-
Notifications
You must be signed in to change notification settings - Fork 66
/
ChangeLog
535 lines (323 loc) · 20.9 KB
/
ChangeLog
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
2011-12-19 Gaia 2.4.3: The "Winter is coming" release (Nicolas Wack <nicolas.wack@upf.edu>)
* feature: classification/evaluation scripts have been enhanced a bit
and should now work in a much easier way.
2011-09-28 Gaia 2.4.2: The "Points can be frozen too!" release (Nicolas Wack <nicolas.wack@upf.edu>)
* feature: added the named constructor FrozenPoint.fromPoint(gaia2.Point) to
be able to correctly build a FrozenPoint from a standard Point while conserving
the layout information
2011-09-14 Gaia 2.4.1: The "Your package has arrived" release (Nicolas Wack <nicolas.wack@upf.edu>)
* bugfix: fixed Debian and CentOS packages.
* bugfix: some small compilation fixes and adjustments
2011-08-16 Gaia 2.4: The "Hell just froze over" release (Nicolas Wack <nicolas.wack@upf.edu>)
* feature: introduction of the new FrozenDataSet and its associated FrozenPoint
and FrozenViews. The FrozenDataSet is an immutable structure, but its memory is
all contiguous which makes NN-queries up to 6x faster in certain cases (more
likely to be 2x faster on average).
* feature: "classic" descriptors in python are now returned as their native
equivalent types instead of as wrappers to RealDescriptor and StringDescriptor.
* bugfix: the usual assortment of bugfixes, unicode related, small memleaks, etc...
----------------------------------------------------------------------------------
2011-04-13 Gaia 2.3.5: The "SearchSpace forever" release (Nicolas Wack <nicolas.wack@upf.edu>)
* bugfix: fixed 2 very hard to hit crashes when reusing a searchspace in the
query optimizer.
2011-04-05 Gaia 2.3.4: The "Appending Enums is Shady" release (Nicolas Wack <nicolas.wack@upf.edu>)
* bugfix: fixed a grave data corruption bug when you tried to append 2 datasets
which have some enumerated values.
2011-03-15 Gaia 2.3.3: The "Need for Speed" release (Nicolas Wack <nicolas.wack@upf.edu>)
* feature: revamped the logging subsystem to have colors and submodules identifiers.
* feature: new gaia2.essentia python submodule that allows to run an Essentia
extractor on all files in a directory.
* optim: fixed a bug where getting an indexed SearchSpace would take way longer
than expected.
* optim: SearchSpace::filterAndSort is now a tiny bit faster.
* bugfix: fixed a memory leak in SvmPredict
2011-01-26 Gaia 2.3.2.5: The "Cyclops loves classifiers!" release (Nicolas Wack <nicolas.wack@upf.edu>)
* feature: added optimized parameters for training the SVM models for Essentia.
* feature: a bit of work on the classification scripts.
* bugfix: Cronolog now runs as user cyclops instead of root.
* bugfix: fixed a crash in the ResultSet::thresholdLimit() method.
2010-12-10 Gaia 2.3.2.4: The "Cyclops eats strings for breakfast" release (Nicolas Wack <nicolas.wack@upf.edu>)
* bugfix: Cyclops client does not automatically convert IDs from strings to integers anymore.
2010-09-27 Gaia 2.3.2.3: The "Cyclops likes CentOS" release (Nicolas Wack <nicolas.wack@upf.edu>)
* bugfix: fixed CentOS package.
2010-09-27 Gaia 2.3.2.2: The "Cyclops back from holidays" release (Nicolas Wack <nicolas.wack@upf.edu>)
* bugfix: fixed a cyclops segfault on queries that use the "thresholdLimit" param.
* bugfix: fixed a crash on RHEL 5.5 where cyclopsmaster would segfault on "status" query.
2010-08-17 Gaia 2.3.2.1: The "Cyclops in holidays" release (Nicolas Wack <nicolas.wack@upf.edu>)
* feature: added thresholdLimit to filter out results by distance instead
of result numbers.
* feature: added Rename algorithm to rename a single descriptor.
* bugfix: workaround for a bug in libyaml that can't read map keys with
more than 1024 chars.
2010-07-08 Gaia 2.3.2: The "Cyclops likes MIREX" release (Nicolas Wack <nicolas.wack@upf.edu>)
* feature: Cyclops, a multithreaded C++ Yaml-RPC query server makes its entry
in Gaia world. It is intended to be run as a daemon, and also comes in a
distributed flavor, with a CyclopsMaster and slaves which it can manage.
* feature: the Mirex scripts from last year have seen a huge makeover, and are
now part of Gaia's standard python library. They are also used to
(semi-)automatically generate the SVM models to be used for Essentia.
* feature: the Gaussianize transformation is now wayyyyy faster to serialize/
deserialize, and the resulting histories are more than 2x smaller in size.
WARNING: this is a binary-incompatible change, and histories created with either
Gaia 2.3 or 2.3.1 will generate incorrect results.
* feature: the Select and Remove transformations are much faster to instantiate.
* feature: the gaiainfo tool is now also much faster (constant time).
* feature: filtering and indexing now work on enumerated descriptors.
* feature: there is now an installer for Gaia2 on Windows, as well as CentOS 5
rpm packages!
* bugfix: gaiamerge's "--reflayout" option was broken since 2.3, it is now fixed.
* bugfix: Points which have 2 descriptors with a common part in their namespace
can now be loaded correctly.
2010-03-11 Gaia 2.3.1: The "GaiaFusion is hot!" release (Nicolas Wack <nicolas.wack@upf.edu>)
* feature: GaiaFusion now works much better, and is also resilient to
discrepancies that might happen in histories when merging multiple chunks.
* feature: can now create programmable filters, ie: filters that can be
constructed directly instead of being parsed from a filter string.
* feature: SearchSpaces can now be built manually.
* feature: new distance: ResistorAverage.
* feature: started importing some python scripts into gaia2 python module.
* bugfix: transforming a dataset now uses the absolute strict minimum amount
of required memory (there was lots of temp memory thrashing before).
* bugfix: the encoding for strings is now forced to utf-8 all the time, even
if your locale is different.
* bugfix: the usual bugfixes, this time mostly about fixed-/variable-length
ambiguities, some SVM fixes, etc...
2009-11-27 Gaia 2.3: The "No-collection" release (Nicolas Wack <nicolas.wack@upf.edu>)
* feature: the collections have been removed from the DataSet class.
Now a dataset only contains points.
* feature: a new tool, GaiaFusion, is introduced. It is a pumped up
version of gaiamerge, which can merge automatically by chunks, and it also
applies the basic required transfo to a dataset (removevl, fixlength, cleaner)
* API: fixlength needs to be applied before cleaner, and actually, it is
a good idea to apply it always as first step.
* API change: DataSet::totalPoints() -> DataSet::size()
* API change: TransfoChain::originalLayout() is now DataSet::originalLayout()
* API change: More generally, the naming of the parameters is more consistent:
- Most of the algorithms take "descriptorNames" and "except" as parameters,
except when it doesn't make sense.
- all composite metrics now take the "distance" and "params" parameters,
which are used for creating the composed distance. (ie: no more forwarding
of half ParameterMaps to composed distance)
- Parameters given to algorithms and metrics are now checked against the ones
that they support, and it will throw an exception if you give them unused
parameters.
* bugfix: fixes a (potentially) huge memory leak when doing queries, as well as
2 other ones in the python bindings when instanting Metrics or Analyzers through
their respective factories
* feature: new distances: ForceIdentity.
----------------------------------------------------------------------------------
2009-09-16 Gaia 2.2.3: The "Mirex 2009" release (Nicolas Wack <nicolas.wack@upf.edu>)
* feature: start of some more high-level tools for doing classification.
This is located in test/classification and is still very experimental.
Use at your own risk. Ultimately, most of this stuff will move to the
standard gaia2 python namespace.
This also contains an implementation of the SIMCA classifier.
* feature: the PCA and RCA transformations now have a parameter to specify
the name of the resulting descriptor.
* bugfix: the usual series of bugfixes, small warts and annoyances,
including the following ones reported in Trac: #139, #142
Also note that Point::load() now correctly loads segmented sigfiles from
Essentia >= 1.1 .
2009-07-06 Gaia 2.2.2.1: The "Third-of-an-imp" release (Nicolas Wack <nicolas.wack@upf.edu>)
* feature: gaia2.utils.mergeDataSet now accepts start and end parameters.
* bugfix: appendCollection works again (failed since the DataSets started
being self-referencing by default).
* bugfix: the order of construction of layouts is not important anymore.
2009-06-18 Gaia 2.2.2: The "Third-of-a-devil" release (Nicolas Wack <nicolas.wack@upf.edu>)
* feature: Gaia now correctly deals with multi-segment points (with respect to
transformations and metrics).
* feature: DataSets are now self-referencing by default. This means that the call
to DataSet::setReferenceDataSet(this) is now superfluous (although still correct).
* feature: new distances: RandomDistance, SemanticClassifier, SummedSegment,
WeightedEuclidean
* feature: Python bindings now include docstrings for the algorithms/metrics.
* feature: new tool: gaiasvm_to_essentia, which converts a Gaia history containing
an SVM trainer into an Essentia SVM model file ready to be used. This works even
when the history contains transfos such as select, remove, normalize, etc...
* feature: AddField transformation now allows to specify default values.
* feature: ResultSets can now limit() the number of points they contain.
* bugfix: Gaia doesn't crash anymore when making a View on a DataSet and a
Distance which was created on a different layout.
* bugfix: fixed a crash when KullbackLeibler distance was used in a multithreaded
query (ie: when TBB was installed when compiling).
* bugfix: some gaiamerge fixes (possible race condition), added the possibility to
specify a reference layout and also the number of threads to be used for merging.
2009-03-12 Gaia 2.2.1: The Wolowicz release (Nicolas Wack <nicolas.wack@upf.edu>)
* feature: Gaia can now be compiled with TBB to have multi-threaded queries.
* bugfix: Gaia now compiles on CentOS4, and fixes the debian packages.
2009-03-06 Gaia 2.2: The Sheldon release (Nicolas Wack <nicolas.wack@upf.edu>)
* feature: WARNING: API-incompatible change
now View.nnSearch doesn't return the results anymore, but instead a ResultSet
which contains all the results (similar to what is returned by a SQL query).
You can call ResultSet::get(nPoints, offset) to get the results you would have
gotten directly before.
ResultSets can be reused for another query, allowing for chained queries.
ResultSets are explicitly shared (ref-counted), meaning that if you make a copy
of one, you will have 2 of them but pointing to the same data. When the last
instance of a ResultSet is destroyed, the memory is returned to the system.
* feature: you can now index views, which makes it much faster to use queries
with filters on them. The flip side of the coin is that you cannot anymore
filter using collections (ie: WHERE ... FROM collec1, collec2)
* feature: you can use a new distance for each query you do on the view, you
don't need to create a new view each time
* feature: in python it is now possible to access descriptor directly with
p[name] and p[name] = value (instead of p.value(name) or p.label(name) )
* feature: new distances: RhythmDistance, WeightedPearsonCorrelation and
BPFCompose
* feature: new transformations: SVM classifier, Extract
* feature: new build target: 'waf dist' to create a redistributable tarball
* bugfix: it is not possible to change a layout anymore once it has been shared
between points.
* bugfix: check for null pointers when applying transformations on datasets.
* bugfix: you can now add descriptors to a layout using their fully qualified
name instead of adding each part consecutively. (ticket #106)
----------------------------------------------------------------------------------
2008-12-10 Gaia 2.1.2: The Zapp Brannigan release (Nicolas Wack <nicolas.wack@upf.edu>)
* fixed a very grave bug in Collection::removePoints() which could lead
to data corruption.
2008-11-25 Gaia 2.1.1: The Bender release (Nicolas Wack <nwack@iua.upf.edu>)
* feature/bugfix: The unicity condition is now enforced upon the IDs of
points within the same collection, both when adding new points and when
loading a DataSet.
* feature: Collection::addPoint* and Collection::removePoint* now have
a worst case complexity of O(n log n) instead of O(n^2) before.
* bugfix: Fixed a (possibly huge) memory leak in the query parser
2008-10-22 Gaia 2.1: The Nibbler release (Nicolas Wack <nwack@iua.upf.edu>)
* Note: now Point::value() returns a copy of the data, not a reference anymore
* feature: Gaia is now ready to load multiple-segments files (segmented by
Essentia) as well as work on it. This feature is not yet fully working but
the file format is fixed so that you can have a smooth upgrade path from the
2.0.x versions. Gaia 2.1 can load any 2.0.x file and load them correctly.
* feature: Added new transformation FixLength, which fixes the length of the
given descriptors, which leads to (sometimes huge) memory savings.
* feature: Added new transformation Enumerate, which allows you to convert
a string descriptor to an enum one. This is "invisible" to the user but
leads to quite big improvements in memory usage as well as better filtering
performance, even possibly indexing.
* feature: The Gaussianize transformation is resurrected from 1.4, Normalize
now takes much more parameters, such as % of outliers, whether to apply
clipping, ...
* feature: Added a new tool "gaiainfo" that returns information about a
given dataset file.
* feature: mergeDataSets now preserves the history of the merged DataSets,
so that it is now possible to add a point to the resulting dataset
* feature: New distance functions: NormalizedAngle (replaces CosineNorm from
the 2.0.x branch) and ExponentialCompress (replaces and extends EuclideanExp
from the 2.0.x branch)
* python: you can now iterate over the points and collections of a DataSet
using ds.points() and ds.collections()
* bugfix: Fixed segfault when deleting a dataset before dereferencing it (by
other datasets)
* bugfix: RCA is functional again! There is also a new parameter, classLabel,
which you can use instead of the classFile you had before if the descriptor
for the class is already in the dataset.
----------------------------------------------------------------------------------
2008-05-16 Gaia 2.0.8: The Super-Arale-chan release (Nicolas Wack <nwack@iua.upf.edu>)
* bugfix: fixed a potential segfault when deleting a dataset before a view
that references it.
* feature: By default asserts are now removed from builds. Add them back
with: './waf configure --with-asserts'
2008-04-25 Gaia 2.0.7: The Arale release (Nicolas Wack <nwack@iua.upf.edu>)
* bugfix: Gaia now loads signature files encoded in utf-8. (ticket #52)
* bugfix: When setting a reference dataset, it now also checks that all the
point names are equal, not only the number of points in each collection.
(ticket #59)
* bugfix: Fixed a nasty bug where some point names mysteriously disappeared
just after merging. (ticket #49, #51 and #60)
2008-03-04 Gaia 2.0.6: The Senbei release (Nicolas Wack <nwack@iua.upf.edu>)
* feature: Added a new transformation MergeRegion which takes several
descriptors and merges them into a single one. This has the nice effect to
require much less memory afterwards, and allows us again to work on datasets
of ~1M points with ~1GB of RAM.
* feature: Added a waf configuration option that allows to compile Gaia to
load "bz2-encrypted" signature files.
* bugfix: Collection::removePointNoSort() has been removed (it could not
possibly work), Collection::removePoints() has been added and does the job
of removing multiple points at once much better. (ticket #36)
* feature: Points can now be loaded directly from a string buffer using the
new Point::loadFromString() method. (ticket #41)
2007-10-10 Gaia 2.0.5: The Midori release (Nicolas Wack <nwack@iua.upf.edu>)
* feature: Added a new function mergePoints(p1, p2) which merges the data
contained in the two points
* feature: There is a version number field available:
const char* gaia2::version in C++ and gaia2.__version__ in python,
thus closing #39
* bugfix: Closed tickets #37 and #38
2007-10-04 Gaia 2.0.4: The Gatchan release (Nicolas Wack <nwack@iua.upf.edu>)
* Fixed the bug (ticket #32) where de-referencing a dataset would make
gaia crash.
* Fixed a potential bug (ticket #29) where morphing a point because of
a wrongly detected layout would not work as expected.
* Algorithms that do not make sense on datasets with less than 2 points
now throw an explicit exception (ticket #31)
2007-09-07 Gaia 2.0.3: The Great King Nikochan release (Nicolas Wack <nwack@iua.upf.edu>)
(Also called the holiday release...)
* Removed the unicity condition for the points (which caused the *NoSort
functions to sort the points anyway).
* Changed the base vector type from QVector to QVarLengthArray, resulting
in memory savings around 35%!! (YMMV)
* Fixed a refcounting bug in SWIG/Python bindings on the reference dataset.
2007-08-16 Gaia 2.0.2: The Ku-Pee-Pee Poh? release (Nicolas Wack <nwack@iua.upf.edu>)
Fixed the (not so) obvious bugs that appeared with 2.0.1 and that needed
immediate fixing:
* Using filters during a query would previously not use the values from
the reference dataset, but those from the current one. Now fixed!
* The FROM list in the query filters was ignored, but not anymore.
* When using containing conditions on point IDs, we do not require
anymore the current dataset to have a reference one.
* Various API changes: all addPoint* and addCollection methods now
return a reference to the object that was just inserted (ticket #22)
You can now also access points in a collection using their index, which
is not recommended but can prove to be useful when iterating over all
points in a collection, for instance.
* Actually changed the namespace from gaia to gaia2 (the previous
Changelog said it was done, but that was a blatant lie!!)
* Syntax of the filter parser additions: BETWEEN ... AND ...
2007-08-10 Gaia 2.0.1: The N'Cha! release (Nicolas Wack <nwack@iua.upf.edu>)
First maintenance release of the 2.0 branch:
* Doxygen documentation improved a lot, now also contains examples &
tutorials
* Build system has been improved a lot, new logging system which allows
to activate or deactivate various runtime debugging levels.
* New algorithms: "RemoveVL" which removes the variable-length
descriptors from a dataset.
* namespace for the base library is now "gaia2", and namespace for the
STL facade is now "gaiastd"; gaia2::GaiaException can also be caught as
gaiastd::GException (instead of the more generic std::exception)
* faster point lookup using binary search (O(log(n)) instead of O(n))
* Views now allow pre-filtering before doing queries.
* Syntax of the filter parser additions: FROM, NOT, IN (...), point.id
The addition of the FROM clause also implies that from now on, all
predicated should be written starting with the WHERE keyword.
* python bindings now handle more gracefully Parameter and ParameterMap
objects, can convert from QString to python string, etc...
* and of course the usual slew of bugfixes, of which the most important
were:
#21: auto-detection of the type when parsing the yaml file sometimes
detected a value which was supposed to be a string as a float,
resulting in the loaded point not being able to be added to a
dataset
#20: query parser was in an invlaid state after parsing an invalid
query resulting in following queries to also fail
#18: the representation of StringDescriptor in python was uncertain.
now it maps exactly the one from python native string and list
of strings
2007-07-11 Gaia 2.0: The Hoyoyo! release (Nicolas Wack <nwack@iua.upf.edu>)
Rewritten everything from scratch, new significant features include:
* More flexible Point layout (ie: possible descriptors), hierarchically
ordered, can use string types as well as real number types, and can also
support the concept of segment (ie: a chunk of the whole audio + scope)
(NB: segments don't work yet in 2.0, but will do so in the next version)
* A DataSet is now composed of Collections, making it easier to work
with multiple collections of tracks (ie: different users of the same
system) within the same DataSet (much easier and faster than having
multiple datasets)
* Better traceability of descriptors. It is now very easy to access
values from a point, know which descriptors were taken into account in
a PCA, etc...
* A simpler and much more generic way of defining distance functions
between points in a space that makes it possible to experiment which one
of those works better for you, to mix different ones to create a new one
that makes musical sense, etc...
* And an overall better architecture that makes it easier to work with
and less prone to bugs...