Initial commit

This commit is contained in:
2026-11-16 16:29:32 +01:00
commit 97e5c9ef4d
37 changed files with 5932 additions and 0 deletions
+167
View File
@@ -0,0 +1,167 @@
"""Tests fuer Datenbank, Repository, Migrationen und Suchindex."""
from __future__ import annotations
from pathlib import Path
from sjaudiosearch.database.migrations import current_version, migrate
from sjaudiosearch.database.repository import Repository
def test_schema_and_migration_version(repo):
assert current_version(repo.conn) == 1
# Migration ist idempotent.
assert migrate(repo.conn) == 1
tables = {
row["name"]
for row in repo.conn.execute(
"SELECT name FROM sqlite_master WHERE type IN ('table', 'view')"
)
}
for table in (
"audio_files",
"transcripts",
"segments",
"words",
"search_index",
"search_trigram",
):
assert table in tables
def test_transcript_stored_with_segments_and_words(repo, add_audio, store_transcript):
record = add_audio("/tmp/folge1.mp3", duration=125.0)
store_transcript(
record.id,
[("Der Professor sah sich um.", 0.0, 4.0), ("Der Schlüssel liegt im alten Haus.", 4.0, 9.5)],
with_words=True,
duration=9.5,
)
stored = repo.get_transcript(record.id)
assert stored is not None
assert stored["model"] == "test-model"
assert stored["language"] == "de"
assert "Schlüssel" in stored["full_text"]
segments = repo.get_segments(record.id)
assert len(segments) == 2
assert segments[1]["start_time"] == 4.0
words = repo.get_words([int(segment["id"]) for segment in segments])
first_segment_words = words[int(segments[0]["id"])]
assert [row["word"] for row in first_segment_words][:2] == ["Der", "Professor"]
updated = repo.get_audio(record.id)
assert updated.status == "done"
assert updated.duration == 125.0
def test_status_management_and_retry(repo, add_audio):
ok = add_audio("/tmp/a.mp3")
failing = add_audio("/tmp/b.mp3")
repo.set_status(ok.id, "processing")
repo.set_status(failing.id, "error", "Decoder kaputt")
counts = repo.count_by_status()
assert counts["processing"] == 1
assert counts["error"] == 1
# Abgestuerzte 'processing'-Eintraege werden zurueckgesetzt
assert repo.reset_stale_processing() == 1
assert repo.get_audio(ok.id).status == "pending"
assert repo.reset_errors() == 1
reset = repo.get_audio(failing.id)
assert reset.status == "pending"
assert reset.error_message is None
def test_queue_skips_duplicates_and_done(repo, add_audio):
done = add_audio("/tmp/done.mp3", status="done")
pending = add_audio("/tmp/pending.mp3")
duplicate = add_audio("/tmp/dup.mp3", duplicate_of=done.id)
queue = repo.transcription_queue()
assert [record.id for record in queue] == [pending.id]
forced = repo.transcription_queue(force=True)
assert {record.id for record in forced} == {done.id, pending.id}
assert duplicate.id not in {record.id for record in forced}
def test_delete_audio_keeps_file_and_removes_index(repo, add_audio, store_transcript, tmp_path):
audio_file = tmp_path / "folge.mp3"
audio_file.write_bytes(b"audio")
record = add_audio(audio_file)
store_transcript(record.id, ["Der Schlüssel liegt im alten Haus."])
assert repo.fts_candidates('"schlüssel"', 10)
assert repo.delete_audio(record.id) is True
assert audio_file.exists()
assert repo.get_audio(record.id) is None
assert repo.get_transcript(record.id) is None
assert repo.fts_candidates('"schlüssel"', 10) == []
def test_foreign_key_cascade(repo, add_audio, store_transcript):
record = add_audio("/tmp/a.mp3")
store_transcript(record.id, ["Erster Satz.", "Zweiter Satz."])
assert repo.get_segments(record.id)
with repo.conn:
repo.conn.execute("DELETE FROM audio_files WHERE id = ?", (record.id,))
segments = repo.conn.execute("SELECT COUNT(*) AS amount FROM segments").fetchone()
words = repo.conn.execute("SELECT COUNT(*) AS amount FROM words").fetchone()
transcripts = repo.conn.execute("SELECT COUNT(*) AS amount FROM transcripts").fetchone()
assert segments["amount"] == 0
assert words["amount"] == 0
assert transcripts["amount"] == 0
def test_reindex_preserves_transcripts(repo, add_audio, store_transcript):
record = add_audio("/tmp/a.mp3")
store_transcript(record.id, ["Der Schlüssel liegt im alten Haus.", "Niemand bemerkte es."])
before_transcript = dict(repo.get_transcript(record.id))
before_segments = [dict(row) for row in repo.get_segments(record.id)]
assert repo.fts_candidates('"schlüssel"', 10)
indexed = repo.rebuild_search_index()
assert indexed == 2
after_transcript = dict(repo.get_transcript(record.id))
after_segments = [dict(row) for row in repo.get_segments(record.id)]
assert before_transcript == after_transcript
assert before_segments == after_segments
assert repo.fts_candidates('"schlüssel"', 10)
def test_saving_new_transcript_replaces_old(repo, add_audio, store_transcript):
record = add_audio("/tmp/a.mp3")
store_transcript(record.id, ["Alter Text mit Schlüssel."])
store_transcript(record.id, ["Neuer Text ohne alles."])
transcript = repo.get_transcript(record.id)
assert transcript["full_text"] == "Neuer Text ohne alles."
assert len(repo.get_segments(record.id)) == 1
assert repo.fts_candidates('"schlüssel"', 10) == []
def test_database_size_and_stats(repo, add_audio, store_transcript):
first = add_audio("/tmp/a.mp3", duration=100.0)
second = add_audio("/tmp/b.mp3", duration=50.0, status="error")
store_transcript(first.id, ["Ein Satz."], duration=100.0)
stats = repo.stats(repo.databases_size())
assert stats.total_files == 2
assert stats.done == 1
assert stats.error == 1
assert stats.total_duration == 150.0
assert stats.transcripts == 1
assert stats.segment_count == 1
assert stats.database_size > 0
assert "test-model" in stats.models