Hosts a to l declare READS: places, teaser, stated seniority and employment

Ashby, BambooHR, Deel, Eightfold, Gem, Greenhouse, HiBob, HR Manager, Jibe,
Jobylon and Lever move to feed_url(slug, ask) and page(get, slug, cursor, ask),
declare READS with the robots decision, and drop robots_page. Rows carry
places (Ashby secondaryLocations, Lever allLocations, Greenhouse offices in
place of a Hybrid placeholder), Jobylon keeps its summary as teaser and maps
experience and employment, Eightfold reads creationTs as the posting date and
Ashby reads workplaceType. Fixtures <name>-places.* and jobylon-teaser.html
drive the conformance checks; the eleven names leave UNDECLARED.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
WolfandClaude Opus 5.5 committed 2026-09-30 09:11:53 +02:00
1 parent 4c0da58da7
commit f2e652a8da
35 files changed
+1290 -391

No files matched your search

+41 -23
View File
@@ -18,7 +18,7 @@ if str(ROOT) not in sys.path:
from app.jobs import discover # noqa: E402
from plugins import ats, readers # noqa: E402
from plugins.readers import eightfold # noqa: E402
from plugins.readers import Ask, eightfold # noqa: E402
FIXTURES = ROOT / "tests" / "fixtures" / "ats"
SLUG = "jobs.ericsson.com/ericsson.com"
@@ -47,6 +47,26 @@ def _search(total: int, seen: list[str] | None = None) -> httpx.MockTransport:
return httpx.MockTransport(handler)
def _crawl(slug: str, transport: httpx.MockTransport) -> tuple[list[dict[str, Any]], bool]:
"""What the runner does with a paged reader: pages until the cursor ends,
complete when as many distinct ids were read as the host stated."""
rows: list[dict[str, Any]] = []
ids: set[str] = set()
total: Any = None
with httpx.Client(transport=transport) as client:
def get(url: str, **_: Any) -> bytes:
return client.get(url).content
cursor: Any = None
while True:
decoded, results, total, cursor = eightfold.page(get, slug, cursor, Ask())
rows += eightfold.rows(decoded, slug)
ids |= {r["id"] for r in results}
if cursor is None or (total is not None and len(ids) >= total):
break
return rows, total is not None and len(ids) >= total
def test_a_page_becomes_rows() -> None:
rows = eightfold.rows(_saved()["data"]["positions"], SLUG)
assert len(rows) == 5
@@ -56,11 +76,14 @@ def test_a_page_becomes_rows() -> None:
assert first["url"] == "https://jobs.ericsson.com/careers/job/563121777024688"
assert first["location"] == "Stockholm, Sweden" # the city is its own region
assert first["department"] == "76 Product Development"
assert first["posted"] == datetime.fromtimestamp(1790703782, tz=UTC)
# `creationTs`, the posting date; `postedTs` (1790703782) is refreshed in bulk.
assert first["posted"] == datetime.fromtimestamp(1785822052, tz=UTC)
assert first["places"] == [{"text": "Stockholm, Sweden", "remote": None}]
assert first["remote"] == "onsite" and (first["description"], first["salary_text"]) == ("", "")
assert mexico["location"] == "Mexico City, Mexico"
assert many["location"] == ("Tecnoparque, Mexico City, Mexico; Santa Fe, Mexico City, Mexico; "
"Mexico City, Mexico; Tlalnepantla, Estado de México, Mexico")
assert [p["text"] for p in many["places"]] == many["location"].split("; ")
@pytest.mark.parametrize(("option", "remote"), [
@@ -89,36 +112,23 @@ def test_a_page_asks_for_the_newest_first_from_its_offset() -> None:
asked.append(url)
return json.dumps(_saved()).encode()
decoded, results, total, cursor = eightfold.page(get, SLUG, None)
decoded, results, total, cursor = eightfold.page(get, SLUG, None, Ask())
assert asked == ["https://jobs.ericsson.com/api/pcsx/search?domain=ericsson.com"
"&sort_by=timestamp&start=0"]
assert (len(decoded), len(results), total, cursor) == (5, 5, 637, 5)
assert eightfold.page(get, SLUG, 630)[3] == 635
assert eightfold.page(get, SLUG, 630, Ask())[3] == 635
assert asked[-1].endswith("&start=630")
def test_a_board_is_read_ten_at_a_time_to_its_total(monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.setattr(ats, "PAGE_DELAY", 0)
def test_a_board_is_read_ten_at_a_time_to_its_total() -> None:
seen: list[str] = []
with httpx.Client(transport=_search(25, seen)) as client:
rows, complete = ats._fetch_rows(client, f"eightfold:{SLUG}")
rows, complete = _crawl(SLUG, _search(25, seen))
assert (len(rows), complete) == (25, True)
assert [u.rsplit("=", 1)[1] for u in seen] == ["0", "10", "20"]
def test_a_board_over_the_page_cap_is_incomplete(monkeypatch: pytest.MonkeyPatch) -> None:
"""The cap is 100 pages of ten: a board of 1,292 (Infineon's) or 2,319
(Microsoft's) is read as its first thousand and never as the whole."""
monkeypatch.setattr(ats, "PAGE_DELAY", 0)
seen: list[str] = []
with httpx.Client(transport=_search(1292, seen)) as client:
rows, complete = ats._fetch_rows(client, "eightfold:infineon.eightfold.ai/infineon.com")
assert (len(rows), complete) == (1000, False) and len(seen) == ats.MAX_PAGES == 100
def test_an_empty_board_is_complete_at_zero(monkeypatch: pytest.MonkeyPatch) -> None:
with httpx.Client(transport=_search(0)) as client:
assert ats._fetch_rows(client, f"eightfold:{SLUG}") == ([], True)
def test_an_empty_board_is_complete_at_zero() -> None:
assert _crawl(SLUG, _search(0)) == ([], True)
@pytest.mark.parametrize("body", [
@@ -126,7 +136,7 @@ def test_an_empty_board_is_complete_at_zero(monkeypatch: pytest.MonkeyPatch) ->
b'{"data": {"positions": null}}'])
def test_an_answer_that_is_not_a_search_raises(body: bytes) -> None:
with pytest.raises(ValueError):
eightfold.page(lambda url, **_: body, SLUG, None)
eightfold.page(lambda url, **_: body, SLUG, None, Ask())
def test_detail_reads_the_text_of_the_ad() -> None:
@@ -164,7 +174,6 @@ def test_the_slug_the_urls_and_the_allow_list() -> None:
for slug in (SLUG, "infineon.eightfold.ai/infineon.com", "Apply.Careers.Microsoft.com/microsoft.com"):
assert eightfold.slug_ok(slug) and ats.is_known_target(f"eightfold:{slug}")
assert eightfold.careers_url("Jobs.Ericsson.com/ericsson.com") == "https://jobs.ericsson.com/careers"
assert eightfold.robots_page(SLUG) == "https://jobs.ericsson.com/careers"
allow = eightfold.allowed("Jobs.Ericsson.com/ericsson.com")
assert allow("jobs.ericsson.com") and allow("JOBS.ERICSSON.COM")
assert not allow("ericsson.com") and not allow("evil.example")
@@ -193,3 +202,12 @@ def test_discovery_reads_the_host_and_domain_from_a_link_that_has_both() -> None
assert ("eightfold", "infineon.eightfold.ai/infineon.com") in found
assert ("eightfold", "") in found
assert "eightfold" not in {h for h, _ in discover._UNREADABLE if h not in readers.HOSTS}
def test_it_declares_a_paged_read_of_ten_with_a_total_and_no_filter_yet() -> None:
reads = eightfold.READS
assert (reads.listing, reads.page_size, reads.text, reads.complete, reads.places) == (
"paged", 10, "detail", "stated", "many")
assert reads.filters.declared() == {} and reads.fields.posted == "listed"
assert reads.robots.startswith("read: robots.txt allows /careers")
assert not hasattr(eightfold, "robots_page")