database.py 24.8 KB
Newer Older
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
#!/usr/bin/env python
# vim: set fileencoding=utf-8 :

###############################################################################
#                                                                             #
# Copyright (c) 2016 Idiap Research Institute, http://www.idiap.ch/           #
# Contact: beat.support@idiap.ch                                              #
#                                                                             #
# This file is part of the beat.backend.python module of the BEAT platform.   #
#                                                                             #
# Commercial License Usage                                                    #
# Licensees holding valid commercial BEAT licenses may use this file in       #
# accordance with the terms contained in a written agreement between you      #
# and Idiap. For further information contact tto@idiap.ch                     #
#                                                                             #
# Alternatively, this file may be used under the terms of the GNU Affero      #
# Public License version 3 as published by the Free Software and appearing    #
# in the file LICENSE.AGPL included in the packaging of this file.            #
# The BEAT platform is distributed in the hope that it will be useful, but    #
# WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY  #
# or FITNESS FOR A PARTICULAR PURPOSE.                                        #
#                                                                             #
# You should have received a copy of the GNU Affero Public License along      #
# with the BEAT platform. If not, see http://www.gnu.org/licenses/.           #
#                                                                             #
###############################################################################


"""Validation of databases"""

import os
import sys

import six
import simplejson
36 37
import itertools
import numpy as np
38
from collections import namedtuple
39 40

from . import loader
Philip ABBET's avatar
Philip ABBET committed
41 42 43
from . import utils

from .dataformat import DataFormat
44
from .outputs import OutputList
Philip ABBET's avatar
Philip ABBET committed
45 46


47 48
#----------------------------------------------------------

Philip ABBET's avatar
Philip ABBET committed
49 50

class Storage(utils.CodeStorage):
Philip ABBET's avatar
Philip ABBET committed
51
    """Resolves paths for databases
Philip ABBET's avatar
Philip ABBET committed
52

Philip ABBET's avatar
Philip ABBET committed
53
    Parameters:
Philip ABBET's avatar
Philip ABBET committed
54

Philip ABBET's avatar
Philip ABBET committed
55
      prefix (str): Establishes the prefix of your installation.
Philip ABBET's avatar
Philip ABBET committed
56

Philip ABBET's avatar
Philip ABBET committed
57 58
      name (str): The name of the database object in the format
        ``<name>/<version>``.
Philip ABBET's avatar
Philip ABBET committed
59

Philip ABBET's avatar
Philip ABBET committed
60
    """
Philip ABBET's avatar
Philip ABBET committed
61

Philip ABBET's avatar
Philip ABBET committed
62
    def __init__(self, prefix, name):
Philip ABBET's avatar
Philip ABBET committed
63

Philip ABBET's avatar
Philip ABBET committed
64 65
        if name.count('/') != 1:
            raise RuntimeError("invalid database name: `%s'" % name)
Philip ABBET's avatar
Philip ABBET committed
66

Philip ABBET's avatar
Philip ABBET committed
67 68
        self.name, self.version = name.split('/')
        self.fullname = name
Philip ABBET's avatar
Philip ABBET committed
69

Philip ABBET's avatar
Philip ABBET committed
70 71
        path = os.path.join(prefix, 'databases', name)
        super(Storage, self).__init__(path, 'python') #views are coded in Python
72 73


74 75
#----------------------------------------------------------

76

77
class Runner(object):
Philip ABBET's avatar
Philip ABBET committed
78
    '''A special loader class for database views, with specialized methods
79

Philip ABBET's avatar
Philip ABBET committed
80
    Parameters:
81

Philip ABBET's avatar
Philip ABBET committed
82
      db_name (str): The full name of the database object for this view
83

Philip ABBET's avatar
Philip ABBET committed
84 85
      module (module): The preloaded module containing the database views as
        returned by :py:func:`beat.core.loader.load_module`.
86

Philip ABBET's avatar
Philip ABBET committed
87
      prefix (str, path): The prefix path for the current installation
88

Philip ABBET's avatar
Philip ABBET committed
89 90
      root_folder (str, path): The path pointing to the root folder of this
        database
91

Philip ABBET's avatar
Philip ABBET committed
92 93 94
      exc (class): The class to use as base exception when translating the
        exception from the user code. Read the documention of :py:func:`run`
        for more details.
95

Philip ABBET's avatar
Philip ABBET committed
96
      *args: Constructor parameters for the database view. Normally, none.
97

Philip ABBET's avatar
Philip ABBET committed
98
      **kwargs: Constructor parameters for the database view. Normally, none.
99

Philip ABBET's avatar
Philip ABBET committed
100
    '''
101

102
    def __init__(self, module, definition, prefix, root_folder, exc=None):
103

Philip ABBET's avatar
Philip ABBET committed
104 105 106 107 108 109 110
        try:
            class_ = getattr(module, definition['view'])
        except Exception as e:
            if exc is not None:
                type, value, traceback = sys.exc_info()
                six.reraise(exc, exc(value), traceback)
            else:
111
                raise # just re-raise the user exception
112

113 114 115 116 117 118 119
        self.obj          = loader.run(class_, '__new__', exc)
        self.ready        = False
        self.prefix       = prefix
        self.root_folder  = root_folder
        self.definition   = definition
        self.exc          = exc or RuntimeError
        self.data_sources = None
120 121


122 123
    def index(self, filename):
        '''Index the content of the view'''
124

125
        parameters = self.definition.get('parameters', {})
126

127
        objs = loader.run(self.obj, 'index', self.exc, self.root_folder, parameters)
128

129 130
        if not isinstance(objs, list):
            raise self.exc("index() didn't return a list")
131

132 133
        if not os.path.exists(os.path.dirname(filename)):
            os.makedirs(os.path.dirname(filename))
134

135 136
        with open(filename, 'wb') as f:
            simplejson.dump(objs, f)
137 138


139 140
    def setup(self, filename, start_index=None, end_index=None, pack=True):
        '''Sets up the view'''
141

142 143
        if self.ready:
            return
144

145 146
        with open(filename, 'rb') as f:
            objs = simplejson.load(f)
147

148 149
        Entry = namedtuple('Entry', sorted(objs[0].keys()))
        objs = [ Entry(**x) for x in objs ]
150

151
        parameters = self.definition.get('parameters', {})
152

153 154
        loader.run(self.obj, 'setup', self.exc, self.root_folder, parameters,
                   objs, start_index=start_index, end_index=end_index)
155 156


157 158 159
        # Create data sources for the outputs
        from .data import DatabaseOutputDataSource
        from .dataformat import DataFormat
160

161 162 163 164 165 166
        self.data_sources = {}
        for output_name, output_format in self.definition.get('outputs', {}).items():
            data_source = DatabaseOutputDataSource()
            data_source.setup(self, output_name, output_format, self.prefix,
                              start_index=start_index, end_index=end_index, pack=pack)
            self.data_sources[output_name] = data_source
167

168
        self.ready = True
169 170


171 172
    def get(self, output, index):
        '''Returns the data of the provided output at the provided index'''
173

Philip ABBET's avatar
Philip ABBET committed
174
        if not self.ready:
175
            raise self.exc("Database view not yet setup")
176

177
        return loader.run(self.obj, 'get', self.exc, output, index)
178 179


Philip ABBET's avatar
Philip ABBET committed
180 181 182
    def __getattr__(self, key):
        '''Returns an attribute of the view - only called at last resort'''
        return getattr(self.obj, key)
183 184


185 186 187 188
    def objects(self):
        return self.obj.objs


189 190
#----------------------------------------------------------

191 192

class Database(object):
Philip ABBET's avatar
Philip ABBET committed
193
    """Databases define the start point of the dataflow in an experiment.
194 195


Philip ABBET's avatar
Philip ABBET committed
196
    Parameters:
197

Philip ABBET's avatar
Philip ABBET committed
198
      prefix (str): Establishes the prefix of your installation.
199

Philip ABBET's avatar
Philip ABBET committed
200
      name (str): The fully qualified database name (e.g. ``db/1``)
201

Philip ABBET's avatar
Philip ABBET committed
202 203 204 205 206 207
      dataformat_cache (dict, optional): A dictionary mapping dataformat names
        to loaded dataformats. This parameter is optional and, if passed, may
        greatly speed-up database loading times as dataformats that are already
        loaded may be re-used. If you use this parameter, you must guarantee
        that the cache is refreshed as appropriate in case the underlying
        dataformats change.
208 209


Philip ABBET's avatar
Philip ABBET committed
210
    Attributes:
211

Philip ABBET's avatar
Philip ABBET committed
212
      name (str): The full, valid name of this database
213

Philip ABBET's avatar
Philip ABBET committed
214 215
      data (dict): The original data for this database, as loaded by our JSON
        decoder.
216

Philip ABBET's avatar
Philip ABBET committed
217
    """
218

Philip ABBET's avatar
Philip ABBET committed
219
    def __init__(self, prefix, name, dataformat_cache=None):
220

Philip ABBET's avatar
Philip ABBET committed
221 222 223 224
        self._name = None
        self.prefix = prefix
        self.dataformats = {} # preloaded dataformats
        self.storage = None
225

Philip ABBET's avatar
Philip ABBET committed
226 227
        self.errors = []
        self.data = None
228

Philip ABBET's avatar
Philip ABBET committed
229 230
        # if the user has not provided a cache, still use one for performance
        dataformat_cache = dataformat_cache if dataformat_cache is not None else {}
231

Philip ABBET's avatar
Philip ABBET committed
232
        self._load(name, dataformat_cache)
233 234


Philip ABBET's avatar
Philip ABBET committed
235 236
    def _load(self, data, dataformat_cache):
        """Loads the database"""
237

Philip ABBET's avatar
Philip ABBET committed
238
        self._name = data
Philip ABBET's avatar
Philip ABBET committed
239

Philip ABBET's avatar
Philip ABBET committed
240 241 242 243 244
        self.storage = Storage(self.prefix, self._name)
        json_path = self.storage.json.path
        if not self.storage.json.exists():
            self.errors.append('Database declaration file not found: %s' % json_path)
            return
Philip ABBET's avatar
Philip ABBET committed
245

Philip ABBET's avatar
Philip ABBET committed
246 247
        with open(json_path, 'rb') as f:
            self.data = simplejson.load(f)
Philip ABBET's avatar
Philip ABBET committed
248

Philip ABBET's avatar
Philip ABBET committed
249 250
        for protocol in self.data['protocols']:
            for _set in protocol['sets']:
Philip ABBET's avatar
Philip ABBET committed
251

Philip ABBET's avatar
Philip ABBET committed
252
                for key, value in _set['outputs'].items():
Philip ABBET's avatar
Philip ABBET committed
253

Philip ABBET's avatar
Philip ABBET committed
254 255
                    if value in self.dataformats:
                        continue
Philip ABBET's avatar
Philip ABBET committed
256

Philip ABBET's avatar
Philip ABBET committed
257 258 259 260 261
                    if value in dataformat_cache:
                        dataformat = dataformat_cache[value]
                    else:
                        dataformat = DataFormat(self.prefix, value)
                        dataformat_cache[value] = dataformat
Philip ABBET's avatar
Philip ABBET committed
262

Philip ABBET's avatar
Philip ABBET committed
263
                    self.dataformats[value] = dataformat
264 265


Philip ABBET's avatar
Philip ABBET committed
266 267 268 269 270
    @property
    def name(self):
        """Returns the name of this object
        """
        return self._name or '__unnamed_database__'
271 272


Philip ABBET's avatar
Philip ABBET committed
273 274 275 276
    @property
    def schema_version(self):
        """Returns the schema version"""
        return self.data.get('schema_version', 1)
277 278


Philip ABBET's avatar
Philip ABBET committed
279 280 281
    @property
    def valid(self):
        return not bool(self.errors)
Philip ABBET's avatar
Philip ABBET committed
282 283


Philip ABBET's avatar
Philip ABBET committed
284 285 286
    @property
    def protocols(self):
        """The declaration of all the protocols of the database"""
287

Philip ABBET's avatar
Philip ABBET committed
288 289
        data = self.data['protocols']
        return dict(zip([k['name'] for k in data], data))
290 291


Philip ABBET's avatar
Philip ABBET committed
292 293
    def protocol(self, name):
        """The declaration of a specific protocol in the database"""
294

Philip ABBET's avatar
Philip ABBET committed
295
        return self.protocols[name]
296 297


Philip ABBET's avatar
Philip ABBET committed
298 299 300
    @property
    def protocol_names(self):
        """Names of protocols declared for this database"""
301

Philip ABBET's avatar
Philip ABBET committed
302 303
        data = self.data['protocols']
        return [k['name'] for k in data]
304 305


Philip ABBET's avatar
Philip ABBET committed
306 307
    def sets(self, protocol):
        """The declaration of a specific set in the database protocol"""
308

Philip ABBET's avatar
Philip ABBET committed
309 310
        data = self.protocol(protocol)['sets']
        return dict(zip([k['name'] for k in data], data))
311 312


Philip ABBET's avatar
Philip ABBET committed
313 314
    def set(self, protocol, name):
        """The declaration of all the protocols of the database"""
315

Philip ABBET's avatar
Philip ABBET committed
316
        return self.sets(protocol)[name]
317 318


Philip ABBET's avatar
Philip ABBET committed
319 320
    def set_names(self, protocol):
        """The names of sets in a given protocol for this database"""
321

Philip ABBET's avatar
Philip ABBET committed
322 323
        data = self.protocol(protocol)['sets']
        return [k['name'] for k in data]
324 325


Philip ABBET's avatar
Philip ABBET committed
326 327
    def view(self, protocol, name, exc=None):
        """Returns the database view, given the protocol and the set name
328

Philip ABBET's avatar
Philip ABBET committed
329
        Parameters:
330

Philip ABBET's avatar
Philip ABBET committed
331
          protocol (str): The name of the protocol where to retrieve the view from
332

Philip ABBET's avatar
Philip ABBET committed
333 334
          name (str): The name of the set in the protocol where to retrieve the
            view from
335

Philip ABBET's avatar
Philip ABBET committed
336 337
          exc (class): If passed, must be a valid exception class that will be
            used to report errors in the read-out of this database's view.
338

Philip ABBET's avatar
Philip ABBET committed
339
        Returns:
340

Philip ABBET's avatar
Philip ABBET committed
341 342
          The database view, which will be constructed, but not setup. You
          **must** set it up before using methods ``done`` or ``next``.
343

Philip ABBET's avatar
Philip ABBET committed
344 345 346 347 348 349 350 351 352
        """

        if not self._name:
            exc = exc or RuntimeError
            raise exc("database has no name")

        if not self.valid:
            message = "cannot load view for set `%s' of protocol `%s' " \
                    "from invalid database (%s)" % (protocol, name, self.name)
353 354 355
            if exc:
                raise exc(message)

Philip ABBET's avatar
Philip ABBET committed
356 357 358 359 360 361 362 363 364 365 366 367 368
            raise RuntimeError(message)

        # loads the module only once through the lifetime of the database object
        try:
            if not hasattr(self, '_module'):
                self._module = loader.load_module(self.name.replace(os.sep, '_'),
                          self.storage.code.path, {})
        except Exception as e:
            if exc is not None:
                type, value, traceback = sys.exc_info()
                six.reraise(exc, exc(value), traceback)
            else:
                raise #just re-raise the user exception
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
        return Runner(self._module, self.set(protocol, name), self.prefix,
                      self.data['root_folder'], exc)


#----------------------------------------------------------


class View(object):

    def index(self, root_folder, parameters):
        """Returns a list of (named) tuples describing the data provided by the view.

        The ordering of values inside the tuples is free, but it is expected
        that the list is ordered in a consistent manner (ie. all train images of
        person A, then all train images of person B, ...).

        For instance, assuming a view providing that kind of data:

        ----------- ----------- ----------- ----------- ----------- -----------
        |  image  | |  image  | |  image  | |  image  | |  image  | |  image  |
        ----------- ----------- ----------- ----------- ----------- -----------
        ----------- ----------- ----------- ----------- ----------- -----------
        | file_id | | file_id | | file_id | | file_id | | file_id | | file_id |
        ----------- ----------- ----------- ----------- ----------- -----------
        ----------------------------------- -----------------------------------
        |             client_id           | |             client_id           |
        ----------------------------------- -----------------------------------

        a list like the following should be generated:

        [
            (client_id=1, file_id=1, image=filename1),
            (client_id=1, file_id=2, image=filename2),
            (client_id=1, file_id=3, image=filename3),
            (client_id=2, file_id=4, image=filename4),
            (client_id=2, file_id=5, image=filename5),
            (client_id=2, file_id=6, image=filename6),
            ...
        ]

        DO NOT store images, sound files or data loadable from a file in the list!
        Store the path of the file to load instead.
        """

        raise NotImplementedError


    def setup(self, root_folder, parameters, objs, start_index=None, end_index=None):

        # Initialisations
        self.root_folder = root_folder
        self.parameters = parameters
        self.objs = objs

        # Determine the range of indices that must be provided
        self.start_index = start_index if start_index is not None else 0
        self.end_index = end_index if end_index is not None else len(self.objs) - 1

        self.objs = self.objs[self.start_index : self.end_index + 1]


    def get(self, output, index):
        """Returns the data of the provided output at the provided index in the list
        of (named) tuples describing the data provided by the view (accessible at
        self.objs)"""

        raise NotImplementedError
437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711


#----------------------------------------------------------


class DatabaseTester:
    """Used while developing a new database view, to test its behavior

    This class tests that, for each combination of connected/not connected
    outputs:

      - Data indices seems consistent
      - All the connected outputs produce data
      - All the not connected outputs don't produce data

    It also report some stats, and can generate a text file detailing the
    data generated by each output.

    By default, outputs are assumed to produce data at constant intervals.
    Those that don't follow this pattern, must be declared as 'irregular'.

    Note that no particular check is done about the database declaration or
    the correctness of the generated data with their data formats. This class
    is mainly used to check that the outputs are correctly synchronized.
    """

    # Mock output class
    class MockOutput:

        def __init__(self, name, connected):
            self.name = name
            self.connected = connected
            self.last_written_data_index = -1
            self.written_data = []

        def write(self, data, end_data_index):
            self.written_data.append(( self.last_written_data_index + 1, end_data_index, data ))
            self.last_written_data_index = end_data_index

        def isConnected(self):
            return self.connected


    class SynchronizedUnit:

        def __init__(self, start, end):
            self.start = start
            self.end = end
            self.data = {}
            self.children = []

        def addData(self, output, start, end, data):
            if (start == self.start) and (end == self.end):
                self.data[output] = self._dataToString(data)
            elif (len(self.children) == 0) or (self.children[-1].end < start):
                unit = DatabaseTester.SynchronizedUnit(start, end)
                unit.addData(output, start, end, data)
                self.children.append(unit)
            else:
                for index, unit in enumerate(self.children):
                    if (unit.start <= start) and (unit.end >= end):
                        unit.addData(output, start, end, data)
                        break
                    elif (unit.start == start) and (unit.end < end):
                        new_unit = DatabaseTester.SynchronizedUnit(start, end)
                        new_unit.addData(output, start, end, data)
                        new_unit.children.append(unit)

                        for i in range(index + 1, len(self.children)):
                            unit = self.children[i]
                            if (unit.end <= end):
                                new_unit.children.append(unit)
                            else:
                                break

                        self.children = self.children[:index] + [new_unit] + self.children[i:]
                        break

        def toString(self):
            texts = {}

            for child in self.children:
                child_texts = child.toString()
                for output, text in child_texts.items():
                    if texts.has_key(output):
                        texts[output] += ' ' + text
                    else:
                        texts[output] = text

            if len(self.data) > 0:
                length = max([ len(x) + 6 for x in self.data.values() ])

                if len(texts) > 0:
                    children_length = len(texts.values()[0])
                    if children_length >= length:
                        length = children_length
                    else:
                        diff = length - children_length
                        if diff % 2 == 0:
                            diff1 = diff / 2
                            diff2 = diff1
                        else:
                            diff1 = diff // 2
                            diff2 = diff - diff1

                        for k, v in texts.items():
                            texts[k] = '|%s%s%s|' % ('-' * diff1, v[1:-1], '-' * diff2)

                for output, value in self.data.items():
                    output_length = len(value) + 6
                    diff = length - output_length
                    if diff % 2 == 0:
                        diff1 = diff / 2
                        diff2 = diff1
                    else:
                        diff1 = diff // 2
                        diff2 = diff - diff1
                    texts[output] = '|-%s %s %s-|' % ('-' * diff1, value, '-' * diff2)

            length = max(len(x) for x in texts.values())
            for k, v in texts.items():
                if len(v) < length:
                    texts[k] += ' ' * (length - len(v))

            return texts

        def _dataToString(self, data):
            if (len(data) > 1) or (len(data) == 0):
                return 'X'

            value = data[data.keys()[0]]

            if isinstance(value, np.ndarray) or isinstance(value, dict):
                return 'X'

            return str(value)


    def __init__(self, name, view_class, outputs_declaration, parameters,
                 irregular_outputs=[], all_combinations=True):
        self.name = name
        self.view_class = view_class
        self.outputs_declaration = {}
        self.parameters = parameters
        self.irregular_outputs = irregular_outputs

        self.determine_regular_intervals(outputs_declaration)

        if all_combinations:
            for L in range(0, len(self.outputs_declaration) + 1):
                for subset in itertools.combinations(self.outputs_declaration.keys(), L):
                    self.run(subset)
        else:
            self.run(self.outputs_declaration.keys())

        print

    def determine_regular_intervals(self, outputs_declaration):
        outputs = OutputList()
        for name in outputs_declaration:
            outputs.add(DatabaseTester.MockOutput(name, True))

        view = self.view_class()
        view.setup('', outputs, self.parameters)

        view.next()

        for output in outputs:
            if output.name not in self.irregular_outputs:
                self.outputs_declaration[output.name] = output.last_written_data_index + 1
            else:
                self.outputs_declaration[output.name] = None


    def run(self, connected_outputs):
        if len(connected_outputs) == 0:
            return

        print "Testing '%s', with %d output(s): %s" % (self.name, len(connected_outputs),
                                                       ', '.join(connected_outputs))

        # Create the mock outputs
        connected_outputs = dict([ x for x in self.outputs_declaration.items()
                                     if x[0] in connected_outputs ])

        not_connected_outputs = dict([ x for x in self.outputs_declaration.items()
                                         if x[0] not in connected_outputs ])

        outputs = OutputList()
        for name in self.outputs_declaration.keys():
            outputs.add(DatabaseTester.MockOutput(name, name in connected_outputs))


        # Create the view
        view = self.view_class()
        view.setup('', outputs, self.parameters)


        # Initialisations
        next_expected_indices = {}
        for name, interval in connected_outputs.items():
            next_expected_indices[name] = 0

        next_index = 0

        def _done():
            for output in outputs:
                if output.isConnected() and not view.done(output.last_written_data_index):
                    return False
            return True


        # Ask for all the data
        while not(_done()):
            view.next()

            # Check the indices for the connected outputs
            for name in connected_outputs.keys():
                if name not in self.irregular_outputs:
                    assert(outputs[name].written_data[-1][0] == next_expected_indices[name])
                    assert(outputs[name].written_data[-1][1] == next_expected_indices[name] + connected_outputs[name] - 1)
                else:
                    assert(outputs[name].written_data[-1][0] == next_expected_indices[name])
                    assert(outputs[name].written_data[-1][1] >= next_expected_indices[name])

            # Check that the not connected outputs didn't produce data
            for name in not_connected_outputs.keys():
                assert(len(outputs[name].written_data) == 0)

            # Compute the next data index that should be produced
            next_index = 1 + min([ x.written_data[-1][1] for x in outputs if x.isConnected() ])

            # Compute the next data index that should be produced by each connected output
            for name in connected_outputs.keys():
                if name not in self.irregular_outputs:
                    if next_index == next_expected_indices[name] + connected_outputs[name]:
                        next_expected_indices[name] += connected_outputs[name]
                else:
                    if next_index > outputs[name].written_data[-1][1]:
                        next_expected_indices[name] = outputs[name].written_data[-1][1] + 1

        # Check the number of data produced on the regular outputs
        for name in connected_outputs.keys():
            print '  - %s: %d data' % (name, len(outputs[name].written_data))
            if name not in self.irregular_outputs:
                assert(len(outputs[name].written_data) == next_index / connected_outputs[name])

        # Check that all outputs ends on the same index
        for name in connected_outputs.keys():
            assert(outputs[name].written_data[-1][1] == next_index - 1)


        # Generate a text file with lots of details (only if all outputs are connected)
        if len(connected_outputs) == len(self.outputs_declaration):
            sorted_outputs = sorted(outputs, key=lambda x: len(x.written_data))

            unit = DatabaseTester.SynchronizedUnit(0, sorted_outputs[0].written_data[-1][1])

            for output in sorted_outputs:
                for data in output.written_data:
                    unit.addData(output.name, data[0], data[1], data[2])

            texts = unit.toString()

            outputs_max_length = max([ len(x) for x in self.outputs_declaration.keys() ])

            with open(self.name.replace(' ', '_') + '.txt', 'w') as f:
                for i in range(1, len(sorted_outputs) + 1):
                    output_name = sorted_outputs[-i].name
                    f.write(output_name + ': ')

                    if len(output_name) < outputs_max_length:
                        f.write(' ' * (outputs_max_length - len(output_name)))

                    f.write(texts[output_name] + '\n')