stack/tests/bcda/test_ui_views.py

"""bcda.ui.views — every page through Streamlit's AppTest with the
service layer stubbed, so no network, no replica, no real store."""

from __future__ import annotations

import logging
import threading
import time
from pathlib import Path

import duckdb
import pytest
from streamlit.testing.v1 import AppTest

from bcda.store import Store
from bcda.ui import service as svc
from bcda.ui import views
from bcda.ui.runner import Runner

SMALL = svc.PRESETS["Small Model Entity"]


def _page(
    name: str, *, conn: svc.Connection | None = SMALL, timeout: float = 60
) -> AppTest:
    at = AppTest.from_string(
        f"from bcda.ui import views\nviews.{name}()\n", default_timeout=timeout
    )
    if conn is not None:
        at.session_state["conn"] = conn
    return at


def _button(at: AppTest, label: str):  # noqa: ANN202
    return next(b for b in at.button if b.label == label)


@pytest.fixture(autouse=True)
def fresh_runner(monkeypatch):
    """Each test gets its own Runner so background runs never leak across tests."""
    r = Runner()
    monkeypatch.setattr(views, "runner", r)
    return r


@pytest.fixture
def fake_client(monkeypatch):
    """A client stand-in; service functions are stubbed to accept it."""

    class FakeClient:
        base_url = svc.SANDBOX
        version = "v3"

        def close(self) -> None:
            pass

    client = FakeClient()
    monkeypatch.setattr(svc, "make_client", lambda conn, output_dir=None: client)
    return client


@pytest.fixture
def tmp_store(monkeypatch, tmp_path):
    monkeypatch.setattr(svc, "store_path", lambda: tmp_path)
    return Store(path=str(tmp_path))


# ── Connect ──────────────────────────────────────────────────


class TestConnect:
    def test_preset_then_preflight_and_auth(self, monkeypatch, fake_client) -> None:
        monkeypatch.setattr(
            svc,
            "preflight",
            lambda c: {
                "ok": True,
                "fhir_version": "4.0.1",
                "software": "r302",
                "release_date": "2026",
                "resources": ["Patient"],
                "health": {"database": "ok"},
                "server_version": {"version": "r302"},
            },
        )
        monkeypatch.setattr(
            svc,
            "auth_check",
            lambda c: {"token_chars": 40, "last_attribution_update": "x"},
        )
        at = _page("connect", conn=None).run()
        assert at.session_state.get("conn") is None
        _button(at, "Use this connection").click().run()
        assert at.session_state["conn"].label == "Large Model Entity"
        _button(at, "Preflight (no token)").click().run()
        assert any(m.value == "4.0.1" for m in at.metric)
        _button(at, "Authenticate + attribution").click().run()
        assert any("Token issued" in s.value for s in at.success)

    def test_preflight_failure_and_auth_error(self, monkeypatch, fake_client) -> None:
        monkeypatch.setattr(
            svc, "preflight", lambda c: {"ok": False, "error": "unreachable"}
        )

        def bad_auth(c):  # noqa: ANN001, ANN202
            raise RuntimeError("401 invalid client")

        monkeypatch.setattr(svc, "auth_check", bad_auth)
        at = _page("connect").run()
        _button(at, "Preflight (no token)").click().run()
        assert any("unreachable" in e.value for e in at.error)
        _button(at, "Authenticate + attribution").click().run()
        assert any("invalid client" in e.value for e in at.error)

    def test_environment_and_custom_sources(self, monkeypatch) -> None:
        monkeypatch.setenv("BCDA_CLIENT_ID", "env-id")
        monkeypatch.setenv("BCDA_CLIENT_SECRET", "env-secret-long-enough")
        at = _page("connect", conn=None).run()
        radio = at.radio[0]
        assert radio.options[0] == "Environment credentials"
        _button(at, "Use this connection").click().run()
        assert at.session_state["conn"].client_id == "env-id"
        radio.set_value("Custom").run()
        assert not [
            b for b in at.button if b.label == "Use this connection"
        ]  # needs id + secret first
        at.text_input(key="custom_id").input("abc").run()
        at.text_input(key="custom_secret").input("def").run()
        at.selectbox(key="custom_version").set_value("v2").run()
        _button(at, "Use this connection").click().run()
        conn = at.session_state["conn"]
        assert (conn.client_id, conn.client_secret, conn.version, conn.label) == (
            "abc",
            "def",
            "v2",
            "custom",
        )


# ── Exports ──────────────────────────────────────────────────


class TestExports:
    def test_requires_connection(self) -> None:
        at = _page("exports", conn=None).run()
        assert any("Connect" in i.value for i in at.info)

    def test_start_export_per_type_and_one_job(
        self, monkeypatch, fake_client, tmp_store, fresh_runner
    ) -> None:
        calls: list[dict] = []
        monkeypatch.setattr(
            svc,
            "export_per_type",
            lambda store, client, **kw: (
                calls.append({"per_type": True, **kw}) or {"Patient": "K1"}
            ),
        )
        monkeypatch.setattr(
            svc,
            "start_export",
            lambda store, client, **kw: calls.append({"per_type": False, **kw}) or "K2",
        )
        at = _page("exports").run()
        at.multiselect(key="exp_types").set_value(["Patient"])
        at.multiselect(key="exp_system_types").set_value(["DDPS"])
        at.selectbox(key="exp_outcome").set_value("complete")
        at.radio(key="exp_since_mode").set_value("date")
        # an empty store has no baseline: the checks block and nothing starts
        _button(at, "Start export").click().run()
        assert fresh_runner.all() == []
        assert "no_baseline" in at.dataframe[0].value["code"].tolist()
        assert any("blocks this export" in e.value for e in at.error)
        # forced through
        at.checkbox(key="exp_force").check()
        _button(at, "Start export").click().run()
        assert any("despite" in w.value for w in at.warning)
        run = fresh_runner.wait(fresh_runner.all()[0].id)
        assert run.state == "done" and run.result == {"Patient": "K1"}
        assert calls[0]["per_type"] and calls[0]["types"] == ["Patient"]
        assert calls[0]["type_filter"].endswith("|DDPS&outcome=complete") and calls[0][
            "since"
        ].endswith("T00:00:00+00:00")
        assert calls[0]["force"] is True
        assert any("_typeFilter=" in c.value for c in at.caption)
        # watermark mode hands the service the sentinel
        at.radio(key="exp_since_mode").set_value("watermark")
        _button(at, "Start export").click().run()
        fresh_runner.wait(fresh_runner.all()[0].id)
        assert calls[-1]["since"] == svc.SINCE_AUTO
        at.radio(key="exp_since_mode").set_value("off")
        at.checkbox(key="exp_per_type").uncheck()
        at.checkbox(key="exp_use_svc_date").check()
        _button(at, "Start export").click().run()
        fresh_runner.wait(fresh_runner.all()[0].id)
        assert (
            calls[-1]["per_type"] is False
            and "service-date=gt" in calls[-1]["type_filter"]
        )

    def test_jobs_adopt_and_cancel(
        self, monkeypatch, fake_client, tmp_store, fresh_runner
    ) -> None:
        jobs = [
            {
                "job_id": "11",
                "status": "completed",
                "types": "Patient,Coverage",
                "endpoint": "Group/all",
                "started": "s",
                "ended": "e",
                "reusable": True,
                "job_url": "u11",
            },
            {
                "job_id": "12",
                "status": "in-progress",
                "types": "(all)",
                "endpoint": "Group/all",
                "started": "s",
                "ended": "",
                "reusable": False,
                "job_url": "u12",
            },
        ]
        monkeypatch.setattr(svc, "server_jobs", lambda c: jobs)
        adopted: list = []
        monkeypatch.setattr(
            svc,
            "adopt_job",
            lambda store, client, url, types=None: adopted.append((url, types)) or "KA",
        )
        cancelled: list = []
        monkeypatch.setattr(svc, "cancel_job", lambda c, url: cancelled.append(url))
        at = _page("exports").run()
        _button(at, "Refresh jobs").click().run()
        assert len(at.dataframe) == 1
        _button(at, "Adopt").click().run()
        run = fresh_runner.wait(fresh_runner.all()[0].id)
        assert run.result == "KA" and adopted == [("u11", ["Patient", "Coverage"])]
        _button(at, "Cancel").click().run()
        assert cancelled == ["u12"] and "server_jobs" not in at.session_state

    def test_jobs_errors(self, monkeypatch, fake_client, tmp_store) -> None:
        def boom(*a, **k):  # noqa: ANN002, ANN003, ANN202
            raise RuntimeError("nope")

        monkeypatch.setattr(svc, "server_jobs", boom)
        at = _page("exports").run()
        _button(at, "Refresh jobs").click().run()
        assert any("Could not list jobs" in e.value for e in at.error)
        jobs = [
            {
                "job_id": "12",
                "status": "in-progress",
                "types": "(all)",
                "endpoint": "Group/all",
                "started": "s",
                "ended": "",
                "reusable": False,
                "job_url": "u12",
            }
        ]
        monkeypatch.setattr(svc, "server_jobs", lambda c: jobs)
        monkeypatch.setattr(svc, "cancel_job", boom)
        _button(at, "Refresh jobs").click().run()
        _button(at, "Cancel").click().run()
        assert any("Cancel failed" in e.value for e in at.error)


# ── Store / Load ─────────────────────────────────────────────


class TestStoreAndLoad:
    def test_store_empty_and_populated(self, tmp_store, tmp_path) -> None:
        at = _page("store").run()
        assert any("No store jobs" in i.value for i in at.info)
        tmp_store._append_job(
            key="K1",
            status="complete",
            endpoint="Group/all",
            types="Patient",
            job_url="u",
        )
        tmp_store._append_file(
            job_key="K1",
            resource_type="Patient",
            url="u1",
            storage_path="p",
            file_size=2048,
            status="complete",
        )
        d = tmp_path / "exports" / "K1"
        d.mkdir(parents=True)
        (d / "Patient.ndjson").write_text("{}\n")
        at = _page("store").run()
        assert at.selectbox(key="store_key").value == "K1"
        assert len(at.dataframe) == 3  # jobs, files, dirs
        assert (
            views._fmt_bytes(2048) == "2.0 KB"
            and views._fmt_bytes(5) == "5 B"
            and views._fmt_bytes(3 * 1024**4) == "3.0 TB"
        )

    def test_load_runs_in_background(self, monkeypatch, fresh_runner) -> None:
        seen = {}
        monkeypatch.setattr(
            svc, "load_and_publish", lambda **kw: seen.update(kw) or {"bcda.patient": 3}
        )
        at = _page("load").run()
        at.checkbox(key="load_skip").check()
        _button(at, "Load now").click().run()
        run = fresh_runner.wait(fresh_runner.all()[0].id)
        assert run.result == {"bcda.patient": 3} and seen == {
            "run_pipeline": True,
            "skip_flatten": True,
        }
        at.run()  # the runs panel now shows the finished run
        assert any("flatten + load" in e.label for e in at.expander)
        _button(at, "Clear finished runs").click().run()
        assert fresh_runner.all() == []


# ── Data ─────────────────────────────────────────────────────


class TestData:
    def test_no_replica(self, monkeypatch) -> None:
        def boom():  # noqa: ANN202
            raise OSError("no replica")

        monkeypatch.setattr(svc, "replica_connect", boom)
        at = _page("data").run()
        assert any("Replica not available" in w.value for w in at.warning)

    def test_empty_replica(self, monkeypatch) -> None:
        con = duckdb.connect()
        monkeypatch.setattr(svc, "replica_connect", lambda: con)
        at = _page("data").run()
        assert any("No `bcda.*`" in i.value for i in at.info)

    def test_tables_summaries_and_sql(self, monkeypatch, tmp_path) -> None:
        db = tmp_path / "r.duckdb"
        con = duckdb.connect(str(db))
        con.execute("create schema bcda")
        con.execute(
            "create table bcda.explanation_of_benefit as select 'DDPS' system_type, 'PDE' eob_type, 'FinalAction' final_action, 1.0 clm_pmt_amt"
        )
        con.close()
        ro = duckdb.connect(str(db), read_only=True)
        monkeypatch.setattr(svc, "replica_connect", lambda: ro)
        at = _page("data").run()
        assert (
            at.metric[0].label == "bcda.explanation_of_benefit"
            and at.metric[0].value == "1"
        )
        assert len(at.dataframe) == 1  # the default summary
        at.selectbox(key="summary_pick").set_value("Coverage rows by plan").run()
        assert any(
            "coverage" in w.value.lower() for w in at.warning
        )  # table absent → warning
        _button(at, "Run").click().run()
        assert any("1 rows" in c.value for c in at.caption)
        at.text_area(key="sql").input("select * from nope").run()
        _button(at, "Run").click().run()
        assert at.error


# ── Dictionary / Upstream ────────────────────────────────────


class TestDictionaryAndUpstream:
    def test_dictionary_search(self, monkeypatch) -> None:
        monkeypatch.setattr(svc, "replica_connect", lambda: duckdb.connect())
        at = _page("dictionary").run()
        assert len(at.dataframe) == 3
        at.text_input(key="dict_search").input("CLM_CNTL_NUM").run()
        assert (
            "Dictionary (" in at.tabs[0].label and at.tabs[0].label != "Dictionary (0)"
        )
        at.checkbox(key="xw_mapped").uncheck().run()
        assert len(at.dataframe) == 3

    def test_dictionary_missing(self, monkeypatch) -> None:
        def boom():  # noqa: ANN202
            raise OSError("no replica")

        monkeypatch.setattr(svc, "replica_connect", boom)
        monkeypatch.setattr(svc, "dictionary_frames", lambda con: {})
        at = _page("dictionary").run()
        assert any("Crosswalk seeds not found" in i.value for i in at.info)

    def test_upstream(self, monkeypatch, fresh_runner) -> None:
        monkeypatch.setattr(svc, "vendor_status", lambda: {"available": False})
        at = _page("upstream").run()
        assert any("not available" in i.value for i in at.info)
        status = {
            "available": True,
            "sources": [
                {
                    "name": "bcda-app",
                    "kind": "github",
                    "origin": "CMSgov/bcda-app",
                    "commit": "abc",
                    "files": 1,
                }
            ],
            "files": [{"path": "bcda-app/LICENSE", "sha256": "0" * 12}],
            "problems": ["missing bcda-app/LICENSE"],
        }
        monkeypatch.setattr(svc, "vendor_status", lambda: status)
        monkeypatch.setattr(svc, "vendor_check", lambda: [])
        at = _page("upstream").run()
        assert any("does not match" in e.value for e in at.error)
        _button(at, "Check upstream for drift (network)").click().run()
        assert fresh_runner.wait(fresh_runner.all()[0].id).result == []
        status["problems"] = []
        at = _page("upstream").run()
        assert any("matches the manifest" in s.value for s in at.success)


# ── runs panel states ────────────────────────────────────────


class TestRunsPanel:
    def test_failed_and_active_runs_render(self, fresh_runner) -> None:
        gate = threading.Event()
        logging.getLogger("bcda.ui.test")

        def fail() -> None:
            raise RuntimeError("kaput")

        fresh_runner.wait(fresh_runner.submit("will fail", fail).id)
        fresh_runner.submit("hangs", gate.wait, 10)
        time.sleep(0.05)
        at = _page("load").run()
        labels = [e.label for e in at.expander]
        assert any("will fail" in lbl and "failed" in lbl for lbl in labels)
        assert any("hangs" in lbl and "running" in lbl for lbl in labels)
        assert any("kaput" in e.value for e in at.error)
        gate.set()
        fresh_runner.wait(fresh_runner.all()[0].id)


# ── Sync ─────────────────────────────────────────────────────


class TestSync:
    def test_empty_store(self, tmp_store) -> None:
        at = _page("sync").run()
        assert not at.exception
        assert any("No complete jobs" in i.value for i in at.info)
        assert at.title[0].value == "Sync"

    def test_with_jobs_and_reconcile(
        self, monkeypatch, tmp_store, fresh_runner
    ) -> None:
        from datetime import datetime, timezone

        from bcda import sync

        from .test_sync import _rows, add_job

        t0 = datetime(2026, 10, 1, tzinfo=timezone.utc)
        add_job(tmp_store, "B1", {"Patient": _rows("Patient")[:5]}, transaction_time=t0)
        at = _page("sync").run()
        assert not at.exception
        assert any("1 pending" in b.label for b in at.button)
        assert any("nothing applied" in c.value for c in at.caption)
        calls: list = []
        monkeypatch.setattr(
            svc,
            "load_and_publish",
            lambda **kw: calls.append(kw) or {"bcda.patient": 5},
        )
        _button(at, "Reconcile now (1 pending)").click().run()
        run = fresh_runner.wait(fresh_runner.all()[0].id)
        assert run.state == "done" and run.result["applied"] == ["B1"]
        assert calls == [{"run_pipeline": True, "skip_flatten": True}]
        assert sync.read_ledger(tmp_store)[0]["job_key"] == "B1"
        # the ledger renders, and the inflated page still has no pending work
        at = _page("sync").run()
        assert not at.exception
        assert any("Reconcile now" == b.label for b in at.button)
        assert any("patient" in d.value.to_string() for d in at.dataframe)
        # a failed ledger row surfaces as an error
        sync._append_ledger(
            tmp_store,
            {
                "applied_seq": 2,
                "job_key": "BAD",
                "kind": "delta",
                "endpoint": "Group/all",
                "type_filter": "",
                "types": "Patient",
                "since": "x",
                "transaction_time": "",
                "transaction_source": "job_row",
                "applied_at": "",
                "ok": False,
                "counts": "{}",
                "notes": '["scope differs"]',
                "error": "SyncInvariantError: boom",
            },
        )
        at = _page("sync").run()
        assert any("boom" in e.value for e in at.error)
        assert any("note(s)" in e.label for e in at.expander)


def test_app_shell_renders() -> None:
    at = AppTest.from_file(
        str(Path(views.__file__).with_name("app.py")), default_timeout=60
    ).run()
    assert not at.exception
    assert any("not connected" in c.value for c in at.sidebar.caption)
    # masthead first, then one page link per page, in order
    assert any("fw-mast" in h.value for h in at.sidebar.html)
    assert [p.label for p in at.sidebar.page_link] == [
        "Connect",
        "Exports",
        "Store",
        "Flatten + load",
        "Sync",
        "Data",
        "Dictionary + crosswalk",
        "Upstream",
    ]
    at.session_state["conn"] = SMALL
    at.run()
    assert any("Small Model Entity" in c.value for c in at.sidebar.caption)