Files
jobboard/tests/test_search.py
T
WolfandClaude Opus 5.5 d04ceae7e0 Prep stage: host words stored raw, mapped centrally in app/prep after every pull
upsert_role stores the host's own words (roles.*_host), the place texts
unresolved when they change, and the company key unresolved on insert, and
marks the role prep_due. app.prep maps seniority, employment, remote and
country through the host tables in app/prep/hosts, resolves places and
company aliases, lists unknown words in `unmapped`, remaps every open role
when the tables' fingerprint changes, and runs rebuild_dupes, which moves
out of the pull's settle. Migration 22 adds the columns and the table and
fills the host words from raw. Cascade: pull, prep, embed, score, enrich,
prep, embed, score, extract, score, summarise. `app prep [unmapped]
[--remap]`, /api/prep/unmapped and a fold on the Sources page.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-30 11:19:22 +02:00

374 lines
16 KiB
Python

"""The `search` plugin over the runner and the provider readers, `country_of`,
and what the pull does with searches. No network: `_runner.fetch` is a stub,
and a fake provider stands in where the pull is what is tested. The readers'
own parsing is tested in `test_readers_<provider>.py`."""
from __future__ import annotations
import json
import sys
import types
from datetime import UTC, datetime, timedelta
from pathlib import Path
from typing import Any, ClassVar
from urllib.parse import parse_qs, urlsplit
import pytest
ROOT = Path(__file__).resolve().parents[1]
if str(ROOT) not in sys.path:
sys.path.insert(0, str(ROOT))
from desk.models import Item # noqa: E402
from app import prep # noqa: E402
from plugins import readers, search # noqa: E402
from plugins.readers import Ask, Fields, Reads, _runner # noqa: E402
from plugins.search import SearchSource # noqa: E402
from app.config import SearchSeed # noqa: E402
from app.jobs import pull # noqa: E402
from pipeline import Tally # noqa: E402
from app.models import country_of, remote_of # noqa: E402
from app.storage import Store # noqa: E402
from prepped import upsert # noqa: E402
FIXTURES = ROOT / "tests" / "fixtures" / "search"
@pytest.fixture
def store(tmp_path: Path) -> Store:
return Store(tmp_path / "roledesk.db")
# ----------------------------------------------------------------- country
@pytest.mark.parametrize(("place", "code"), [
("Berlin, Germany", "DE"),
("København K", "DK"),
("Kongens Lyngby", "DK"),
("Boston, MA", "US"),
("Berlin, DE", "DE"), # a state code only when nothing else matched
("Remote - US", "US"),
("London | Remote", "GB"),
("Northern Ireland", "GB"), # the longer name wins over "ireland"
("Remote - Europe", "EU"),
("Home based - EMEA", "EU"),
("London, EMEA", "GB"), # a country beats the region it is in
("London or New York", ""),
("Remote", ""),
("", ""),
("Wrocław", "PL"), # ł and đ do not decompose under NFKD
("Łódź, Poland", "PL"),
("Bogota, CO", "CO"), # a state code that is a country's too
("Paderborn, DE", ""), # ... is no US state without a US place
("Sliema, MT", ""),
("Casablanca, MA", "MA"),
("Cambridge, MA", "US"),
("Chicago, IL", "US"),
("Madrid, MD, Spain", "ES"),
("Dublin, London", "EU"), # several European countries
("Anywhere in France, Belgium, Spain", "EU"),
("Amsterdam, Netherlands; London, United Kingdom; Remote - Europe", "EU"),
("Remote, EU or US", ""), # a region and a country outside it
("London, Ontario", "CA"), # a province beats the city before it
("Bellevue, Washington", "US"),
("Tilburg, NB, Netherlands", "NL"), # a country beats a province code
("Atlanta, Georgia", "US"),
("Tbilisi, Georgia", "GE"),
("Malaysia", "MY"),
("Belgrade, Serbia", "RS"),
("Dubai", "AE"),
("United Arab Emirates", "AE"),
("Buenos Aires, , Argentina", "AR"),
("Leicester (GB)", "GB"), # a code in parentheses when nothing else
("Cagliari (CA)", "IT"), # ... and not over a name
("Hybrid", ""),
("Distributed", ""),
])
def test_country_of(place: str, code: str) -> None:
assert country_of(place) == code
@pytest.mark.parametrize(("place", "mode"), [
("Remote", "remote"),
("Distributed", "remote"),
("Home based - Worldwide", "remote"),
("Remote - US", "remote"),
("Hybrid", "hybrid"),
("Remote or hybrid, Berlin", "hybrid"),
("In-Office", "onsite"),
("Berlin, Germany", ""),
("", ""),
])
def test_remote_of(place: str, mode: str) -> None:
assert remote_of(place) == mode
def test_upsert_fills_country_from_place_unless_the_plugin_says(store: Store) -> None:
def item(source_id: str, **raw: Any) -> Item:
it = Item(source_id=source_id, location="https://x.example/1", title="Engineer",
body="", published_at=datetime(2026, 9, 20, tzinfo=UTC), raw=raw)
it.source = "ats"
return it
upsert(store, item("a", place="Copenhagen, Denmark"))
upsert(store, item("b", place="Remote", country="EU"))
upsert(store, item("c", place="Somewhere"))
assert {r.source_id: r.country for r in store.roles()} == {"a": "DK", "b": "EU", "c": ""}
# ----------------------------------------------------- read through the runner
def _web(monkeypatch: pytest.MonkeyPatch, answer: Any) -> list[str]:
"""`_runner.fetch` answering `answer(url)`, remembering the URLs asked."""
asked: list[str] = []
def fake_fetch(client: object, url: str, allow: Any, **kwargs: Any) -> bytes:
assert allow(urlsplit(url).hostname), url
asked.append(url)
return answer(url)
monkeypatch.setattr(_runner, "fetch", fake_fetch)
monkeypatch.setattr(search, "PAGE_DELAY", 0)
return asked
def _jobindex_page(url: str) -> bytes:
"""A full Jobindex page (20 items) whatever page is asked."""
page = int(parse_qs(urlsplit(url).query)["page"][0])
items = "".join(f"<item><title>Role {page}-{n}, Acme</title>"
f"<link>https://www.jobindex.dk/vis-job/r{page}x{n}</link>"
f"<description>&lt;p&gt;Teaser {n}&lt;/p&gt;</description></item>"
for n in range(20))
return f"<rss><channel><title>Jobindex</title>{items}</channel></rss>".encode()
@pytest.mark.parametrize(("pages", "asked"), [(None, 5), (2, 2), (30, 10), ("x", 5)])
def test_the_saved_pages_param_bounds_the_pages_read(
monkeypatch: pytest.MonkeyPatch, pages: Any, asked: int) -> None:
urls = _web(monkeypatch, _jobindex_page)
params: dict[str, Any] = {"q": "python", "area": 1} | ({} if pages is None else
{"pages": pages})
rows, complete = search.run_search("jobindex:cph", params)
assert len(urls) == asked and len(rows) == 20 * asked
assert not complete # Jobindex states no total: never complete
query = parse_qs(urlsplit(urls[0]).query)
assert (query["q"], query["geoareaid"]) == (["python"], ["1"])
def test_a_jobindex_role_hands_its_teaser_over_in_raw_not_in_the_body(
monkeypatch: pytest.MonkeyPatch) -> None:
_web(monkeypatch, lambda url: (FIXTURES / "jobindex.xml").read_bytes())
source = SearchSource({"searches": [{"url": "jobindex:cph", "label": "Copenhagen",
"params": {"q": "x", "pages": 1}}]})
items = list(source.fetch())
paid = items[0]
assert paid.source_id == "jobindex:h1701165"
assert (paid.body, paid.excerpt) == ("", "")
assert paid.raw["teaser"].startswith("Do you want to minimise")
assert paid.raw["company"] == "DSV" and paid.raw["country"] == "DK"
assert paid.raw["places"] == [{"text": "Hedehusene", "remote": None}]
assert paid.raw["place"] == "Hedehusene"
assert (paid.raw["provider"], paid.raw["search"]) == ("jobindex", "jobindex:cph")
assert source.state["search:complete:jobindex:cph"] == ""
def _adzuna(count: int | None, stock: int = 60) -> Any:
def answer(url: str) -> bytes:
page = int(urlsplit(url).path.rsplit("/", 1)[1])
start = (page - 1) * 50
results = [{"id": str(n), "title": f"Role {n}", "redirect_url": f"https://x.example/{n}"}
for n in range(start, min(start + 50, stock))]
body: dict[str, Any] = {"results": results}
if count is not None:
body["count"] = count
return json.dumps(body).encode()
return answer
def test_a_search_without_its_key_is_refused_before_any_request(
monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.delenv("ADZUNA_APP_ID", raising=False)
monkeypatch.delenv("ADZUNA_APP_KEY", raising=False)
urls = _web(monkeypatch, _adzuna(60))
with pytest.raises(ValueError, match="app_id"):
search.run_search("adzuna:berlin", {"q": "x", "country": "de"}, {})
assert urls == []
def test_the_key_comes_from_the_config_before_the_environment(
monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.setenv("ADZUNA_APP_ID", "env-id")
monkeypatch.setenv("ADZUNA_APP_KEY", "env-key")
urls = _web(monkeypatch, _adzuna(60))
search.run_search("adzuna:berlin", {"q": "x", "country": "de"},
{"adzuna": {"app_id": "cfg-id"}})
query = parse_qs(urlsplit(urls[0]).query)
assert (query["app_id"], query["app_key"]) == (["cfg-id"], ["env-key"])
assert urlsplit(urls[0]).path == "/v1/api/jobs/de/search/1"
@pytest.mark.parametrize(("count", "complete"), [(60, True), (75, False)])
def test_an_adzuna_search_is_complete_when_it_read_its_count(
monkeypatch: pytest.MonkeyPatch, count: int, complete: bool) -> None:
_web(monkeypatch, _adzuna(count))
rows, done = search.run_search("adzuna:berlin", {"q": "x", "country": "de"},
{"adzuna": {"app_id": "i", "app_key": "k"}})
assert (len(rows), done) == (60, complete)
assert rows[0]["place"] == ""
def test_a_platsbanken_search_sends_region_and_offset_and_stops_at_its_total(
monkeypatch: pytest.MonkeyPatch) -> None:
def answer(url: str) -> bytes:
query = parse_qs(urlsplit(url).query)
start = int(query["offset"][0])
hits = [{"id": str(n), "headline": f"Role {n}", "webpage_url": f"https://x.example/{n}",
"workplace_addresses": [{"municipality": "Lund"}, {"municipality": "Malmö"}]}
for n in range(start, min(start + 100, 200))]
return json.dumps({"total": {"value": 200}, "hits": hits}).encode()
urls = _web(monkeypatch, answer)
rows, done = search.run_search("platsbanken:skane", {"q": "data", "region": "CaRE_1nn_cSU"})
assert (len(rows), done) == (200, True)
offsets = [parse_qs(urlsplit(u).query)["offset"] for u in urls]
assert offsets == [["0"], ["100"]]
assert parse_qs(urlsplit(urls[0]).query)["region"] == ["CaRE_1nn_cSU"]
assert rows[0]["place"] == "Lund; Malmö"
def test_an_unknown_provider_is_refused() -> None:
with pytest.raises(ValueError, match="provider"):
search.run_search("indeed:x", {"q": "x"})
def test_provider_of() -> None:
assert search.provider_of("jobindex:copenhagen-backend") == "jobindex"
assert search.provider_of("adzuna:berlin") == "adzuna"
assert search.provider_of("platsbanken:skane-data") == "platsbanken"
assert search.provider_of("indeed:x") == ""
assert search.provider_of("jobindex:") == ""
# ------------------------------------------------------------------- pull
class FakeConfig:
sources: ClassVar[dict[str, object]] = {}
searches: ClassVar[list[SearchSeed]] = []
path = None
def _fake_provider(monkeypatch: pytest.MonkeyPatch, ids: list[str], complete: bool) -> None:
"""A provider `fake` that lists `ids` in one page and states a total of
as many, or of one more when the search is to be incomplete."""
module: Any = types.ModuleType("fake")
module.NAME, module.LABEL = "fake", "Fake"
module.READS = Reads(scope="search", listing="paged", text="listing", complete="stated",
places="one", fields=Fields(country="stated"), robots="test")
module.allowed = lambda slug: lambda host: host == "search.fake.example"
def page(get: Any, slug: str, cursor: Any, ask: Ask) -> Any:
assert ask.params["q"]
return ids, [{"id": i} for i in ids], len(ids) + (0 if complete else 1), None
module.page = page
module.rows = lambda decoded, slug: [
{"source_id": i, "title": f"Engineer {i}", "company": "Acme",
"url": f"https://x.example/{i}", "places": [{"text": "København", "remote": None}],
"country": "DK", "remote": None, "posted": None, "description": ""}
for i in decoded]
assert not readers.problems(module, "fake")
monkeypatch.setitem(readers.PROVIDERS, "fake", module)
def _pull(store: Store) -> dict[str, Any]:
results = [pull.run_batch(FakeConfig(), store, **batch) # type: ignore[arg-type]
for batch in pull._boards(FakeConfig(), store)] # type: ignore[arg-type]
closed = pull.close_boards(FakeConfig(), store, Tally(results=tuple(results))) # type: ignore[arg-type]
prep.run(store)
return closed
def test_a_complete_search_closes_what_it_stopped_listing(
store: Store, monkeypatch: pytest.MonkeyPatch) -> None:
store.add_subscription("search", "fake:cph", label="Copenhagen", board="fake",
params={"q": "backend"})
_fake_provider(monkeypatch, ["h1", "h2"], complete=True)
_pull(store)
roles = {r.source_id: r for r in store.roles()}
assert set(roles) == {"fake:h1", "fake:h2"}
assert roles["fake:h1"].country == "DK"
assert roles["fake:h1"].subscription_id is not None
_fake_provider(monkeypatch, ["h2"], complete=True)
result = _pull(store)
assert {r.source_id for r in store.roles()} == {"fake:h2"}
assert (result["searches"], result["closed"]) == (1, 1)
def test_an_incomplete_search_closes_nothing_until_stale(
store: Store, monkeypatch: pytest.MonkeyPatch) -> None:
store.add_subscription("search", "fake:cph", label="Copenhagen", board="fake",
params={"q": "backend"})
_fake_provider(monkeypatch, ["h1", "h2"], complete=False)
_pull(store)
_fake_provider(monkeypatch, ["h2"], complete=False)
_pull(store)
assert {r.source_id for r in store.roles()} == {"fake:h1", "fake:h2"}
old = (datetime.now(UTC) - timedelta(days=pull.STALE_DAYS + 1)).isoformat()
with store.transaction() as conn:
conn.execute("UPDATE roles SET last_seen = ? WHERE source_id = 'fake:h1'", (old,))
_pull(store)
assert {r.source_id for r in store.roles()} == {"fake:h2"}
def test_seed_searches_imports_once(store: Store) -> None:
config = FakeConfig()
config.searches = [SearchSeed(name="Copenhagen backend", provider="jobindex",
params={"q": "backend", "area": 1})]
assert pull.seed_searches(config, store) == ["Copenhagen backend"] # type: ignore[arg-type]
sub = store.subscriptions(source="search")[0]
assert (sub.target, sub.params) == ("jobindex:copenhagen-backend",
{"q": "backend", "area": 1})
store.unsubscribe(sub.id)
with store.transaction() as conn:
conn.execute("DELETE FROM subscriptions")
assert pull.seed_searches(config, store) == [] # type: ignore[arg-type]
def test_search_source_stamps_complete_and_ids(monkeypatch: pytest.MonkeyPatch) -> None:
_fake_provider(monkeypatch, ["h1"], complete=True)
source = SearchSource({"searches": [{"url": "fake:cph", "label": "Copenhagen",
"params": {"q": "x"}}]})
items = list(source.fetch())
assert [i.source_id for i in items] == ["fake:h1"]
assert items[0].raw["search"] == "fake:cph"
assert source.state["search:complete:fake:cph"] == "1"
assert source.state["search:ids:fake:cph"] == '["fake:h1"]'
def test_a_search_that_fails_is_skipped_and_stamps_nothing(
monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.delenv("ADZUNA_APP_ID", raising=False)
monkeypatch.delenv("ADZUNA_APP_KEY", raising=False)
source = SearchSource({"searches": [{"url": "adzuna:berlin", "label": "Berlin",
"params": {"q": "x", "country": "de"}}]})
assert list(source.fetch()) == []
assert "search:ok:adzuna:berlin" not in source.state
def test_a_search_switched_off_mid_pull_ends_the_pull_closed(
store: Store, monkeypatch: pytest.MonkeyPatch) -> None:
sub = store.add_subscription("search", "fake:cph", label="Copenhagen",
board="fake", params={"q": "backend"})
_fake_provider(monkeypatch, ["h1"], complete=False)
batches = pull._boards(FakeConfig(), store) # type: ignore[arg-type]
result = pull.run_batch(FakeConfig(), store, **batches[0]) # type: ignore[arg-type]
store.update_subscription(sub.id, subscribed=False, fetch=False)
pull.close_boards(FakeConfig(), store, Tally(results=(result,))) # type: ignore[arg-type]
assert store.roles() == []
# And a batch that starts after the switch reads nothing.
assert pull.run_batch(FakeConfig(), store, **batches[0])["roles"] == 0 # type: ignore[arg-type]