diff --git a/app/places.py b/app/places.py index ae564a6..0505189 100644 --- a/app/places.py +++ b/app/places.py @@ -13,23 +13,29 @@ of three tokens, and several tokens are joined by OR: 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`; +`matches` answers whether one role's folded place and country fit a token, in +Python: a region is over a hundred phrases, and as `LIKE` clauses that costs +several hundred milliseconds a query, as a set intersection on the words of a +line it costs a few. `register` gives a connection that answer as the SQL +function `place_match`, and `match_sql` returns the clause that calls it. `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. +chip's text back from the address. This module opens no database; the caller +hands over the places it holds, or its connection. """ from __future__ import annotations import re +import sqlite3 from collections import Counter from collections.abc import Iterable from dataclasses import dataclass +from functools import lru_cache from app.models import _CITIES, _COUNTRY_NAMES, _REGIONS, _place_fold -__all__ = ["REGIONS", "Region", "Suggestion", "Token", "fold", "label", "match_sql", - "parse", "suggest"] +__all__ = ["REGIONS", "SQL_FUNCTION", "Region", "Suggestion", "Token", "fold", "known", + "label", "match_sql", "matches", "parse", "register", "suggest"] def fold(text: str) -> str: @@ -202,14 +208,16 @@ def _expand(text: str) -> tuple[str, ...]: 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.""" + has one of either as whole words, or is in a region it `includes`, by key. + Only `names` are what the region is suggested and read as. Spellings are + shown with their marks and folded when used.""" key: str label: str names: tuple[str, ...] towns: tuple[str, ...] country: str = "" + includes: tuple[str, ...] = () def _split(text: str) -> tuple[str, ...]: @@ -281,12 +289,12 @@ _FUNEN = Region( Ærø, Ærøskøbing, Langeland, Rudkøbing, Tåsinge, Marstal, Nordfyn"""), "DK") -# The Øresund region is the two capital regions either side of the sound. +# The Øresund region is the two capital regions either side of the sound. It is +# suggested on its own names only; a line in either region is in it. _ORESUND = Region( "oresund", "Øresund", - _split("Øresund, Öresund, Oresund, Oresund Region, Øresundsregionen, Öresundsregionen") - + _HOVEDSTADEN.names + _SKANE.names, - _HOVEDSTADEN.towns + _SKANE.towns) + _split("Øresund, Öresund, Oresund, Oresund Region, Øresundsregionen, Öresundsregionen"), + (), includes=("hovedstaden", "skane")) REGIONS: dict[str, Region] = { r.key: r for r in (_SKANE, _HOVEDSTADEN, _ZEALAND, _ORESUND, _JUTLAND, _FUNEN)} @@ -294,11 +302,13 @@ REGIONS: dict[str, Region] = { def _phrases(region: Region) -> tuple[str, ...]: """Every folded phrase that puts a line in the region, with the other - spellings of its towns.""" + spellings of its towns, and those of the regions it includes.""" found: dict[str, None] = {} for text in (*region.names, *region.towns): for form in _expand(fold(text)): found[form] = None + for key in region.includes: + found.update(dict.fromkeys(_phrases(REGIONS[key]))) return tuple(found) @@ -321,6 +331,24 @@ for _region in REGIONS.values(): _NOT_A_TOWN |= frozenset(_REGION_BY_NAME) +def _group(folded: str) -> str: + """The key of the town a fold is: the first spelling of its alias group, else + the fold with `aa` as `a`.""" + return _GROUPS[folded][0] if folded in _GROUPS else _variants(folded)[-1] + + +def _known_towns() -> dict[str, str]: + """Every town the tables name, as a fold, to its group: the alias groups, the + towns of the regions and the cities `country_of` knows.""" + names = [*_GROUPS, *(fold(city) for city in _CITIES), + *(fold(town) for region in REGIONS.values() for town in region.towns)] + return {form: _group(form) for name in names for form in _variants(name)} + + +_KNOWN = _known_towns() +_KNOWN_LONGEST = max(len(name.split()) for name in _KNOWN) + + # --------------------------------------------------------------------- tokens @@ -369,57 +397,118 @@ def label(token: str | Token) -> str: return (token if isinstance(token, Token) else parse(token)).label -# ------------------------------------------------------------------------ SQL +# ------------------------------------------------------------------- matching -_NEVER = ("0", []) +SQL_FUNCTION = "place_match" -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 known(token: str) -> bool: + """Whether the token names something: a country of the table (or `EU`), a + region of `REGIONS`, a place with a value. `region:atlantis`, `country:ZZ` + and `place:` are not.""" + return _known(parse(token)) -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 _known(tok: Token) -> bool: + if tok.kind == "country": + return tok.value in _COUNTRY_LABELS or tok.value == "EU" + if tok.kind == "region": + return tok.value in REGIONS + return bool(tok.value) -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] +@dataclass(frozen=True) +class _Matcher: + """What a token asks of a role. `phrases` are whole-word phrases of the + folded place; `countries` are `(code, mentions)`: the role's own country, or + a line with one of the mentions while `country` is empty or `EU`, since a + board listing several offices has no single country.""" + + phrases: frozenset[str] + countries: tuple[tuple[str, frozenset[str]], ...] + longest: int + + +@lru_cache(maxsize=1024) +def _matcher(token: str) -> _Matcher | None: + """The token read once: `None` when it names nothing. + + 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.""" + tok = parse(token) + if not _known(tok): + return None + phrases: set[str] = set() + countries: list[tuple[str, frozenset[str]]] = [] + if tok.kind == "country": + countries.append((tok.value, frozenset(_COUNTRY_MENTIONS.get(tok.value, ())))) + elif tok.kind == "region": + phrases.update(_PHRASES[tok.value]) + else: + phrases.update(_expand(tok.value)) + phrases.update(_PHRASES.get(_REGION_BY_NAME.get(tok.value, ""), ())) + if code := _COUNTRY_BY_NAME.get(tok.value): + countries.append((code, frozenset(_COUNTRY_MENTIONS.get(code, ())))) + phrases.discard("") + longest = max((len(p.split()) for p in (*phrases, *(m for _, ms in countries for m in ms))), + default=1) + return _Matcher(frozenset(phrases), tuple(countries), longest) + + +@lru_cache(maxsize=32768) +def _runs(place_fold: str, longest: int) -> frozenset[str]: + """Every run of up to `longest` words of a folded line, each a whole-word + phrase the line holds.""" + words = place_fold.split() + return frozenset(" ".join(words[i:i + n]) + for n in range(1, longest + 1) for i in range(len(words) - n + 1)) + + +def matches(place_fold: str, country: str, token: str) -> bool: + """Whether a role whose folded place and country are these is found by the + token. `country:DK` is `country == 'DK'`, or a line that names Denmark (or + a Danish city) while `country` is empty or `EU`. `region:skane` is any of + its names or towns as whole words, `place:malmo` the town and its aliases. + A token that names nothing (`known` is false) matches no role.""" + match = _matcher(token) + if match is None: + return False + runs = None + if match.phrases: + runs = _runs(place_fold, match.longest) + if not match.phrases.isdisjoint(runs): + return True + for code, mentions in match.countries: + if country == code: + return True + if mentions and country in ("", "EU"): + runs = _runs(place_fold, match.longest) if runs is None else runs + if not mentions.isdisjoint(runs): + return True + return False + + +def _place_match(place_fold: str | None, country: str | None, token: str) -> int: + return int(matches(place_fold or "", country or "", token)) + + +def register(conn: sqlite3.Connection) -> None: + """Give a connection `place_match(place_fold, country, token)`, which is 1 when + `matches` is true. `match_sql` returns a clause that calls it.""" + conn.create_function(SQL_FUNCTION, 3, _place_match, deterministic=True) 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.""" + """SQL over the two columns for one token, with one `?` parameter. It calls + `place_match`, so the connection needs `register`. Several tokens are joined + with OR by the caller. Bare text is a place. A token that names nothing + (`known` is false) is `0`, which 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 + if not _known(tok): + return "0", [] + return f"{SQL_FUNCTION}({place}, {country}, ?)", [str(tok)] # ---------------------------------------------------------------- suggestions @@ -442,9 +531,9 @@ def _prefixes(q: str, candidates: Iterable[str]) -> bool: 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) +#: Where one town ends in a line: `;` and `|` 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) @@ -471,16 +560,32 @@ def _towns(place: str) -> list[str]: return found +def _towns_in(folded: str) -> list[str]: + """The known towns a fold holds as whole words, longest first from the left, + as their groups: `munich berlin` is Munich and Berlin, `copenhagen office` is + Copenhagen, `hoje taastrup` is one town and not also Taastrup.""" + words = folded.split() + found: list[str] = [] + i = 0 + while i < len(words): + for n in range(min(_KNOWN_LONGEST, len(words) - i), 0, -1): + if (group := _KNOWN.get(" ".join(words[i:i + n]))) is not None: + found.append(group) + i += n + break + else: + i += 1 + return list(dict.fromkeys(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) + sets = {key: frozenset(_PHRASES[key]) for key in keys} + longest = max((len(p.split()) for phrases in sets.values() for p in phrases), 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)} + runs = _runs(text, longest) for key, phrases in sets.items(): if not phrases.isdisjoint(runs): counts[key] += roles @@ -516,20 +621,32 @@ def suggest(q: str, places: Iterable[tuple[str, str, int]], limit: int = 12) \ regions = [Suggestion(f"region:{key}", REGIONS[key].label, "region", counts[key]) for key in named if counts[key]] + # A line's town is a group. A line that is a town is that group's spelling; + # one that only holds a known town ("Copenhagen Office", "Munich - Berlin") + # counts under each such town and shows as none; anything else is its own. + counts_by_town: Counter[str] = Counter() 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] + if folded not in _KNOWN and (inside := _towns_in(folded)): + for group in inside: + counts_by_town[group] += roles + continue + group = _group(folded) + counts_by_town[group] += roles 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)]): + for group, total in counts_by_town.items(): + forms = {group, *written.get(group, ())} + if not _prefixes(text, [form for folded in forms 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()))) + seen = spellings.get(group) + shown = min(seen, key=lambda s: (-seen[s], s)) if seen \ + else _TOWN_LABELS.get(group) or group.title() + towns.append(Suggestion(f"place:{fold(shown)}", shown, "place", total)) countries.sort(key=lambda s: (-s.count, s.label)) regions.sort(key=lambda s: (-s.count, s.label)) diff --git a/tests/test_places.py b/tests/test_places.py index d1f83df..98e3395 100644 --- a/tests/test_places.py +++ b/tests/test_places.py @@ -3,6 +3,7 @@ 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`.""" +# ruff: noqa: RUF001 (real place strings, with the letters boards write) from __future__ import annotations import sqlite3 @@ -10,7 +11,8 @@ import sqlite3 import pytest from app import places -from app.places import REGIONS, Suggestion, Token, fold, label, match_sql, parse, suggest +from app.places import (REGIONS, Suggestion, Token, fold, known, label, match_sql, matches, + parse, register, suggest) #: (place, country) as a board wrote them and `country_of` read them. ROWS = [ @@ -37,9 +39,15 @@ ROWS = [ ] +def connect() -> sqlite3.Connection: + db = sqlite3.connect(":memory:") + register(db) + return db + + @pytest.fixture() def table() -> sqlite3.Connection: - db = sqlite3.connect(":memory:") + db = connect() 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]) @@ -153,7 +161,7 @@ def test_copenhagen_is_every_spelling(table: sqlite3.Connection, spelling: str) @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 = connect() 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]) @@ -223,7 +231,7 @@ def test_a_country_is_also_a_line_that_names_it_when_country_is_unset( def test_a_country_name_does_not_reach_a_role_of_another_country() -> None: - db = sqlite3.connect(":memory:") + db = connect() 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"), @@ -261,6 +269,15 @@ def test_jutland_and_funen(table: sqlite3.Connection) -> None: assert found(table, "region:funen") == {"Odense, REGION_DNK 1852 South Denmark, Denmark"} +def test_oresund_finds_a_line_of_either_region_but_is_named_by_its_own_words() -> None: + assert matches("lund sweden", "SE", "region:oresund") + assert matches("kobenhavn k", "DK", "region:oresund") + assert not matches("aarhus", "DK", "region:oresund") + assert "skane" not in {fold(n) for n in REGIONS["oresund"].names} + assert REGIONS["oresund"].includes == ("hovedstaden", "skane") + assert parse("region:Capital Region of Denmark").value == "hovedstaden" + + 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")) @@ -279,27 +296,28 @@ def test_a_token_that_names_nothing_finds_nothing(table: sqlite3.Connection) -> assert found(table, "place:nowhere-in-particular") == set() -def test_the_columns_are_the_callers(table: sqlite3.Connection) -> None: +def test_the_columns_are_the_callers() -> 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:") + assert sql == "place_match(r.pf, r.c, ?)" and params == ["country:DK"] + db = connect() 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"): +def test_sql_has_one_parameter_per_question_mark_and_a_token_is_normalised() -> None: + for token in ("country:DK", "region:oresund", "place:copenhagen", "place:denmark", "Skåne"): sql, params = match_sql(token) - assert sql.startswith("(") and sql.endswith(")") - assert sql.count("?") == len(params), token + assert sql.startswith("place_match(") and sql.count("?") == len(params) == 1, token + assert match_sql("country:dk")[1] == ["country:DK"] + assert match_sql("Kgs. Lyngby")[1] == ["place:kgs lyngby"] + for token in ("region:atlantis", "country:ZZ", "place:", ""): + assert match_sql(token) == ("0", []), token def test_a_wildcard_in_a_value_is_a_letter() -> None: - assert places._escape("50%_off\\") == "50\\%\\_off\\\\" - db = sqlite3.connect(":memory:") + db = connect() 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")]) @@ -307,9 +325,26 @@ def test_a_wildcard_in_a_value_is_a_letter() -> None: 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",)] + + +def test_known_is_false_for_a_token_that_names_nothing() -> None: + for token in ("region:atlantis", "country:ZZ", "place:", "region:", "country:", "", ":"): + assert not known(token), token + for token in ("country:DK", "country:EU", "country:dk", "region:skane", "region:Skåne", + "place:nowhere-in-particular", "Skåne", "country:Denmark"): + assert known(token), token + + +def test_matches_is_false_for_a_token_that_names_nothing() -> None: + assert not matches("copenhagen", "DK", "region:atlantis") + assert not matches("copenhagen", "ZZ", "country:ZZ") + assert not matches("copenhagen", "DK", "place:") + + +def test_null_columns_match_nothing_and_do_not_raise() -> None: + db = connect() + assert db.execute("SELECT place_match(NULL, NULL, 'place:lund')").fetchone() == (0,) + assert db.execute("SELECT place_match(NULL, 'DK', 'country:DK')").fetchone() == (1,) # ----------------------------------------------------------------- the tables @@ -318,7 +353,8 @@ def test_a_wildcard_in_a_value_is_a_letter() -> None: 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 + assert region.key == key and region.label and region.names + assert region.towns or region.includes def test_the_bare_word_zealand_is_no_name_since_new_zealand_has_it() -> None: @@ -348,6 +384,11 @@ PLACES = [ ("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), + ("Copenhagen Office", "DK", 16), ("Copenhagen Hub - Løvstræde", "DK", 14), + ("DK-Copenhagen-MSO", "DK", 2), + ("Munich", "DE", 40), ("München", "DE", 10), ("Munich Office", "DE", 5), + ("Munich (DE)", "DE", 3), ("Munich - Berlin", "DE", 2), + ("Garching bei München (Munich)", "DE", 4), ("Kalundborg, Region Zealand, Denmark", "DK", 14), ] @@ -365,7 +406,8 @@ def test_skane_is_suggested_however_it_is_typed() -> None: 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) + 69 + 56 + 35 + 34 + 16 + 18 + 26 + 10 + 27 + 1 + + 16 + 14 + 2 + 14) assert tokens("swe")[0] == "country:SE" assert tokens("sverige")[0] == "country:SE" assert tokens("danmark")[0] == "country:DK" @@ -394,7 +436,7 @@ 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 towns[0].label == "København" and towns[0].count == 69 + 35 + 34 + 1 + 56 + 1 + 16 + 14 + 2 assert [s.token for s in suggest("kobenhavn", PLACES) if s.kind == "place"] \ == [towns[0].token] @@ -425,9 +467,10 @@ def test_the_word_must_start_not_end() -> None: assert "place:albertslund" not in tokens("lund") -def test_new_zealand_is_a_country_and_zealand_a_region_only_when_it_has_roles() -> None: +def test_a_country_or_region_with_no_open_role_is_not_suggested() -> None: assert tokens("zea")[:1] == ["country:NZ"] - assert "region:zealand" not in tokens("zea") # no Zealand role in PLACES + assert "region:zealand" in tokens("zea") # Kalundborg + assert "region:funen" not in tokens("fun") # no Funen role in PLACES assert "country:AU" not in tokens("aus") # no role of its own @@ -451,3 +494,704 @@ def test_a_suggested_token_finds_the_roles_it_counted(table: sqlite3.Connection) assert parse(s.token).kind == s.kind assert found(table, s.token), s.token assert label(s.token) == s.label or s.kind == "place" + + +# ------------------------------------------------- the function against LIKE + +#: `countryplace` of open roles on the live database, distinct: a random +#: sample plus, for each token below, lines that hold it. Some lines have no country. +REAL = """ +NL 's-Hertogenbosch, Netherlands + 3 Locations + 5 Locations +DK Aabenraa eller København +DK Aalborg, Copenhagen, Kolding, Aarhus, Hybrid position +DK Aalborg, Denmark; Aarhus, Denmark; Copenhagen, Denmark; Odense, Denmark; Søborg, Denmark +DK Aarhus C, Denmark +DK Aarhus N, Denmark +DK Aarhus or Valby +EU Aarhus, Denmark; København, Denmark; Stockholm, Sweden + Aix-en-Provence +DK Allerød or Stockholm (SE) +DK Allerød, Copenhagen, or Stockholm, Hybrid position +KZ Almaty, Almatı oblısı, Kazakhstan +EU Amsterdam, Netherlands; Berlin, Germany; København, Denmark; Vilnius, Lithuania +EU Amsterdam, Netherlands; Berlin, Germany; Munich, Germany; Remote, Germany +EU Amsterdam, Netherlands; Manchester, United Kingdom +EU Amsterdam; Germany +FR Angers, France +DE Ansbach, Mittelfranken, Germany +MX Apaseo el Grande, GUANAJUATO, Mexico +US Arizona; California; Washington +NL Arnhem, GELDERLAND, Netherlands +BE Asse, , Belgium + Athens, Greece; Cyprus; Frankfurt am Main, Germany; Munich, Germany; Tel Aviv, Israel + Athens, Greece; London, United Kingdom; Tel Aviv, Israel + Auckland +US Austin; New York City; Palo Alto +AU Australian Capital Territory, Australia +DK Bad Homburg; Copenhagen +DK Bagsværd +DK Ballerup or Aarhus V +DK Ballerup, , Denmark +DK Ballerup, Ballerup +DK Ballerup, Denmark +IN Bangalore, KA, India +CN Bao Ding, Hebei, China +ES Barcelona, Catalonia +EU Barcelona, Spain; Stockholm, Stockholm County, Sweden +EU Barcelona; Berlin; London; Munich; Paris +EU Barcelona; Geneva +DE Barsinghausen, Germany +PL Bartoszyce, Warminsko-mazurskie, Poland +MY Bayan lepas, Pulau Pinang, Malaysia +EU Belgium; Netherlands +RS Belgrade - Savski Venac, Serbia +EU Belgrade, Serbia; Berlin, Germany; Munich, Germany +US Bellevue, United States of America +US Bellevue, WA, United States +US Bellevue, Washington; Denver, Colorado; Seattle, Washington; Washington, D.C. +IN Bengaluru, India +RS Beograd, , Serbia +NO Bergen, , Norway +NL Berkel en Rodenrijs, ROTTERDAM, Netherlands +DE Berlin Office; Munich Office +DE Berlin Office; Munich Office; Remotely in Germany +DE Berlin, Berlin, Germany +DE Berlin, Germany +EU Berlin, Germany; Amsterdam, Netherlands; Copenhagen, Denmark +EU Berlin, Germany; Copenhagen, Denmark +EU Berlin, Germany; London, United Kingdom +EU Berlin, Germany; London, United Kingdom; Stockholm, Sweden +EU Berlin, Germany; Munich, Germany; Prague, Czech Republic +EU Berlin, Germany; Paris, France +DE Berlin; Cologne; Frankfurt; Hamburg; Munich; Stuttgart +DE Berlin; Frankfurt; Munich +DE Berlin; Frankfurt; Munich; Stuttgart +EU Berlin; London; Munich +EU Berlin; London; Munich; Paris +EU Berlin; London; Munich; Paris; Stockholm; Tallinn +DE Berlin; Munich +CH Bern, Switzerland +NL Best, Noord-Brabant, Netherlands +DE Bietigheim-Bissingen, Baden-Württemberg, Germany +DK Billund Kornmarken, Denmark +DK Billund, Denmark +DK Billund, København +DK Billund, København K +DK Billund; Copenhagen Hub - Løvstræde +EU Billund; London; Copenhagen Hub - Løvstræde +DK Birkerød +SE Bjuv +DK Bjæverskov +DE Bochum, NW, Germany +DE Bonn, Germany +FR Bordeaux +JP Bosch Corporation_Shibuya_BGSJ, Tokyo, Japan +US Boston, MA, United States +US Bozeman, MT, United States +DK Brabrand, Denmark + Brest +US Bridgewater, United States of America +FR Briey, France +GB Bristol, UK; London, UK +SI Brnik - Aerodrom, , Slovenia +DK Bronshoj, Denmark +BE Bruxelles, Bruxelles, Belgium +DK Brøndby +DK Brøndby eller Fredericia +DK Brøndby eller Fredericia og mulighed for hjemmearbejde +US Burlington, United States + Busan +DE Butzbach, HE, Germany +RO Buzau, Com Caragele, Constanta, Romania +SE Båstad +DE Böblingen, Baden-Württemberg, Germany +CA CA-Toronto +AU Camberwell, Australia +GB Carlisle, England, United Kingdom +DE Carlsfelder Str. 3, 06188 Land, Germany +DE Castroper Hellweg 109, 44805 B, Germany +MX Celaya Planta, GUANAJUATO, Mexico +US Cement, United States + Chandigarh +CN Chengdu, Sichuan, China; XiAn, Shaanxi, China +FR Cherbourg-en-Cotentin, Manche, France +US Chicago, Atlanta, US-Remote, Toronto +CL Chile + China - Hong Kong; Taiwan - Taipei +MX Ciudad de México, CDMX, Mexico +MX Colon, QUERETARO, Mexico +DK Copenhagen +DK Copenhagen Hub - Løvstræde, Denmark +DK Copenhagen Hub - Løvstræde; Billund +DK Copenhagen Hub - Østergade, Denmark +DK Copenhagen K, Denmark; Aarhus C, Denmark +DK Copenhagen O, Denmark +DK Copenhagen Office +DK Copenhagen S, Region Capital of Denmark, Denmark +DK Copenhagen V, Denmark +DK Copenhagen V, Denmark; Aarhus C, Denmark +DK Copenhagen or Aarhus + Copenhagen, Denmark; Tel Aviv, Israel +DK Copenhagen, Hovedstaden, Denmark +DK Copenhagen; Billund LEGO Campus + Copenhagen; Zurich; Singapore; Warsaw; Toronto +AR Corrientes, , Argentina + Corte Madera +NL Culemborg, Netherlands + Curitiba +PE Cusco, Cusco, Peru +DE DE Store Hamburg Spitalerstraße, Germany +US Dallas, United States +AU Dandenong South, VICTORIA, Australia +GB Darlington, England, United Kingdom +AU Deakin ACT, Australia +EU Denmark; Vilnius, Lithuania +DK Dianalund +CN Dongguan, Guangdong, China +DE Dortmund, Germany +DE Dortmund, North Rhineland Westphalia, DE +1 more… +IE Dublin, CA (ZICC) +EU Dublin, Ireland; London +GB Dublin, United Kingdom + Dunkerque + Durban +EU EMEA; Germany; London, United Kingdom; Paris, France; Remote - Netherlands +DE Ergolding, BY, Germany +DK Esbjerg, Region South Denmark, Denmark +DE Eschborn, HE, Germany +DE Eschenburg, HE, Germany + Ezhou +IT Fagagna, , Italy +GB Farnborough, Hampshire, United Kingdom +DK Farum +DK Finland, København, Litauen, Polen, Sverige +EU Finland; France; Germany; Netherlands; Remote - Europe; Sweden; United Kingdom +EU Finland; Remote - Denmark; Stockholm, Sweden + Finland; United Kingdom; Korea, Republic of; Denmark; France; India; Poland; United States + Finland; United Kingdom; United States; Belgium +DE Fins Only-DE-Munich-MSO +EU Fleet Place Office; Amsterdam Office; Berlin Office + France; India; Finland +DE Frankfurt; Berlin; Munich +DE Frankfurt; Munich +DK Fredericia or Ballerup +DK Fredericia, REGION_DNK 1852 South Denmark, Denmark +DK Frederiksberg +DK Frederikssund og mulighed for hjemmearbejde +DK Fårevejle +GB GB-London +US Garden City, TX, United States +EU Geneva; London +DE Gerlingen, Baden-Württemberg, Germany +DE Germany Scoober Office +EU Germany, Remote; The Netherlands, Remote + Germany; Portugal; Canada; United States +DK Glostrup +DK Glostrup, Denmark +PL Goleniów, Województwo zachodniopomorskie, Poland +US Greenville, South Carolina, United States, South Carolina, United States of America +PL Grodzisk Mazowiecki (mazowieckie), Poland +SE Göteborg, Västra Götaland, Sweden; Lund, Skåne, Sweden +AT Haag, Austria +DE Hannover, NDS, Germany +VN Hanoi, , Vietnam +DK Headquarters, Denmark +US Heber, AZ, United States +SE Helsingborg, Sweden +FI Helsinki Metropolitan Area +FI Helsinki, Finland +EU Helsinki, Finland; Stockholm, Sweden; Tallinn, Estonia +DK Herlev og mulighed for hjemmearbejde +DE Herzogenrath, North Rhine Westphalia, Germany +DE Hildesheim, NDS, Germany +DK Hillerød +DK Hillerød, Capital Region of Denmark, Denmark +DE Hoher Hagen 3, 58513 Lüdensche, Germany +DK Holbæk +DE Homburg, , Germany +DE Homburg/Saar, SL, Germany +EU Home Based - Americas; Home Based - APAC; Home based - EMEA; Home based - Worldwide +GB Home Based - Americas; Home based - EMEA; Office Based - London, UK +MX Home-based, Mexico +DK Hvalsø +DK Hvidovre + Hwaseong-si + Hybrid +GB Hybrid - London, England +IN Hyderabad; Noida +SE Härryda, Sweden +SE Hässleholm +SE Höör +JP Ibaraki, Osaka, Japan +US Illinois, USA, Remote; Minnesota, USA, Remote +DE Immenstadt im Allgäu, BY, Germany + Indija +IT Induno Olona, Italy + Israel; Remote - Europe; United Kingdom +DE Jacob-A.-Morand-Str. 18, 07552, Germany +JP Japan, Tokyo + Jena +CZ Jihlava 1, Kraj Vysočina, Czechia +NL Joure, Netherlands +DK Kalundborg, Region Zealand, Denmark +DK Kastrup +SE Katrineholm, Sweden +DE Kelsterbach, HE, Germany + Kempten im Allgäu +IL Kfar Saba, Israel + Konstanz +SE Kristianstad, Sweden +MY Kuala Lumpur – Co-labs, Kuala Lumpur – Co-labs, Malaysia +MY Kuala Lumpur, Malaysia +SE Kungsängen, Sweden +BE Kuurne, Belgium +DK København +DK København K +DK København K eller Odense C +DK København K og mulighed for hjemmearbejde +DK København NV +DK København S +DK København SV eller Aalborg og mulighed for hjemmearbejde +DK København SV, Odense C +DK København V eller Korsør +DK København V og mulighed for hjemmearbejde +DK København Ø og Frederiksberg C samt mulighed for hjemmearbejde +DK København Ø, Denmark +EU København, Capital RegionDenmark, Denmark; Moss, Østfold, Norway; Oslo, Oslo, Norway +DK København, Denmark +DK København, Denmark; Aarhus, Denmark; Aalborg, Denmark; Kolding, Denmark +DK København, Denmark; Aarhus, Denmark; Kolding, Denmark; Aalborg, Denmark +EU København, Denmark; Stockholm, Sweden +DK København, København SV +DK København, København SV, Odense, Odense C +DK København, København SV, Viby J, Aarhus +DK København, Roskilde +DK Køge +DK Køge, Denmark +ES La Muela (Zaragoza), Z, Spain + La Roche-sur-Yon +US Lake Zurich, Illinois, United States +SE Landskrona +EU Lausanne; London +DK Lem St, Denmark +DK Lem St, Region Central Jutland, Denmark +NL Lemelerveld, Netherlands +MX León, GUA, Mexico +IT Liguria, Italy +DE Lohr am Main, BY, Germany +SE Lomma + London, Canada; Remote, United States; Oak Brook, United States +GB London, London, United Kingdom + London, Paris, Toronto, New York +EU London, United Kingdom; Oslo, Norway + London, United Kingdom; Remote, United States +GB London; Billund Project House +EU London; Copenhagen +GB London; Remote (UK) +EU London; Warsaw; Copenhagen +EU London; Zurich +SE Luleå, Sweden +SE Lund, Skåne län, Sweden +SE Lund, Skåne, Sweden +US Madison, Wisconsin, United States of America +SE Malmo, Sweden +SE Malmö +SE Malmö, Skåne County, Sweden +SE Malmö, Skåne Län, Sweden +SE Malmö, Sweden +DK Maribo +DE Meschede, Nordrhein-Westfalen, Germany + Metz +US Miami +US Miami, Florida, United States +DK Middelfart, Denmark +AU Midland, Australia +IT Milan Office +EU Milan, Italy; Paris, Paris, France +US Minneapolis, Minnesota, United States, Minnesota, United States of America +US Minneola, KS, United States + Monterrey +MX Monterrey, Mexico +US Mountain View, CA, USA; San Francisco, CA, USA +US Mountain View, California +US Mountain View, California; Plano, Texas; San Francisco, California +DE Munich +EU Munich - Berlin - London - Paris +EU Munich - Berlin - London - Paris - Warsaw - Tallinn +EU Munich - Berlin - Paris +DE Munich, Bavaria, Germany +DK Munkebo, Denmark +LU Munsbach, Luxembourg + Mátészalka + Möhnesee +DE München +DE München (DE) +DE München, BAYERN, Germany +DE München, BY, Germany +DE München, Germany +US NYC +US NYC, SF, Chi, Remote + Nagpur +DK Nakskov, Denmark +DK Nakskov, Region Zealand, Denmark + Nanjing +JP Nasushiobara, Tochigi, Japan +GB New Britain + New York City, Toronto, Chicago, or Remote +US New York, San Francisco, Seattle, Atlanta, Chicago + New York; Hong Kong; London; Singapore; Sydney +RS Niš, Serbia +DK Nordhavn eller Aalborg og mulighed for hjemmearbejde +DK Nordhavn eller Viby J +DK Nordhavn or Åbyhøj +NO Norway +AU Nundah, Australia +SE Nybro, Sweden +DK Nykøbing F +DK Nykøbing F, Denmark +DK Næstved +DK Nørre Asmindrup +DE Nürnberg +CA ON, Toronto, Canada +DE Oberkochen; München +PL Oborniki, Poland +DK Odense C, Denmark +DK Odense M, Denmark +DK Odense, Denmark +DK Odense, REGION_DNK 1852 South Denmark, Denmark +GB Office Based - London, UK +TW Office Based - Taipei, Taiwan +US Oklahoma, United States +CA Ontario - Remote +US Ontario, CA, United States of America +US Orange County , California, United States +SE Oskarshamn, Sweden +EU Paris, France; Copenhagen, Denmark; Amsterdam, Netherlands; Berlin, Germany +FR Paris, France; Tassin-la-Demi-Lune, France; Bordeaux, France +EU Paris;Geneva +US Paw Paw, IL, United States +FR Perigny, France +FR Perpignan, Pyrénées-Orientales, France +SE Perstorp +US Phillipsburg, New Jersey, United States +DE Pinkertweg 16, 22113 Hamburg, Germany +MX Planta Celaya, , Mexico +EU Poland; Hungary +PT Porto, Portugal +EU Porto, Portugal; Madrid, Spain; Berlin, Germany +CZ Prague +EC Quito, , Ecuador +DK Randers SV, København V or Krakow +GB Redhill (GB) +DK Region Hovedstaden +DK Region Sjælland +IL Rehovot, Israel +AU Remote - Australia +CA Remote - CAN British Columbia + Remote - Portugal; United States +US Remote - USA WA, United States of America +US Remote - United States +DK Remote OR Copenhagen +US Remote US MD, United States of America + Remote job + Remote, AMER + Remote, Canada; Remote, Poland; Remote, United Kingdom; Remote, United States + Remote, Canada; Remote, United Kingdom; Remote, US +EU Remote, Germany; Dublin, Ireland; Lisbon, Portugal +US Remote, US, Georgia +US Remote, Washington, USA, United States of America +BG Remote- Bulgaria + Remote-NORAM +US Remote-US +RS Remote; Serbia +DE Remscheid +CA Retail Store Chinook Ctr, Canada +US Retail Store Ontario Mills, United States of America +HR Rijeka, Croatia +DK Ringkøbing, Denmark +DK Ringsted +DK Ringsted, Ringsted +BR Rio de Janeiro, Brasil +DE Rosenheim +DK Roskilde +DK Roskilde og mulighed for hjemmearbejde +DK Roskilde, Denmark +PL Rosówek, Poland + Roßdorf +AU Ryde, Australia +DE Röntgenstr. 33, 32107 Bad Salz, Germany +DK Rønne or Køge + Saint-Etienne + Saint-Malo +US San Bernardino, California, United States, California, United States of America +US San Francisco, CA; Washington, DC +US San Francisco, Seattle, New York, Chicago, Remote in the US +CR San Jose, Costa Rica +HN San Pedro Sula, , Honduras +MT San Ġwann, Malta +DE Schwäbisch Gmünd, BW, Germany +US Seattle, Washington, United States +US Seattle, Washington, United States, Washington, United States of America +US Seymour, United States +CN Shanghai, Shanghai, China; Suzhou, Jiangsu, China +PL Siedlc, Poland +SE Sigtuna, Sweden +SG Singapore - Singapore +SG Singapore, SG +DK Slagelse +SE Sollentuna, Sweden +SE Solna +BR Sorocaba, SP, Brazil +DK Sorø +GB Southampton, England, United Kingdom +SE Staffanstorp +US Stamford, Connecticut, United States, Connecticut, United States of America +SE Stockholm HQ +SE Stockholm, Stockholm County, Sweden +EU Stockholm, Stockholm County, Sweden; London, London, United Kingdom; Barcelona, Spain +SE Stockholm, Stockholm County, Sweden; Malmö, Sweden +EU Stockholm, Sweden; Berlin, Germany +EU Stockholm, Sweden; Jorvas, Kirkkonummi, Finland +EU Stockholm; London; Milan; Warsaw +CZ Strakonice, Czechia +SE Svedala +SE Sweden +SE Sweden - Linköping +SE Sweden - Lund +GB Swindon, United Kingdom +AU Sydney Or Melbourne +US Syracuse, New York, United States, New York, United States of America +PT São João, Aveiro, Portugal +SE Södertälje, Sweden +DK Taastrup +DK Taastrup og mulighed for hjemmearbejde +TW Taichung, Taichung City, Taiwan +TW Taipei, Taiwan, Taiwan + Taipei; Geneva +CN Taiyuan, Shanxi, China +EE Tallinn, Estonia +GE Tbilisi, Georgia +MX Tecnoparque, Mexico City, Mexico; Santa Fe, Mexico City, Mexico; Mexico City, Mexico +MX Tepotzotlan, Estado de México, Mexico +US The Woodlands, United States of America +RO Timisoara, Romania +GB Tiverton, United Kingdom +SE Tomelilla +MX Tonalá, Mexico + Tonsley SA + Toronto, New York, San Francisco, Atlanta, Seattle +PL Toruń, PM, Poland +PL Toruń, Poland +SE Trelleborg + Troisdorf +NO Trondheim, Norway + Tunis +DE Tübingen (DE) +US US-IL-Chicago-MSO + USA (Remote), Spain (Remote), UK (Remote) +US USA - IL - Chicago, United States of America +US USA - UT - Remote, United States of America +GB United Kingdom Remote +EU United Kingdom; Paris, Ile-de-France, France + United States; Stockholm, Sweden; Melbourne, Victoria, Australia + United States; Sweden +CN Urumqi, Xinjiang, China +FR Valbonne, Provence-Alpes-Côte d'Azur, France +DK Valby, Aalborg +DK Vallensbæk Strand +CA Vancouver + Vannes +FI Vantaa, Uusimaa, Finland +SE Varberg, Sweden +SE Vellinge +US Vernon Hills, IL, United States +DK Vi søger i hele landet +US Victorville, California, United States +DK Videbæk, Denmark +DK Videbæk, Region Central Jutland, Denmark +EU Vilnius, Lithuania; Copenhagen K, Denmark +EU Vilnius, Lithuania; Copenhagen V, Denmark +DK Værløse +EU Warsaw, Poland; Vilnius, Lithuania +PL Warszawa +AU Wensleydale, VIC, Australia +DE Wiesbaden, HE, Germany +CA Windsor, Canada +US Windsor, United States +GB Worksop, UK + Wuhan +CN XiAn, Shaanxi, China +SE Ystad +CN Zhuhai, Guangdong, China +SE Älmhult, Sweden +SE Ängelholm +SE Ängelholm, Sweden +SE Åby, Sweden +DK Åbyhøj eller Ballerup +SE Örebro, Sweden +SE Östra Göinge +CN 东莞, , China +""" + +LINES = [tuple(reversed(line.split("\t", 1))) for line in REAL.strip("\n").split("\n")] + +#: The tokens of every kind that are known. (`country:ZZ` and the rest that name +#: nothing are `known` false and match nothing, which the old LIKE did not say.) +ORACLE_TOKENS = [ + "country:DK", "country:SE", "country:DE", "country:GB", "country:US", "country:EU", + "country:NO", "country:CH", + "region:skane", "region:hovedstaden", "region:zealand", "region:oresund", + "region:jutland", "region:funen", + "place:copenhagen", "place:kobenhavn", "place:malmo", "place:lund", "place:munich", + "place:zurich", "place:london", "place:remote", "place:aarhus", "place:kgs lyngby", + "place:the hague", "place:scania", "place:capital region of denmark", + "place:denmark", "place:germany", "place:skane", "Skåne", "Berlin", +] + + +def like_words(column: str, phrases) -> tuple[str, list]: + """The clause `match_sql` was before it called a function: each phrase a + `LIKE` over the padded column. Slow, and the definition `matches` must keep.""" + phrases = [p for p in dict.fromkeys(phrases) if p] + clause = f"' ' || {column} || ' ' LIKE ? ESCAPE '\\'" + return " OR ".join([clause] * len(phrases)) or "0", [f"% {p} %" for p in phrases] + + +def like_country(code: str) -> tuple[str, list]: + names, params = like_words("place_fold", places._COUNTRY_MENTIONS.get(code, ())) + if not params: + return "country = ?", [code] + return f"country = ? OR (country IN ('', 'EU') AND ({names}))", [code, *params] + + +def like_sql(token: str) -> tuple[str, list]: + tok = parse(token) + if tok.kind == "country": + sql, params = like_country(tok.value) + elif tok.kind == "region": + sql, params = like_words("place_fold", places._PHRASES[tok.value]) + else: + region = places._REGION_BY_NAME.get(tok.value, "") + sql, params = like_words("place_fold", [*places._expand(tok.value), + *places._PHRASES.get(region, ())]) + if tok.value in places._COUNTRY_BY_NAME: + more, extra = like_country(places._COUNTRY_BY_NAME[tok.value]) + sql, params = f"{sql} OR {more}", [*params, *extra] + return f"({sql})", params + + +@pytest.fixture(scope="module") +def real() -> sqlite3.Connection: + db = connect() + 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 LINES]) + return db + + +def test_there_are_enough_real_lines() -> None: + assert len(set(LINES)) >= 300 + assert {"DK", "SE", "DE", "EU", ""} <= {country for _, country in LINES} + + +@pytest.mark.parametrize("token", ORACLE_TOKENS) +def test_matches_agrees_with_the_like_it_replaces(real: sqlite3.Connection, token: str) -> None: + sql, params = like_sql(token) + by_like = {row[0] for row in real.execute( + f"SELECT rowid FROM roles WHERE {sql}", params)} + by_function = {row[0] for row in real.execute( + "SELECT rowid, place_fold, country FROM roles") if matches(row[1], row[2], token)} + assert by_function == by_like, token + sql, params = match_sql(token) + assert {row[0] for row in real.execute( + f"SELECT rowid FROM roles WHERE {sql}", params)} == by_like + + +@pytest.mark.parametrize("kind, tokens", [ + ("country", [t for t in ORACLE_TOKENS if t.startswith("country:")]), + ("region", [t for t in ORACLE_TOKENS if t.startswith("region:")]), + ("place", [t for t in ORACLE_TOKENS if t.startswith("place:")]), +]) +def test_each_kind_finds_something_in_the_real_lines(real: sqlite3.Connection, kind: str, + tokens: list[str]) -> None: + """Agreement on nothing would prove little: most tokens hit, some lines do not.""" + hits = [sum(matches(fold(place), country, t) for place, country in LINES) for t in tokens] + assert sum(1 for n in hits if n) >= len(tokens) * 0.75, (kind, hits) + assert all(n < len(LINES) for n in hits) + + +def test_a_region_over_thousands_of_lines_is_fast() -> None: + """The reason `matches` is a function: 160 `LIKE` clauses took 400 ms per query. + The bound is generous, a machine under load is not the test's business.""" + import time + + db = connect() + db.execute("CREATE TABLE roles (place TEXT, place_fold TEXT, country TEXT)") + rows = [(f"{place} {i}", fold(f"{place} {i}"), country) + for i in range(10) for place, country in LINES] + db.executemany("INSERT INTO roles VALUES (?, ?, ?)", rows) + assert len(rows) >= 5000 + for token in ("region:oresund", "region:hovedstaden", "region:skane", "country:DK"): + sql, params = match_sql(token) + start = time.perf_counter() + n = db.execute(f"SELECT count(*) FROM roles WHERE {sql}", params).fetchone()[0] + assert time.perf_counter() - start < 0.5, token + assert n > 0 + + +# --------------------------------------------- what is suggested, and on what + + +def test_oresund_is_not_suggested_for_the_names_of_the_regions_it_holds() -> None: + for q in ("ska", "skane", "den", "capital", "hoved", "scania"): + assert "region:oresund" not in tokens(q), q + for q in ("ore", "øre", "oresund", "öresund"): + assert "region:oresund" in tokens(q), q + + +def test_a_region_is_suggested_on_its_own_names() -> None: + assert tokens("ska")[:1] == ["region:skane"] + assert "region:hovedstaden" in tokens("den") + assert "region:zealand" in tokens("den") or "region:zealand" in tokens("zea") + + +def test_a_line_that_holds_a_town_counts_under_it() -> None: + munich = [s for s in suggest("mun", PLACES) if s.kind == "place"] + assert [s.token for s in munich] == ["place:munich"] + # Munich 40, München 10, Office 5, (DE) 3, Munich - Berlin 2, Garching (Munich) 4. + assert munich[0].count == 40 + 10 + 5 + 3 + 2 + 4 + assert munich[0].label == "Munich" + + +def test_a_line_that_holds_two_towns_counts_under_both() -> None: + berlin = [s for s in suggest("ber", PLACES) if s.kind == "place"] + assert [(s.label, s.count) for s in berlin] == [("Berlin", 100 + 2)] + + +def test_copenhagen_with_an_office_or_a_hub_is_copenhagen() -> None: + towns = [s for s in suggest("cope", PLACES) if s.kind == "place"] + assert [s.token for s in towns] == ["place:kobenhavn"] + assert towns[0].count == 69 + 35 + 34 + 1 + 56 + 1 + 16 + 14 + 2 + for text in ("Copenhagen Office", "Hub", "MSO"): + assert all(text not in s.label for s in towns) + + +def test_a_town_the_tables_do_not_know_stays_as_it_is() -> None: + labels = {s.label for q in "abcdefghijklmnopqrstuvwxyz" + for s in suggest(q, PLACES, limit=999) if s.kind == "place"} + assert {"Albertslund", "Billund", "Auckland"} <= labels + assert not {"Copenhagen Office", "Munich Office", "Munich - Berlin", "Munich (DE)"} & labels + + +def test_a_line_that_is_only_a_known_town_in_a_word_is_not_a_hit_for_another() -> None: + """`Lund` is not in `Billund`: a line holds a town as whole words or not at all.""" + assert [s.label for s in suggest("lund", PLACES) if s.kind == "place"] == ["Lund"] + + +def test_a_bar_separates_places_like_a_semicolon() -> None: + rows = [("Denmark | Remote", "DK", 1), ("Lund | Malmö", "SE", 2)] + assert [s.label for s in suggest("den", rows) if s.kind == "place"] == [] + assert [(s.label, s.count) for s in suggest("mal", rows) if s.kind == "place"] \ + == [("Malmö", 2)]