Blame for webapp/test_reference_selector.py

"""Region selector: every model the registry lists (and whose file is
present) is offered in one Region list, and the page names the reference of
the chosen model; deprecated and preview models are not offered.

Model files are stand-ins (small files under the registry file names) and
loading and settings reading are stubbed, so this runs without checkpoints.
Expectations are derived from the registry, so a release that adds models
needs no change here.
"""
import html
import re
import sys
from pathlib import Path

import pytest

pytest.importorskip("httpx")
sys.path.insert(0, str(Path(__file__).resolve().parent.parent))

from fastapi.testclient import TestClient  # noqa: E402
from taxonova.models import list_models  # noqa: E402

import webapp.app as webapp_app  # noqa: E402
from webapp import model_catalog as mc  # noqa: E402

LISTED = [s for s in list_models() if s.status not in ("deprecated", "preview")]
HIDDEN = [s for s in list_models(include_hidden=True) if s.status in ("deprecated", "preview")]
SETTINGS = {"operating_point": None, "species_gate": False, "species_groups": "kept", "forward_primer": None}


def _options(page: str, select_id: str):
    block = re.search(rf'<select id="{select_id}"[^>]*>(.*?)</select>', page, re.S).group(1)
    return [(v, html.unescape(t).strip(), bool(sel))
            for v, sel, t in re.findall(r'<option value="([^"]*)"( selected)?>([^<]*)</option>', block)]


@pytest.fixture
def server(tmp_path, monkeypatch):
    """Start the app on a models dir that has every registry file; returns the page."""
    for spec in list_models(include_hidden=True):
        (tmp_path / spec.filename).write_bytes(b"stand-in")
    monkeypatch.setenv("TAXONOVA_WEBAPP_MODELS_DIR", str(tmp_path))
    loads = []
    monkeypatch.setattr(webapp_app, "load_model", lambda path: loads.append(path) or {"path": path})
    monkeypatch.setattr(webapp_app, "read_settings", lambda path: dict(SETTINGS))

    def start():
        webapp_app.startup()
        return TestClient(webapp_app.app).get("/").text

    start.loads = loads
    return start


def _ref(spec):
    return mc.reference_label(spec.reference, spec.reference_version)


def test_every_listed_model_is_offered_and_hidden_ones_are_not(server):
    page = server()
    assert list(webapp_app.app.state.models) == [s.id for s in LISTED]
    # one Region list, no reference selector
    assert 'id="reference"' not in page
    regions = _options(page, "model")
    operon_only = {_ref(s) for s in LISTED if all(mc.region_key(t.id) == "operon" for t in LISTED if _ref(t) == _ref(s))}
    amplicons = [k for k in mc.REGION_LABELS if k != "operon"]
    rank = dict({k: i for i, k in enumerate(amplicons)}, operon=len(amplicons) + 1)
    expected = [s.id for s in sorted(LISTED, key=lambda s: (_ref(s) in operon_only,
                                                            rank.get(mc.region_key(s.id), len(amplicons))))]
    assert [v for v, _, _ in regions] == expected
    assert regions[0] == ("v4", "V4", True) and sum(sel for _, _, sel in regions) == 1
    if {"v4v5", "v3v4", "16s", "operon"} <= {s.id for s in LISTED}:
        ids = [v for v, _, _ in regions]
        assert ids.index("v4") < ids.index("v4v5") < ids.index("v3v4") < ids.index("16s") < ids.index("operon")
    labels = [t for _, t, _ in regions]
    assert len(labels) == len(set(labels)), labels
    assert "Reference: " + _ref(mc.registry.get_model("v4")) in page
    for spec in HIDDEN:
        assert f'"{spec.id}"' not in page, spec.id
    visible = re.sub(r'<script id="cfg".*?</script>', "", page, flags=re.S)
    assert "preview" not in visible.lower()


def test_start_up_loads_the_default_model_only(server):
    server()
    assert len(server.loads) == 1 and server.loads[0].endswith(mc.registry.get_model("v4").filename)
    assert webapp_app.app.state.cache.loaded_ids() == ["v4"]


def test_every_region_has_a_plain_label():
    for m in mc.discover_models(Path("/nonexistent")):
        assert m.region_label != m.id or m.id in mc.REGION_LABELS, m.id
        assert "16S V" not in m.region_label, m.region_label
    labels = {m.id: m.region_label for m in mc.discover_models(Path("/nonexistent"))}
    if "v4v5" in labels:
        assert labels["v4v5"] == "V4-V5"
    # a region the web app has never seen is named from the registry entry
    assert mc.region_label("v1v3", "16S V1-V3") == "V1-V3"
    assert mc.region_label("v1v9", "16S V1-V9 (whole gene)") == "V1-V9"


def test_a_model_needs_its_file(server, tmp_path):
    victim = LISTED[-1]
    (tmp_path / victim.filename).unlink()
    server()
    assert victim.id not in webapp_app.app.state.models


def test_deprecated_ids_get_a_message_naming_the_command_line(server):
    server()
    deprecated = [s for s in HIDDEN if s.status == "deprecated"]
    if not deprecated:
        pytest.skip("this taxonova release lists no deprecated models")
    for spec in deprecated:
        with pytest.raises(webapp_app.InputError) as e:
            webapp_app.resolve_model(spec.id)
        msg = str(e.value)
        assert "deprecated" in msg and f"taxonova classify reads.fasta --model {spec.id}" in msg, msg
        assert f"taxonova=={mc.package_version()}" in msg
        # its earlier file names name the same deprecated model
        for alias in spec.legacy_filenames:
            with pytest.raises(webapp_app.InputError, match="deprecated"):
                webapp_app.resolve_model(alias)


#: ids and file names of the 26.11.0 V4-V5 models of the references withdrawn
#: in 26.11.1: the web app must not know them (no deprecated ids replace them)
WITHDRAWN = ("v4v5-silva", "v4v5-rdp", "v4v5-silva-26.11", "silva_v4v5", "rdp_v4v5",
             "taxonova-V4V5-silva-26.11.pt", "taxonova-V4V5-rdp-26.11.pt", "taxonova-V4V5-silva-26.11.1.pt")


def test_unknown_and_withdrawn_ids_are_refused(server):
    server()
    for bad in WITHDRAWN + ("v4-nonexistent", "../../etc/passwd"):
        assert mc.canonical_id(bad) is None, bad
        assert mc.unavailable_message(bad) is None, bad
        with pytest.raises(webapp_app.InputError, match="Unknown model"):
            webapp_app.resolve_model(bad)


def test_reference_labels():
    assert mc.reference_label("Greengenes2", "2024.09") == "Greengenes2 2024.09"
    assert mc.reference_label("GROND", "GTDB R207") == "GTDB R207 (GROND)"