", "DK", 35), + ("København Ø", "DK", 34), ("Kgs. Lyngby", "DK", 16), ("Billund, Denmark", "DK", 18), + ("Aarhus N, Region Central Jutland, Denmark", "DK", 26), ("Albertslund", "DK", 10), + ("Denmark", "DK", 27), ("Remote OR Copenhagen", "DK", 1), + ("Malmö", "SE", 214), ("Lund", "SE", 127), ("Sweden - Lund", "SE", 29), + ("Malmö, Skåne County, Sweden", "SE", 3), ("Stockholm, Sweden", "SE", 287), + ("Aarhus, Denmark; København, Denmark; Stockholm, Sweden", "EU", 1), + ("Remote", "", 40), ("Hybrid", "", 12), ("Europe", "EU", 20), ("", "", 5), + ("Berlin, Germany", "DE", 100), ("Auckland, New Zealand", "NZ", 25), + ("The Hague, Netherlands", "NL", 9), ("Den Haag", "NL", 4), +] + + +def tokens(q: str, **kw) -> list[str]: + return [s.token for s in suggest(q, PLACES, **kw)] + + +def test_skane_is_suggested_however_it_is_typed() -> None: + for q in ("Skåne", "skane", "SKANE", "ska", "Scania", "scan"): + assert "region:skane" in tokens(q), q + hit = next(s for s in suggest("skane", PLACES) if s.token == "region:skane") + assert hit == Suggestion("region:skane", "Skåne", "region", 214 + 127 + 29 + 3) + + +def test_a_country_counts_the_roles_that_carry_its_code() -> None: + hit = suggest("den", PLACES)[0] + assert hit == Suggestion("country:DK", "Denmark", "country", + 69 + 56 + 35 + 34 + 16 + 18 + 26 + 10 + 27 + 1) + assert tokens("swe")[0] == "country:SE" + assert tokens("sverige")[0] == "country:SE" + assert tokens("danmark")[0] == "country:DK" + + +def test_countries_come_first_then_regions_then_towns() -> None: + kinds = [s.kind for s in suggest("den", PLACES)] + assert kinds == sorted(kinds, key=["country", "region", "place"].index) + assert kinds[0] == "country" and "region" in kinds and kinds[-1] == "place" + + +def test_towns_are_grouped_and_labelled_by_their_commonest_spelling() -> None: + lund = [s for s in suggest("lund", PLACES) if s.kind == "place"] + assert lund == [Suggestion("place:lund", "Lund", "place", 127 + 29)] + kbh = [s for s in suggest("kob", PLACES) if s.kind == "place"] + assert kbh[0].label == "København" and kbh[0].token == "place:kobenhavn" + + +def test_a_postal_district_is_not_a_town_of_its_own() -> None: + labels = [s.label for s in suggest("kob", PLACES) if s.kind == "place"] + assert "København K" not in labels and "København Ø" not in labels + assert [s.label for s in suggest("aarhus", PLACES) if s.kind == "place"] == ["Aarhus"] + + +def test_spellings_of_one_town_are_one_suggestion() -> None: + towns = [s for s in suggest("cop", PLACES) if s.kind == "place"] + assert len(towns) == 1 + # København, K and Ø, the one in a list of three towns; Copenhagen, and "Remote OR Copenhagen". + assert towns[0].label == "København" and towns[0].count == 69 + 35 + 34 + 1 + 56 + 1 + assert [s.token for s in suggest("kobenhavn", PLACES) if s.kind == "place"] \ + == [towns[0].token] + + +def test_only_the_first_comma_part_of_each_semicolon_part_is_a_town() -> None: + labels = {s.label for q in "abcdefghijklmnopqrstuvwxyz" + for s in suggest(q, PLACES, limit=999) if s.kind == "place"} + assert {"Aarhus", "Stockholm", "Malmö", "Lund", "Berlin"} <= labels + assert not {"Denmark", "Sweden", "Region Central Jutland", "Remote", "Hybrid", "Europe"} \ + & labels + + +def test_remote_hybrid_and_country_names_are_no_town() -> None: + for q in ("rem", "hyb", "eur", "denm"): + assert [s for s in suggest(q, PLACES) if s.kind == "place"] == [], q + + +def test_a_word_prefix_finds_a_multi_word_town() -> None: + assert tokens("haag") == tokens("hague") == tokens("den haag") == tokens("the hag") \ + == ["place:the hague"] + assert tokens("kgs ly") == ["place:kgs lyngby"] + assert tokens("lyngby") == ["place:kgs lyngby"] + + +def test_the_word_must_start_not_end() -> None: + assert "place:lund" not in tokens("und") + assert "place:billund" not in tokens("lund") + assert "place:albertslund" not in tokens("lund") + + +def test_new_zealand_is_a_country_and_zealand_a_region_only_when_it_has_roles() -> None: + assert tokens("zea")[:1] == ["country:NZ"] + assert "region:zealand" not in tokens("zea") # no Zealand role in PLACES + assert "country:AU" not in tokens("aus") # no role of its own + + +def test_limit_and_the_empty_query() -> None: + assert suggest("", PLACES) == [] and suggest(" ", PLACES) == [] and suggest("---", PLACES) == [] + assert len(suggest("a", PLACES, limit=3)) == 3 + assert len(suggest("a", PLACES)) <= 12 + + +def test_towns_are_ordered_by_roles() -> None: + towns = [s for s in suggest("l", PLACES) if s.kind == "place"] + assert [s.count for s in towns] == sorted((s.count for s in towns), reverse=True) + + +def test_suggest_takes_any_iterable() -> None: + assert suggest("lund", iter(PLACES)) == suggest("lund", PLACES) + + +def test_a_suggested_token_finds_the_roles_it_counted(table: sqlite3.Connection) -> None: + for s in suggest("kob", PLACES) + suggest("lund", PLACES) + suggest("skane", PLACES): + assert parse(s.token).kind == s.kind + assert found(table, s.token), s.token + assert label(s.token) == s.label or s.kind == "place"