File size: 25,011 Bytes
06555b5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
"""
SQLite as alternative storage backend for a TableGroup's data.

For the most part, translation of a TableGroup's tableSchema to SQL works as expected:

- each table is converted to a `CREATE TABLE` statement
- each column specifies a column in the corresponding `CREATE TABLE` statement
- `foreignKey` constraints are added according to the corresponding `tableSchema` property.

List-valued foreignKeys are supported as follows: For each pair of tables related through a
list-valued foreign key, an association table is created. To make it possible to distinguish
multiple list-valued foreign keys between the same two tables, the association table has
a column `context`, which stores the name of the foreign key column from which a row in the
assocation table was created.

Other list-valued columns work in two different ways: If the atomic datatype is `string`, the
specified separator is used to create a concatenated string representation in the database field.
Otherwise, the list of values is serialized as JSON.

SQL table and column names can be customized by passing a translator callable when instantiating
a :class:`Database`.

SQLite support has the following limitations:

- regex constraints on strings (as specified via a :class:`csvw.Datatype`'s format attribute) are
  not enforced by the database.
"""
import json
import typing
import decimal
import pathlib
import sqlite3
import functools
import contextlib
import collections

import attr

import csvw
from csvw.datatypes import DATATYPES
from csvw.metadata import TableGroup


def identity(s):
    return s


TYPE_MAP = {
    'string': (
        'TEXT',
        identity,
        identity),
    'integer': (
        'INTEGER',
        identity,
        identity),
    'boolean': (
        'INTEGER',
        lambda s: s if s is None else int(s),
        lambda s: s if s is None else bool(s)),
    'decimal': (
        'REAL',
        lambda s: s if s is None else float(s),
        lambda s: s if s is None else decimal.Decimal(s)),
    'hexBinary': (
        'BLOB',
        identity,
        identity),
}


class SchemaTranslator(typing.Protocol):
    def __call__(self, table: str, column: typing.Optional[str] = None) -> str:
        ...  # pragma: no cover


class ColumnTranslator(typing.Protocol):
    def __call__(self, column: str) -> str:
        ...  # pragma: no cover


def quoted(*names):
    return ','.join('`{0}`'.format(name) for name in names)


def insert(db: sqlite3.Connection,
           translate: SchemaTranslator,
           table: str,
           keys: typing.Sequence[str],
           *rows: list,
           single: typing.Optional[bool] = False):
    """
    Insert a sequence of rows into a table.

    :param db: Database connection.
    :param translate: Callable translating table and column names to proper schema object names.
    :param table: Untranslated table name.
    :param keys: Untranslated column names.
    :param rows: Sequence of rows to insert.
    :param single: Flag signaling whether to insert all rows at once using `executemany` or one at \
    a time, allowing for more focused debugging output in case of errors.
    """
    if rows:
        sql = "INSERT INTO {0} ({1}) VALUES ({2})".format(
            quoted(translate(table)),
            quoted(*[translate(table, k) for k in keys]),
            ','.join(['?' for _ in keys]))
        try:
            db.executemany(sql, rows)
        except:  # noqa: E722 - this is purely for debugging.
            if not single:
                for row in rows:
                    insert(db, translate, table, keys, row, single=True)
            else:
                print(sql)
                print(rows)
                raise


def select(db: sqlite3.Connection, table: str) -> typing.Tuple[typing.List[str], typing.Sequence]:
    cu = db.execute("SELECT * FROM {0}".format(quoted(table)))
    cols = [d[0] for d in cu.description]
    return cols, list(cu.fetchall())


@attr.s
class ColSpec:
    """
    A `ColSpec` captures sufficient information about a :class:`csvw.Column` for the DB schema.
    """
    name = attr.ib()
    csvw_type = attr.ib(default='string', converter=lambda s: s if s else 'string')
    separator = attr.ib(default=None)
    db_type = attr.ib(default=None)
    convert = attr.ib(default=None)
    read = attr.ib(default=None)
    required = attr.ib(default=False)
    csvw = attr.ib(default=None)

    def __attrs_post_init__(self):
        if self.csvw_type in TYPE_MAP:
            self.db_type, self.convert, self.read = TYPE_MAP[self.csvw_type]
        else:
            self.db_type = 'TEXT'
            self.convert = DATATYPES[self.csvw_type].to_string
            self.read = DATATYPES[self.csvw_type].to_python
        if self.separator and self.db_type != 'TEXT':
            self.db_type = 'TEXT'

    def check(self, translate: ColumnTranslator) -> typing.Optional[str]:
        """
        We try to convert as many data constraints as possible into SQLite CHECK constraints.

        :param translate: Callable to translate column names between CSVW metadata and DB schema.
        :return: A string suitable as argument of an SQL CHECK constraint.
        """
        if not self.csvw:
            return
        c, cname = self.csvw, translate(self.name)
        constraints = []
        if (c.minimum is not None) or (c.maximum is not None):
            func = {
                'date': 'date',
                'datetime': 'datetime',
            }.get(self.csvw_type)
            if c.minimum is not None:
                if func:
                    constraints.append("{2}(`{0}`) >= {2}('{1}')".format(cname, c.minimum, func))
                else:
                    constraints.append('`{0}` >= {1}'.format(cname, c.minimum))
            if c.maximum is not None:
                if func:
                    constraints.append("{2}(`{0}`) <= {2}('{1}')".format(cname, c.maximum, func))
                else:
                    constraints.append('`{0}` <= {1}'.format(cname, c.maximum))
        elif any(cc is not None for cc in [c.length, c.minLength, c.maxLength]):
            if c.length:
                constraints.append('length(`{0}`) = {1}'.format(cname, c.length))
            if c.minLength:
                constraints.append('length(`{0}`) >= {1}'.format(cname, c.minLength))
            if c.maxLength:
                constraints.append('length(`{0}`) <= {1}'.format(cname, c.maxLength))
        return ' AND '.join(constraints)

    def sql(self, translate: ColumnTranslator) -> str:
        _check = self.check(translate)
        return '`{0}` {1}{2}{3}'.format(
            translate(self.name),
            self.db_type,
            ' NOT NULL' if self.required else '',
            ' CHECK ({0})'.format(_check) if _check else '')


@attr.s
class TableSpec(object):
    """
    A `TableSpec` captures sufficient information about a :class:`csvw.Table` for the DB schema.

    .. note::

        We support "light-weight" many-to-many relationships by allowing list-valued foreign key
        columns in CSVW. In the database these columns are turned into an associative table, adding
        the name of the column as value a `context` column. Thus, multiple columns in a table my be
        specified as targets of many-to-many relations with the same table.

        .. seealso:: `<https://en.wikipedia.org/wiki/Associative_entity>`_
    """
    name = attr.ib()
    columns = attr.ib(default=attr.Factory(list))
    foreign_keys = attr.ib(default=attr.Factory(list))
    many_to_many = attr.ib(default=attr.Factory(collections.OrderedDict))
    primary_key = attr.ib(default=None)

    @classmethod
    def from_table_metadata(cls,
                            table: csvw.Table,
                            drop_self_referential_fks: typing.Optional[bool] = True) -> 'TableSpec':
        """
        Create a `TableSpec` from the schema description of a `csvw.metadata.Table`.

        :param table: `csvw.metadata.Table` instance.
        :param drop_self_referential_fks: Flag signaling whether to drop self-referential foreign \
        keys. This may be necessary, if the order of rows in a CSVW table does not guarantee \
        referential integrity when inserted in order (e.g. an eralier row refering to a later one).
        :return: `TableSpec` instance.
        """
        spec = cls(name=table.local_name, primary_key=table.tableSchema.primaryKey)
        list_valued = {c.header for c in table.tableSchema.columns if c.separator}
        for fk in table.tableSchema.foreignKeys:
            # We only support Foreign Key references between tables!
            if not fk.reference.schemaReference:
                if len(fk.columnReference) == 1 and fk.columnReference[0] in list_valued:
                    # List-valued foreign keys are turned into a many-to-many relation!
                    assert len(fk.reference.columnReference) == 1, \
                        'Composite key {0} in table {1} referenced'.format(
                            fk.reference.columnReference,
                            fk.reference.resource)
                    assert spec.primary_key and len(spec.primary_key) == 1, \
                        'Table {0} referenced by list-valued foreign key must have non-composite ' \
                        'primary key'.format(spec.name)
                    spec.many_to_many[fk.columnReference[0]] = TableSpec.association_table(
                        spec.name,
                        spec.primary_key[0],
                        fk.reference.resource.string,
                        fk.reference.columnReference[0],
                    )
                elif not (drop_self_referential_fks and fk.reference.resource.string == spec.name):
                    spec.foreign_keys.append((
                        sorted(fk.columnReference),
                        fk.reference.resource.string,
                        sorted(fk.reference.columnReference),
                    ))
        for c in table.tableSchema.columns:
            if c.header not in spec.many_to_many:
                datatype = c.inherit('datatype')
                spec.columns.append(ColSpec(
                    name=c.header,
                    csvw_type=datatype.base if datatype else datatype,
                    separator=c.inherit('separator'),
                    required=c.inherit('required'),
                    csvw=c.inherit('datatype'),
                ))
        return spec

    @classmethod
    def association_table(cls, atable, apk, btable, bpk) -> 'TableSpec':
        """
        List-valued foreignKeys are supported as follows: For each pair of tables related through a
        list-valued foreign key, an association table is created. To make it possible to distinguish
        multiple list-valued foreign keys between the same two tables, the association table has
        a column `context`, which stores the name of the foreign key column from which a row in the
        assocation table was created.
        """
        afk = ColSpec('{0}_{1}'.format(atable, apk))
        bfk = ColSpec('{0}_{1}'.format(btable, bpk))
        if afk.name == bfk.name:
            afk.name += '_1'
            bfk.name += '_2'
        return cls(
            name='{0}_{1}'.format(atable, btable),
            columns=[afk, bfk, ColSpec('context')],
            foreign_keys=[
                ([afk.name], atable, [apk]),
                ([bfk.name], btable, [bpk]),
            ]
        )

    def sql(self, translate: SchemaTranslator) -> str:
        """
        :param translate:
        :return: The SQL statement to create the table.
        """
        col_translate = functools.partial(translate, self.name)
        clauses = [col.sql(col_translate) for col in self.columns]
        if self.primary_key:
            clauses.append('PRIMARY KEY({0})'.format(quoted(
                *[col_translate(c) for c in self.primary_key])))
        for fk, ref, refcols in self.foreign_keys:
            clauses.append('FOREIGN KEY({0}) REFERENCES {1}({2}) ON DELETE CASCADE'.format(
                quoted(*[col_translate(c) for c in fk]),
                quoted(translate(ref)),
                quoted(*[translate(ref, c) for c in refcols])))
        return "CREATE TABLE IF NOT EXISTS `{0}` (\n    {1}\n)".format(
            translate(self.name), ',\n    '.join(clauses))


def schema(tg: csvw.TableGroup,
           drop_self_referential_fks: typing.Optional[bool] = True) -> typing.List[TableSpec]:
    """
    Convert the table and column descriptions of a `TableGroup` into specifications for the
    DB schema.

    :param tg: CSVW TableGroup.
    :param drop_self_referential_fks: Flag signaling whether to drop self-referential foreign \
    keys. This may be necessary, if the order of rows in a CSVW table does not guarantee \
    referential integrity when inserted in order (e.g. an eralier row refering to a later one).
    :return: A pair (tables, reference_tables).
    """
    tables = {}
    for tname, table in tg.tabledict.items():
        t = TableSpec.from_table_metadata(
            table, drop_self_referential_fks=drop_self_referential_fks)
        tables[t.name] = t
        for at in t.many_to_many.values():
            tables[at.name] = at

    # We must determine the order in which tables must be created!
    ordered = collections.OrderedDict()
    i = 0

    # We loop through the tables repeatedly, and whenever we find one, which has all
    # referenced tables already in ordered, we move it from tables to ordered.
    while tables and i < 100:
        i += 1
        for table in list(tables.keys()):
            if all((ref[1] in ordered) or ref[1] == table for ref in tables[table].foreign_keys):
                # All referenced tables are already created (or self-referential).
                ordered[table] = tables.pop(table)
                break
    if tables:  # pragma: no cover
        raise ValueError('there seem to be cyclic dependencies between the tables')

    return list(ordered.values())


class Database(object):
    """
    Represents a SQLite database associated with a :class:`csvw.TableGroup` instance.

    :param tg: `TableGroup` instance defining the schema of the database.
    :param fname: Path to which to write the database file.
    :param translate: Schema object name translator.
    :param drop_self_referential_fks: Flag signaling whether to drop or enforce self-referential \
    foreign-key constraints.

    .. warning::

        We write rows of a table to the database sequentially. Since CSVW does not require ordering
        rows in tables such that self-referential foreign-key constraints are satisfied at each row,
        we don't enforce self-referential foreign-keys by default in order to not trigger "false"
        integrity errors. If data in a CSVW Table is known to be ordered appropriately, `False`
        should be passed as `drop_self_referential_fks` keyword parameter to enforce
        self-referential foreign-keys.
    """
    def __init__(
            self,
            tg: TableGroup,
            fname: typing.Optional[typing.Union[pathlib.Path, str]] = None,
            translate: typing.Optional[SchemaTranslator] = None,
            drop_self_referential_fks: typing.Optional[bool] = True,
    ):
        self.translate = translate or Database.name_translator
        self.fname = pathlib.Path(fname) if fname else None
        self.init_schema(tg, drop_self_referential_fks=drop_self_referential_fks)
        self._connection = None  # For in-memory dbs we need to keep the connection!

    def init_schema(self, tg, drop_self_referential_fks=True):
        self.tg = tg
        self.tables = schema(
            self.tg, drop_self_referential_fks=drop_self_referential_fks) if self.tg else []

    @property
    def tdict(self) -> typing.Dict[str, TableSpec]:
        return {t.name: t for t in self.tables}

    @staticmethod
    def name_translator(table: str, column: typing.Optional[str] = None) -> str:
        """
        A callable with this signature can be passed into DB creation to control the names
        of the schema objects.

        :param table: CSVW name of the table before translation
        :param column: CSVW name of a column of `table` before translation
        :return: Translated table name if `column is None` else translated column name
        """
        # By default, no translation is done:
        return column or table

    def connection(self) -> typing.Union[sqlite3.Connection, contextlib.closing]:
        if self.fname:
            return contextlib.closing(sqlite3.connect(str(self.fname)))
        if not self._connection:
            self._connection = sqlite3.connect(':memory:')
        return self._connection

    def select_many_to_many(self, db, table, context) -> dict:
        if context is not None:
            context_sql = "WHERE context = '{0}'".format(context)
        else:
            context_sql = ''
        sql = """\
SELECT {0}, group_concat({1}, ' '), group_concat(COALESCE(context, ''), '||')
FROM {2} {3} GROUP BY {0}""".format(
                quoted(self.translate(table.name, table.columns[0].name)),
                quoted(self.translate(table.name, table.columns[1].name)),
                quoted(self.translate(table.name)),
                context_sql)
        cu = db.execute(sql)
        return {
            r[0]: [(k, v) if context is None else k
                   for k, v in zip(r[1].split(), r[2].split('||'))] for r in cu.fetchall()}

    def separator(self, tname: str, cname: str) -> typing.Optional[str]:
        """
        :return: separator for the column specified by db schema names `tname` and `cname`.
        """
        for name in self.tdict:
            if self.translate(name) == tname:
                for col in self.tdict[name].columns:
                    if self.translate(name, col.name) == cname:
                        return col.separator

    def split_value(self, tname, cname, value) -> typing.Union[typing.List[str], str, None]:
        sep = self.separator(tname, cname)
        return (value or '').split(sep) if sep else value

    def read(self) -> typing.Dict[str, typing.List[typing.OrderedDict]]:
        """
        :return: A `dict` where keys are SQL table names corresponding to CSVW tables and values \
        are lists of rows, represented as dicts where keys are the SQL column names.
        """
        res = collections.defaultdict(list)
        with self.connection() as conn:
            for tname in self.tg.tabledict:
                #
                # FIXME: how much do we want to use DB types? Probably as much as possible!
                # Thus we need to convert on write **and** read!
                #
                convert, seps, refs = {}, {}, collections.defaultdict(dict)
                table = self.tdict[tname]  # The TableSpec object.

                # Assemble the conversion dictionary:
                for col in table.columns:
                    convert[self.translate(tname, col.name)] = [col.name, identity]
                    if col.csvw_type in TYPE_MAP:
                        convert[self.translate(tname, col.name)][1] = TYPE_MAP[col.csvw_type][2]
                    else:
                        convert[self.translate(tname, col.name)][1] = \
                            DATATYPES[col.csvw_type].to_python
                    if col.separator:
                        if col.csvw_type == 'string':
                            seps[self.translate(tname, col.name)] = col.separator
                        else:
                            seps[self.translate(tname, col.name)] = 'json'

                # Retrieve the many-to-many relations:
                for col, at in table.many_to_many.items():
                    for pk, v in self.select_many_to_many(conn, at, col).items():
                        refs[pk][self.translate(tname, col)] = v

                cols, rows = select(conn, self.translate(tname))
                for row in rows:
                    d = collections.OrderedDict()
                    for k, v in zip(cols, row):
                        if k in seps:
                            if v is None:
                                d[k] = None
                            elif not v:
                                d[k] = []
                            elif seps[k] == 'json':
                                d[k] = json.loads(v)
                            else:
                                d[k] = [convert[k][1](v_) for v_ in (v or '').split(seps[k])]
                        else:
                            d[k] = convert[k][1](v) if v is not None else None
                    pk = d[self.translate(tname, table.primary_key[0])] \
                        if table.primary_key and len(table.primary_key) == 1 else None
                    d.update({k: [] for k in table.many_to_many})
                    d.update(refs.get(pk, {}))
                    res[self.translate(tname)].append(d)
        return res

    def association_table_context(self, table, column, fkey):
        """
        Context for association tables is created calling this method.

        Note: If a custom value for the `context` column is created by overwriting this method,
        `select_many_to_many` must be adapted accordingly, to make sure the custom
        context is retrieved when reading the data from the db.

        :param table:
        :param column:
        :param fkey:
        :return: a pair (foreign key, context)
        """
        # The default implementation takes the column name as context:
        return fkey, column

    def write_from_tg(self, _force=False, _exists_ok=False, _skip_extra=False):
        return self.write(
            force=_force,
            _exists_ok=_exists_ok,
            _skip_extra=_skip_extra,
            **self.tg.read())

    def write(self, *, force=False, _exists_ok=False, _skip_extra=False, **items):
        """
        Creates a db file with the core schema.

        :param force: If `True` an existing db file will be overwritten.
        """
        if self.fname and self.fname.exists():
            if not force:
                raise ValueError('db file already exists, use force=True to overwrite')
            else:
                self.fname.unlink()

        with self.connection() as db:
            for table in self.tables:
                db.execute(table.sql(translate=self.translate))

            db.execute('PRAGMA foreign_keys = ON;')
            db.commit()

            refs = collections.defaultdict(list)  # collects rows in association tables.
            for t in self.tables:
                if t.name not in items:
                    continue
                rows, keys = [], []
                cols = {c.name: c for c in t.columns}
                for i, row in enumerate(items[t.name]):
                    pk = row[t.primary_key[0]] \
                        if t.primary_key and len(t.primary_key) == 1 else None
                    values = []
                    for k, v in row.items():
                        if k in t.many_to_many:
                            assert pk
                            at = t.many_to_many[k]
                            atkey = tuple([at.name] + [c.name for c in at.columns])
                            # We distinguish None - meaning NULL - and [] - meaning no items - as
                            # values of list-valued columns.
                            for vv in (v or []):
                                fkey, context = self.association_table_context(t, k, vv)
                                refs[atkey].append((pk, fkey, context))
                        else:
                            if k not in cols:
                                if _skip_extra:
                                    continue
                                else:
                                    raise ValueError(
                                        'unspecified column {0} found in data'.format(k))
                            col = cols[k]
                            if isinstance(v, list):
                                # Note: This assumes list-valued columns are of datatype string!
                                if col.csvw_type == 'string':
                                    v = (col.separator or ';').join(
                                        col.convert(vv) or '' for vv in v)
                                else:
                                    v = json.dumps(v)
                            else:
                                v = col.convert(v) if v is not None else None
                            if i == 0:
                                keys.append(col.name)
                            values.append(v)
                    rows.append(tuple(values))
                insert(db, self.translate, t.name, keys, *rows)

            for atkey, rows in refs.items():
                insert(db, self.translate, atkey[0], atkey[1:], *rows)

            db.commit()