|
| 1 | +""" |
| 2 | +Tests for skip_duplicates behavior with secondary unique constraints. |
| 3 | +
|
| 4 | +Verifies that skip_duplicates=True on PostgreSQL skips primary key |
| 5 | +duplicates while still raising on secondary unique constraint violations. |
| 6 | +Resolves #1049. |
| 7 | +""" |
| 8 | + |
| 9 | +import time |
| 10 | + |
| 11 | +import pytest |
| 12 | + |
| 13 | +import datajoint as dj |
| 14 | +from datajoint.errors import DuplicateError |
| 15 | + |
| 16 | + |
| 17 | +@pytest.fixture(scope="function") |
| 18 | +def schema_by_backend(connection_by_backend, db_creds_by_backend): |
| 19 | + """Create a fresh schema per test, parameterized across backends.""" |
| 20 | + backend = db_creds_by_backend["backend"] |
| 21 | + test_id = str(int(time.time() * 1000))[-8:] |
| 22 | + schema_name = f"djtest_skipdup_{backend}_{test_id}"[:64] |
| 23 | + |
| 24 | + if connection_by_backend.is_connected: |
| 25 | + try: |
| 26 | + connection_by_backend.query( |
| 27 | + f"DROP DATABASE IF EXISTS {connection_by_backend.adapter.quote_identifier(schema_name)}" |
| 28 | + ) |
| 29 | + except Exception: |
| 30 | + pass |
| 31 | + |
| 32 | + schema = dj.Schema(schema_name, connection=connection_by_backend) |
| 33 | + yield schema |
| 34 | + |
| 35 | + if connection_by_backend.is_connected: |
| 36 | + try: |
| 37 | + connection_by_backend.query( |
| 38 | + f"DROP DATABASE IF EXISTS {connection_by_backend.adapter.quote_identifier(schema_name)}" |
| 39 | + ) |
| 40 | + except Exception: |
| 41 | + pass |
| 42 | + |
| 43 | + |
| 44 | +def test_skip_duplicates_pk_match(schema_by_backend): |
| 45 | + """skip_duplicates=True silently skips rows whose PK already exists.""" |
| 46 | + |
| 47 | + @schema_by_backend |
| 48 | + class Item(dj.Manual): |
| 49 | + definition = """ |
| 50 | + item_id : int |
| 51 | + --- |
| 52 | + name : varchar(100) |
| 53 | + email : varchar(100) |
| 54 | + unique index (email) |
| 55 | + """ |
| 56 | + |
| 57 | + Item.insert1(dict(item_id=1, name="Alice", email="alice@example.com")) |
| 58 | + |
| 59 | + # Same PK, different values — should be silently skipped |
| 60 | + Item.insert1( |
| 61 | + dict(item_id=1, name="Bob", email="bob@example.com"), |
| 62 | + skip_duplicates=True, |
| 63 | + ) |
| 64 | + |
| 65 | + # Original row unchanged |
| 66 | + row = (Item & "item_id=1").fetch1() |
| 67 | + assert row["name"] == "Alice" |
| 68 | + assert row["email"] == "alice@example.com" |
| 69 | + |
| 70 | + |
| 71 | +def test_skip_duplicates_unique_violation_raises_on_postgres(schema_by_backend, db_creds_by_backend): |
| 72 | + """On PostgreSQL, skip_duplicates=True still raises on secondary unique violations. |
| 73 | +
|
| 74 | + Regression test for #1049: a row with a *new* PK but a *conflicting* |
| 75 | + secondary unique index value must raise DuplicateError on PostgreSQL. |
| 76 | + """ |
| 77 | + if db_creds_by_backend["backend"] != "postgresql": |
| 78 | + pytest.skip("PostgreSQL-specific: ON CONFLICT (pk) DO NOTHING preserves unique constraints") |
| 79 | + |
| 80 | + @schema_by_backend |
| 81 | + class Item(dj.Manual): |
| 82 | + definition = """ |
| 83 | + item_id : int |
| 84 | + --- |
| 85 | + name : varchar(100) |
| 86 | + email : varchar(100) |
| 87 | + unique index (email) |
| 88 | + """ |
| 89 | + |
| 90 | + Item.insert1(dict(item_id=1, name="Alice", email="alice@example.com")) |
| 91 | + |
| 92 | + # New PK (2) but email conflicts with existing row (1) |
| 93 | + with pytest.raises(DuplicateError): |
| 94 | + Item.insert1( |
| 95 | + dict(item_id=2, name="Bob", email="alice@example.com"), |
| 96 | + skip_duplicates=True, |
| 97 | + ) |
| 98 | + |
| 99 | + |
| 100 | +def test_skip_duplicates_unique_on_mysql(schema_by_backend, db_creds_by_backend): |
| 101 | + """On MySQL, skip_duplicates=True silently skips secondary unique conflicts. |
| 102 | +
|
| 103 | + Documents the known MySQL asymmetry: ON DUPLICATE KEY UPDATE catches |
| 104 | + all unique key conflicts, not just primary key. |
| 105 | + """ |
| 106 | + if db_creds_by_backend["backend"] != "mysql": |
| 107 | + pytest.skip("MySQL-specific: ON DUPLICATE KEY UPDATE catches all unique keys") |
| 108 | + |
| 109 | + @schema_by_backend |
| 110 | + class Item(dj.Manual): |
| 111 | + definition = """ |
| 112 | + item_id : int |
| 113 | + --- |
| 114 | + name : varchar(100) |
| 115 | + email : varchar(100) |
| 116 | + unique index (email) |
| 117 | + """ |
| 118 | + |
| 119 | + Item.insert1(dict(item_id=1, name="Alice", email="alice@example.com")) |
| 120 | + |
| 121 | + # New PK (2) but email conflicts — MySQL silently skips |
| 122 | + Item.insert1( |
| 123 | + dict(item_id=2, name="Bob", email="alice@example.com"), |
| 124 | + skip_duplicates=True, |
| 125 | + ) |
| 126 | + |
| 127 | + # Only the original row exists |
| 128 | + assert len(Item()) == 1 |
| 129 | + assert (Item & "item_id=1").fetch1()["name"] == "Alice" |
| 130 | + |
| 131 | + |
| 132 | +def test_skip_duplicates_no_unique_index(schema_by_backend): |
| 133 | + """skip_duplicates=True works normally on tables without secondary unique indexes.""" |
| 134 | + |
| 135 | + @schema_by_backend |
| 136 | + class Simple(dj.Manual): |
| 137 | + definition = """ |
| 138 | + item_id : int |
| 139 | + --- |
| 140 | + name : varchar(100) |
| 141 | + """ |
| 142 | + |
| 143 | + Simple.insert1(dict(item_id=1, name="Alice")) |
| 144 | + |
| 145 | + # Same PK, different name — silently skipped |
| 146 | + Simple.insert1(dict(item_id=1, name="Bob"), skip_duplicates=True) |
| 147 | + assert (Simple & "item_id=1").fetch1()["name"] == "Alice" |
| 148 | + |
| 149 | + # New PK — inserted |
| 150 | + Simple.insert1(dict(item_id=2, name="Bob"), skip_duplicates=True) |
| 151 | + assert len(Simple()) == 2 |
| 152 | + |
| 153 | + |
| 154 | +def test_skip_duplicates_composite_unique(schema_by_backend, db_creds_by_backend): |
| 155 | + """skip_duplicates=True with a composite secondary unique index.""" |
| 156 | + if db_creds_by_backend["backend"] != "postgresql": |
| 157 | + pytest.skip("PostgreSQL-specific unique constraint enforcement") |
| 158 | + |
| 159 | + @schema_by_backend |
| 160 | + class Record(dj.Manual): |
| 161 | + definition = """ |
| 162 | + record_id : int |
| 163 | + --- |
| 164 | + first_name : varchar(100) |
| 165 | + last_name : varchar(100) |
| 166 | + data : varchar(255) |
| 167 | + unique index (first_name, last_name) |
| 168 | + """ |
| 169 | + |
| 170 | + Record.insert1(dict(record_id=1, first_name="Alice", last_name="Smith", data="v1")) |
| 171 | + |
| 172 | + # New PK but composite unique (first_name, last_name) conflicts |
| 173 | + with pytest.raises(DuplicateError): |
| 174 | + Record.insert1( |
| 175 | + dict(record_id=2, first_name="Alice", last_name="Smith", data="v2"), |
| 176 | + skip_duplicates=True, |
| 177 | + ) |
| 178 | + |
| 179 | + |
| 180 | +def test_skip_duplicates_batch_mixed(schema_by_backend, db_creds_by_backend): |
| 181 | + """Batch insert with skip_duplicates=True: PK duplicates skipped, unique conflicts raise.""" |
| 182 | + if db_creds_by_backend["backend"] != "postgresql": |
| 183 | + pytest.skip("PostgreSQL-specific unique constraint enforcement") |
| 184 | + |
| 185 | + @schema_by_backend |
| 186 | + class Item(dj.Manual): |
| 187 | + definition = """ |
| 188 | + item_id : int |
| 189 | + --- |
| 190 | + email : varchar(100) |
| 191 | + unique index (email) |
| 192 | + """ |
| 193 | + |
| 194 | + Item.insert1(dict(item_id=1, email="alice@example.com")) |
| 195 | + |
| 196 | + # Batch: row 2 is new (OK), row 1 is PK dup (skip), row 3 conflicts on email |
| 197 | + with pytest.raises(DuplicateError): |
| 198 | + Item.insert( |
| 199 | + [ |
| 200 | + dict(item_id=2, email="bob@example.com"), |
| 201 | + dict(item_id=1, email="duplicate-pk@example.com"), # PK dup — skipped |
| 202 | + dict(item_id=3, email="alice@example.com"), # unique conflict — error |
| 203 | + ], |
| 204 | + skip_duplicates=True, |
| 205 | + ) |
0 commit comments