comparison cpack/dml/api/perspectives.pl @ 0:718306e29690 tip

commiting public release
author Daniel Wolff
date Tue, 09 Feb 2016 21:05:06 +0100
parents
children
comparison
equal deleted inserted replaced
-1:000000000000 0:718306e29690
1 /* Part of DML (Digital Music Laboratory)
2 Copyright 2014-2015 Samer Abdallah, University of London
3
4 This program is free software; you can redistribute it and/or
5 modify it under the terms of the GNU General Public License
6 as published by the Free Software Foundation; either version 2
7 of the License, or (at your option) any later version.
8
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
13
14 You should have received a copy of the GNU General Public
15 License along with this library; if not, write to the Free Software
16 Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18
19 :- module(perspectives, []).
20
21 /** <module> VIS API Perspectives
22
23 Todo
24
25 - Chord sequences
26 - Standardise data structures
27 */
28 :- use_module(library(http/http_dispatch), [http_link_to_id/3]).
29 :- use_module(library(semweb/rdf_db)).
30 :- use_module(library(semweb/rdf_label)).
31 :- use_module(library(dcg_core)).
32 :- use_module(library(insist)).
33 :- use_module(library(computations)).
34 :- use_module(library(backend_json)).
35 :- use_module(library(dataset)).
36 :- use_module(library(memo)).
37 :- use_module(library(async)).
38 :- use_module(library(mlserver)).
39 :- use_module(api(dmlvis)).
40 :- use_module(api(archive)).
41
42 % :- setting(memoise_failures,boolean,false,"Whether or not to record failed computations to avoid retrying").
43 :- setting(default_recompute_policy,oneof([none,failed,force]),none,'Default policy on recomputing memoised computations').
44 :- setting(default_vamp_on_demand,boolean,false,'Default policy on doing VAMP computations on demand').
45
46 % registry of perspectives.
47 dmlvis:perspective( getRecordingPerspective, Name,
48 [+uri(URI),vamp_on_demand(V)-false|Params],
49 cc(perspectives:rla(Pred,[vamp_on_demand(V)],URI))
50 ) :- rec_persp(Name, Params, Pred).
51
52 dmlvis:perspective( getCollectionPerspective, Name,
53 [+cid(CID),recompute(R)-none,vamp_on_demand(V)-false,coverage(C)-summary|Params],
54 cc(perspectives:cla(Pred,[recompute(R),vamp_on_demand(V),coverage(C)],CID))
55 ) :- coll_persp(Name, Params, Pred).
56
57 :- meta_predicate rla(2,+,+,-,-), cla(2,+,+,-,-).
58
59 rla(Pred,Opts,URI,Result,stable) :-
60 option(vamp_on_demand(V), Opts, false),
61 with_global(vamp_on_demand, V, call(Pred,URI,Result)).
62
63 cla(Pred,Opts,CID,Result,stable) :-
64 check_collection(CID),
65 call(Pred,Opts,CID,Result1),
66 option(coverage(Cov),Opts,full),
67 insist(filter_coverage(Cov,Result1,Result), invalid_coverage_parameter(Cov)).
68
69 :- op(1050,xfy,=>).
70 G1 => G2 --> (call_dcg(G1) -> call_dcg(G2); []).
71
72 filter_coverage(full) --> [].
73 filter_coverage(summary) --> dtrans(coverage,C1,C2) => {summarise_coverage(C1,C2)}.
74 summarise_coverage --> foldl(replace_list_with_length,[failed,errors],[failed_count,errors_count]).
75 replace_list_with_length(Key) --> dtrans(Key,List,Length) => {length(List,Length)}.
76 replace_list_with_length(Key1,Key2) --> ddel(Key1,List) => {length(List,Len)}, dput(Key2,Len).
77 dtrans(Key,Val1,Val2,D1,D2) :- get_dict(Key,D1,Val1,D2,Val2).
78 ddel(Key,Val,D1,D2) :- del_dict(Key,D1,Val,D2).
79 dput(Key,Val,D1,D2) :- put_dict(Key,D1,Val,D2).
80 % dget(Key,Val,D,D) :- get_dict(Key,D,Val).
81
82 check_collection(CID) :-
83 insist(dataset_size(CID,Size), unknown_collection(CID)),
84 debug(dmlvis(perspective),'Doing collection level analysis on ~d items.',[Size]).
85
86 rec_persp( transcription, [], output_link(transcription(0))).
87 rec_persp( transcription_fine, [], output_link(transcription(1))).
88 rec_persp( chords, [], output_link(chords)).
89 rec_persp( chord_notes, [], output_link(chord_notes)).
90 rec_persp( beatroot, [], output_link(beats(beatroot))).
91 rec_persp( key, [], output_link(key)).
92 rec_persp( key_tonic, [], output_link(tonic)).
93 rec_persp( beats, [], output_link(beats(qm))).
94 rec_persp( tempo, [], output_link(tempo)).
95 rec_persp( chromagram, [], output_link(chromagram)).
96 rec_persp( mfcc, [], output_link(mfcc)).
97
98 rec_persp( spectrogram, [offset(O)-0,length(L)-60], spectrogram_link(O,L)).
99 rec_persp( tempo_nonuniform, [], nonuniform_tempo).
100 rec_persp( tempo_uniform, [period(DT)-1,lang(L)-ml ], uniform_tempo(L,DT)).
101 rec_persp( tempo_normalised, [num_samples(N)-20,lang(L)-ml ], normalised_tempo(L,N)).
102 rec_persp( chord_histogram, [], chord_histogram).
103 rec_persp( midi_pitch_histogram, [weighting(W)-none], pitch_histogram(W)).
104 rec_persp( pitch_histogram, [weighting(W)-none, quant(Q)-5, min(Min)-0, max(Max)-127, lang(L)-ml ],
105 freq_histogram(L,Min,Max,Q,W)).
106 rec_persp( tempo_histogram, [period(DT)-1, num_bins(N)-50, min(Min)-20, max(Max)-360, lang(L)-ml ],
107 tempo_histogram(L,DT,Min,Max,N)).
108
109 %% coll_persp(P:perspective(A), Params:list(param), Pred:pred(+options,+dataset,-A)) is nondet.
110 %
111 % Database of collection perspectives. The first argument is an atom denoting a perspective
112 % which returns results of type A. Params must be defined as in dmlvis:options_optspec/2.
113 % Pred must accept a list of options and a collection (dataset) id and produce a result.
114 coll_persp( mean_tempo_curve, [num_samples(N)-20,lang(L)-ml], mem(collection_tempo_curve(L,N))).
115 coll_persp( midi_pitch_histogram, [weighting(W)-none], mem(collection_pitch_histogram(W))).
116 coll_persp( pitch_histogram, [weighting(W)-none, quant(Q)-5, min(Min)-20, max(Max)-100, lang(L)-ml],
117 mem(collection_freq_histogram(L,Min,Max,Q,W))).
118 coll_persp( tempo_histogram, [period(DT)-1, num_bins(N)-50, min(Min)-20, max(Max)-100, lang(L)-ml],
119 mem(collection_tempo_histogram(L,DT,Min,Max,N))).
120 coll_persp( pitch_lookup, [+midi_pitch(P), weighting(W)-none, limit(Lim)-5000,offset(Off)-0],
121 nomem(collection_pitch_lookup(W,P,Lim,Off))).
122
123 % using python back-end
124 coll_persp( tonic_relative_pitch_class_histogram, [],
125 mem(py_hist(transcription_tonic_duration, tonic_norm_semitone_hist:aggregate, [opts{normalisation:piece}]))).
126 coll_persp( tonic_histogram, [], mem(py_hist(tagged(tonic), key_tonic_hist:aggregate, []))).
127 coll_persp( pitch_class_histogram, [], mem(py_hist(tagged(transcription), semitone_hist:aggregate, []))).
128 coll_persp( tuning_stats, [], mem(py_cla(tagged(transcription(1)),tuning_stats:per_file,[]))).
129 coll_persp( tuning_stats_by_year, [], mem(py_cla(transcription_date,tuning_stats_byyear:per_file,[]))).
130 coll_persp( places_hist, [], nomem(py_cla(list_places,places_hist:per_file,[]))).
131 coll_persp( key_relative_chord_seq,
132 [ spm_minlen(MinLen)-2, spm_maxseqs(MaxSeqs)-500, spm_algorithm(Alg)-'CM-SPADE',
133 spm_ignore_n(Ignn)-1, spm_maxtime(Smaxt)-60, spm_minsupport(Smins)-50 ],
134 mem(py_cla( keys_chords,chord_seq_key_relative:aggregate,
135 [opts{ spm_minlen:MinLen, spm_maxseqs:MaxSeqs, spm_algorithm:Alg,
136 spm_ignore_n:Ignn,spm_maxtime:Smaxt,spm_minsupport:Smins } ]))).
137
138 coll_persp( similarity,
139 [ sim_downsample(SimDown)-1,sim_clusters(SimClusters)-40,sim_reclimit(Limo)-2000,
140 sim_type(SimType)-'euclidean',sim_features(SimFeat)-'chromagram',
141 sim_compressor(SimComp)-'zlib'],
142 mem(py_cla(similarity_bundle,similarity:per_file,
143 [opts{sim_type:SimType,sim_clusters:SimClusters,sim_downsample:SimDown,
144 sim_reclimit:Limo,sim_features:SimFeat,sim_compressor:SimComp}]))).
145
146 % adaptor to ignore collection perspective options parameter
147 nomem(Goal,Opts,CID,Result) :-
148 option(vamp_on_demand(V), Opts, false),
149 with_global(vamp_on_demand, V,
150 with_progress_stack(call(Goal,CID,Result))).
151
152 dmlvis:param( recompute, [oneof([none,failed,force]), default(Def),
153 description('Controls handling of memoised collection level results')]) :-
154 setting(default_recompute_policy,Def).
155 dmlvis:param( vamp_on_demand, [boolean, default(Def),
156 description('Whether to run VAMP plugins if results are not already available')]) :-
157 setting(default_vamp_on_demand,Def).
158 dmlvis:param( coverage, [oneof([full,summary]), default(summary),
159 description('How much detail to provide about recordings not successfully included in CLA')]).
160 dmlvis:param( offset, [number, default(0), description('Offset into signal in seconds')]).
161 dmlvis:param( length, [number, default(60), description('Length of signal extract in seconds')]).
162 dmlvis:param( weighting, [oneof([none,dur,vel]), default(none), description('Weighting for pitch_histogram perspective')]).
163 dmlvis:param( quant, [nonneg, default(5), description('Subdivisions of a semitone for freq_histogram')]).
164 dmlvis:param( period, [number, default(1), description('Sampling period in seconds')]).
165 dmlvis:param( num_bins, [nonneg, default(50), description('Number of bins for histogram')]).
166 dmlvis:param( num_samples, [nonneg, default(50), description('Number of samples for normalised histogram')]).
167 dmlvis:param( max, [nonneg, default(100), description('Max pitch for pitch histogram')]).
168 dmlvis:param( min, [nonneg, default(20), description('Min pitch for pitch histogram')]).
169 dmlvis:param( lang, [oneof([ml,r]), default(r), description('Numerical computations language')]).
170
171 /* chord sequence parameters */
172 dmlvis:param( spm_minlen, [nonneg, default(2), description('Minimum length of chord sequence')]).
173 dmlvis:param( spm_maxseqs, [nonneg, default(500), description('Maximum number of sequences to return')]).
174 dmlvis:param( spm_algorithm, [atom, default('CM-SPADE'), description('CM-SPADE, TKS or ClaSP')]).
175 dmlvis:param( spm_ignore_n, [nonneg, default(1), description('Ignore failed chord detections')]).
176 dmlvis:param( spm_maxtime, [nonneg, default(60), description('Max. runtime for SPM algorithm')]).
177 dmlvis:param( spm_minsupport, [nonneg, default(50), description('Minimal Support in Percent')]).
178
179
180 /* similarity parameters */
181 dmlvis:param( sim_type, [atom, default('euclidean'), description('Tpye of similarity measure: euclidean, compression')]).
182 dmlvis:param( sim_clusters, [nonneg, default(40), description('Number of clusters for vector Quantisation (40-200)')]).
183 dmlvis:param( sim_downsample, [nonneg, default(1), description('Downsample the audio analysis to a resolution of 1 second')]).
184 dmlvis:param( sim_reclimit, [nonneg, default(2000), description('Maximum number of recordings in dataset')]).
185 dmlvis:param( sim_features, [atom, default('chromagram'), description('Feature basis of the similarity estimation, any combination, separated by comma: chromagram,mfcc,chords')]).
186 dmlvis:param( sim_compressor, [atom, default('zlib'), description('Compressor for similarity estimation: zlib, zxd')]).
187
188
189 :- rdf_meta transform_computation(+,r,r).
190 transform_computation(Class,In,Out) :-
191 ( transform(Class,Fn), computation(Fn,In,Out) *-> true
192 ; ( nb_current(vamp_on_demand,true)
193 -> insist(transform(Class,Fn), unrecognised_transform_class(Class)), % picks first match
194 format(string(Desc),"Running computation ~w on ~w.",[Fn,In]),
195 simple_task(Desc,computation_memo(Fn,In,Out))
196 ; throw(missing_computation(Class,In))
197 )
198 ).
199
200 :- rdf_meta transform_op(+,+,r,-).
201 transform_op(TName,Op,In,Out) :-
202 transform_computation(TName,In,X),
203 csv_op(Op,X,Out).
204
205 % ------- recording level perspectives ------------
206
207 spectrogram_link(Offs,Len,URI,_{image_url:Link}) :-
208 http_link_to_id(spectrogram_window, [uri(URI), offset(Offs), length(Len)], Link).
209
210 output_link(TransformName,Input,_{csv:Output}) :-
211 transform_computation(TransformName,Input,Output).
212
213 chord_histogram(URI,_{values:Chords, counts:Counts}) :-
214 transform_op(chords,chord_hist,URI,Hist),
215 unzip(Hist,Chords,Counts).
216
217 pitch_histogram(W,URI,_{values:NNs, counts:Counts}) :-
218 transform_op(transcription,pitch_hist(W),URI,Hist),
219 unzip(Hist,NNs,Counts).
220
221 freq_histogram(ml,Min,Max,Q,W,URI,_{edges:Edges, counts:Counts}) :-
222 microtone_map(Min,Max,Q,Map),
223 transform_op(transcription(1),freq_hist(Map,W),URI,Counts),
224 map_edges(ml,Map,Edges).
225 freq_histogram(r,Min,Max,Q,W,URI,_{edges:Edges, counts:Counts}) :-
226 microtone_map(Min,Max,Q,Map),
227 transform_op(transcription(1),freq_hist_r(Map,W),URI,Counts),
228 map_edges(r,Map,Edges).
229
230 nonuniform_tempo(URI,_{times:Times, values:Values}) :-
231 transform_op(tempo,tempo,URI,Result),
232 unzip(Result,Times,Values).
233
234 uniform_tempo(ml,DT,URI,_{times:Times, values:Values}) :-
235 transform_op(tempo,uniform_tempo(DT),URI,Result),
236 Result=Times-Values.
237
238 uniform_tempo(r,DT,URI,_{times:Times, values:Values}) :-
239 transform_op(tempo,uniform_tempo_r(DT),URI,Result),
240 Result=Times-Values.
241
242 normalised_tempo(ml,N,URI,_{times:Times, values:Values}) :-
243 transform_op(tempo,normalised_tempo(N),URI,Result),
244 Result=Times-Values.
245
246 normalised_tempo(r,N,URI,_{times:Times, values:Values}) :-
247 transform_op(tempo,normalised_tempo_r(N),URI,Result),
248 Result=Times-Values.
249
250 tempo_histogram(ml,DT,Min,Max,N,URI,_{edges:Edges, counts:Counts}) :-
251 insist(Min>0, domain_error(min,"positive value",Min)),
252 Map=expmap(Min,Max,N),
253 map_edges(ml,Map,Edges),
254 transform_op(tempo,tempo_hist(DT,Map),URI,Result),
255 Result=_-Counts.
256 tempo_histogram(r,DT,Min,Max,N,URI,_{edges:Edges, counts:Counts}) :-
257 insist(Min>0, domain_error(min,"positive value",Min)),
258 Map=expmap(Min,Max,N),
259 map_edges(r,Map,Edges),
260 transform_op(tempo,tempo_hist_r(DT,Map),URI,Result),
261 Result=_-Counts.
262
263
264 % ------- collection level perspectives ------------
265
266 collection_pitch_histogram(W,CID,Result) :-
267 Min-Max = 20-100, % !!! FIXME
268 numlist(Min,Max,NNs),
269 dataset_histogram(CID, dense_pitch_hist(Min,Max,W), _{values:NNs}, Result).
270
271 collection_pitch_lookup(Weighting, Pitch, Lim, Offset, CID, Result) :-
272 map_reduce_dataset(rec_pitch_hist(Weighting), pitch_lookup_cont(Pitch,Lim,Offset), CID, Result).
273
274 pitch_lookup_cont(Pitch,Lim,Offset,RecHists, _{items:Items}) :-
275 findall( _{ uri: Rec, label:Label, count:Count, prob:Prob },
276 offset(Offset, limit(Lim, order_by( [desc(Prob)],
277 ( member(Rec-Hist,RecHists),
278 rdf_display_label(Rec,Label),
279 pitch_hist_prob(Hist,Pitch,Count,Prob)
280 )))),
281 Items).
282
283 % collection_pitch_lookup_alt(Weighting, Pitch, Lim, Offset, CID, _{ items:Items, coverage:Coverage}) :-
284 % findall_map_coverage(dataset_item(CID), rec_transcription, RecTrans, Coverage),
285 % findall( _{ uri: Rec, label:Label, count:Count, prob:Prob },
286 % offset(Offset, limit(Lim, order_by( [desc(Prob)],
287 % ( csv_pitch_count_prob(Weighting,Trans,Pitch,Count,Prob),
288 % member(Rec-Trans,RecTrans),
289 % rdf_display_label(Rec,Label)
290 % )))),
291 % Items).
292 %
293 % rec_transcription(Rec,Rec-Transcription) :- transform_computation(transcription,Rec,Transcription).
294
295 collection_freq_histogram(Lang,Min,Max,Q,W,CID,Result) :-
296 Map=binmap(Min,Max,(Max-Min)*Q+1),
297 map_edges(Lang,Map,Edges),
298 dataset_histogram(CID, dense_freq_hist(Lang,Map,W),_{edges:Edges}, Result).
299
300 collection_tempo_histogram(Lang,DT,Min,Max,N,CID,Result) :-
301 insist(Min>0, domain_error(min,"positive value",Min)),
302 Map=expmap(Min,Max,N),
303 map_edges(Lang,Map,Edges),
304 dataset_histogram(CID, tempo_hist(Lang,DT,Map), _{edges:Edges}, Result).
305
306 collection_tempo_curve(Lang,N,CID, Result) :-
307 map_reduce_dataset(tempo_curve(Lang,N), tempo_curves_stats(Lang), CID, Result).
308
309
310 dataset_histogram(CID, Mapper, Dict, Result) :-
311 dataset_map_fold_reduce(CID,Mapper,with_dl(fold_hist),finish_hist(Dict),nothing,Result).
312
313 fold_hist([], S, S) :- !.
314 fold_hist(Xs, just(C1), just(C2)) :- !, insist(seqmap(maplist(add),Xs,C1,C2)).
315 fold_hist([X|Xs], nothing, just(C)) :- insist(seqmap(maplist(add),Xs,X,C)).
316
317 finish_hist(Dict,just(Counts),Hist) :- put_dict(counts,Dict,Counts,Hist).
318
319 py_hist(Mapper, PyFunction, Args, CID, _{counts:H,values:D,coverage:C,py_coverage:PYC}) :-
320 py_cla(Mapper,PyFunction,Args, CID, _{stats:_{counts:H,domain:D},coverage:C,py_coverage:PYC}).
321
322 py_cla(Mapper, PyFunction, Args, CID, Result) :-
323 map_reduce_dataset(Mapper, py_cla_cont(PyFunction,Args), CID, Result).
324
325 py_cla_cont(PyFunction,Args, Ok, _{stats:Result, py_coverage:Coverage}) :-
326 python_apply(PyFunction,[Ok|Args],Reply),
327 Reply = _{result:Result, stats:Coverage}.
328
329
330 % CLA mappers
331 rec_pitch_hist(W,Rec,Rec-Hist) :- transform_op(transcription,pitch_hist(W),Rec,Hist).
332
333 dense_pitch_hist(Min,Max,W,Rec,DenseHist) :-
334 transform_op(transcription,pitch_hist(W),Rec,SparseHist),
335 sparse_to_dense(Min,Max,SparseHist,DenseHist).
336
337 dense_freq_hist(ml,Map,W,Rec,Counts) :-
338 transform_op(transcription(1),freq_hist(Map,W),Rec,Counts).
339 dense_freq_hist(r,Map,W,Rec,Counts) :-
340 transform_op(transcription(1),freq_hist_r(Map,W),Rec,Counts).
341
342 tempo_hist(ml,DT,Map,Rec,Counts) :-
343 transform_op(tempo,tempo_hist(DT,Map),Rec,Result),
344 Result=_-Counts.
345 tempo_hist(r,DT,Map,Rec,Counts) :-
346 transform_op(tempo,tempo_hist_r(DT,Map),Rec,Result),
347 Result=_-Counts.
348
349 tempo_curve(ml,N,Rec,Values) :-
350 transform_op(tempo,normalised_tempo(N),Rec,Result),
351 Result=_-Values.
352 tempo_curve(r,N,Rec,Values) :-
353 transform_op(tempo,normalised_tempo_r(N),Rec,Result),
354 Result=_-Values.
355
356 transcription_tonic_duration(Rec, _{transcription: Transcription, tonic: Tonic, duration:0 }) :-
357 tagged(transcription,Rec,Transcription),
358 tagged(tonic,Rec,Tonic).
359
360 transcription_date(Rec, _{transcription: Transcription, date:Date}) :-
361 tagged(transcription(1),Rec,Transcription),
362 insist(recording_property(Rec,date,Date),missing_property(Rec,date)).
363
364 keys_chords(Rec, _{keys: Keys, chords:Chords}) :-
365 tagged(key,Rec,Keys),
366 tagged(chords,Rec,Chords).
367
368 similarity_bundle(Rec, _{chromagram: Chromagram, mfcc:Mfcc, keys: Keys, chords:Chords, list:_{uri:Rec, label:Label}}) :-
369 % nb_getval(vamp_on_demand,Vamp),
370 % concurrent_maplist(tagged_parallel(Vamp,Rec),[chromagram,mfcc,key,chords],[Chromagram,Mfcc,Keys,Chords]),
371 maplist(tagged,[chromagram,mfcc,key,chords],[Rec,Rec,Rec,Rec],[Chromagram,Mfcc,Keys,Chords]),
372 insist(recording_property(Rec,label,Label),missing_property(Rec,label)).
373
374 tagged_parallel(Vamp,Rec,Transform,Result) :-
375 nb_setval(vamp_on_demand,Vamp),
376 with_progress_stack(tagged(Transform,Rec,Result)).
377
378 list_places(Rec, _{place:Place,list:_{uri:Rec, label:Label}}) :-
379 insist(recording_property(Rec,place,Place),missing_property(Rec,place)),
380 insist(recording_property(Rec,label,Label),missing_property(Rec,label)).
381
382 % for later...
383 % tagged_list(Spec,Rec,Dict) :-
384 % maplist(tagged_item(Rec),Spec,Pairs),
385 % dict_create(Dict,_,Pairs).
386 % tagged_item(Rec,Key:Transform,Key:Value) :- tagged(Transform,Rec,Value).
387
388 tagged(Transform,Input,csv{value:Path}) :-
389 transform_computation(Transform,Input,R), uri_absolute_path(R,Path).
390
391 % ---------------------------------------------------
392
393 :- initialization time(memo_attach(memo(perspectives),[])).
394
395 :- persistent_memo cla_memo(+spec:ground,+cid:atom,-result:any).
396 cla_memo(Spec,CID,Result) :-
397 debug(perspectives(cla),'cla_mem: ~q',[call(Spec,CID,Result)]),
398 with_progress_stack(call(Spec,CID,Result)).
399
400 %% mem(+Spec:pred(+cid,-A),+Opts:options,+CID:cid,-Result:A) is det.
401 %
402 % Asynchronous memoised collection-level computation.
403 % Spec must be a ground term that can be called with two arguments: the id of a
404 % collection and a variable, which must be bound to an arbitrary result term on exit.
405 % If the computation has already be done and memoised in cla_memo/3, then the result is
406 % retrieved. Otherwise, the computation is started asynchronously and an exception
407 % describing the state of the computation will be thrown.
408 % ==
409 % * dml_error(10, _{status:already_waiting, position:n})
410 % the goal was previously added and is now waiting at position n in the queue.
411 % * dml_error(11, _{status:already_running, progress:Progress})
412 % the goal was previously added and is currently running, with some progress information.
413 % * dml_error(12, _{status:initiate, position:N})
414 % Means the goal has been added to the work queue of the thread pool at position N.
415 % ==
416 % Options are options passed to control interaction with async_memo.
417
418 mem(Spec,Opts,CID,Result) :-
419 option(vamp_on_demand(V), Opts, false),
420 async_memo(vis_cla,cla_memo(Spec,CID,Result),Status,
421 [ progress_levels([elapsed,summary,partial_result]),
422 globals([vamp_on_demand-V])|Opts ]),
423 ( Status=done(_-ok) -> true
424 ; status_response(Status,Code,Dict),
425 ( Status=done(_) -> Dict1=Dict
426 ; estimate_run_time(Spec,CID,ERT),
427 put_dict(ert,Dict,ERT,Dict1)
428 ),
429 throw(dml_error(Code,Dict1))
430 ).
431
432 % very crude estimate
433 estimate_run_time(Spec,CID0,ERT) :-
434 findall(Size-Dur, ( browse(cla_memo(Spec,CID,_),comp(_,_,Dur)-ok),
435 dataset_size(CID,Size)), Pairs),
436 length(Pairs,N),
437 ( N=0 -> ERT is -1
438 ; maplist(computations:pair,Sizes,Durs,Pairs),
439 sumlist(Sizes,TotalSize),
440 sumlist(Durs,TotalDur),
441 dataset_size(CID0,Size),
442 ERT is TotalDur*Size/TotalSize
443 ).
444
445 status_response(spawned(ID,Pos), 12, Info) :-
446 Info = _{status:initiated, id:ID, position:Pos }.
447 status_response(waiting(ID,T,Pos), 10, Info) :-
448 Info = _{status:already_waiting, id:ID, submit_time:TS, position:Pos },
449 time_to_string(T,TS).
450 status_response(running(ID,TStart,_,nothing), 11, Info) :-
451 Info = _{ id:ID, status:already_running, start_time:TS },
452 time_to_string(TStart,TS).
453 status_response(running(ID,TStart,_,just(Time-[Progress,Partial])), 11, Info) :-
454 maplist(progress_json,Progress,Progs),
455 time_to_string(TStart,TS),
456 Elapsed is Time-TStart,
457 ( member(stepwise(_,Done/Total),Progress), Done>0
458 -> ETA is Elapsed*(Total-Done)/Done
459 ; ETA is -1
460 ),
461 Info1 = _{ id:ID
462 , status:already_running
463 , start_time:TS
464 , elapsed_time:Elapsed
465 , progress:Progs
466 , eta:ETA
467 },
468 ( Partial=just(R)
469 -> put_dict(partial_result,Info1,R,Info)
470 ; Info=Info1
471 ).
472
473 status_response(recomputing(ID,Pos,Meta), 13, Info) :-
474 Info = _{status:recomputing, id:ID, position:Pos, meta:MD},
475 meta_dict(Meta,MD).
476 status_response(done(Meta), 14, _{status:failed, meta:MD}) :-
477 meta_dict(Meta,MD).
478
479 meta_dict(comp(_,Time,Dur)-Result, _{ date: Date, duration:Dur, reason:Reason}) :-
480 format_result(Result,Reason),
481 time_to_string(Time,Date).
482
483 format_result(fail,'Unspecified failure').
484 format_result(ex(Ex),Description) :- message_to_string(Ex,Description).
485
486 time_to_string(Time,String) :- format_time(string(String),'%FT%T%:z',Time).
487
488 progress_json(A,A) :- atomic(A), !.
489 progress_json(stepwise(Desc,Done/Total), _{ task:Task, total:Total, done:Done }) :- !,
490 progress_json(Desc,Task).
491 progress_json(T,A) :- message_to_string(T,A).
492
493 prolog:message(map_fold(_Mapper,_Folder)) --> ['Map-fold'].
494
495 :- multifile thread_pool:create_pool/1.
496 thread_pool:create_pool(vis_cla) :-
497 current_prolog_flag(cpu_count,N),
498 thread_pool_create(vis_cla, N, [backlog(20)]).
499
500 % ------------ computations with progress ------------------
501
502 map_reduce_dataset(Mapper,Reducer,CID,Result) :-
503 dataset_map_fold_reduce(CID,Mapper,append_dl,with_dl(Reducer),H-H,Result).
504
505 append_dl(HH-TT,H-HH,H-TT).
506 with_dl(P,H-[],A) :- call(P,H,A).
507 with_dl(P,H-[],A,B) :- call(P,H,A,B).
508
509 dataset_map_fold_reduce(CID,Mapper,Folder,Reducer,S0,Result) :-
510 dataset_items(CID,Items),
511 with_cont( 'Map-fold-reduce',
512 map_fold_with_progress( safe_call(Mapper),
513 safe_fold(Folder),
514 Items, s(0,F-F,E-E,S0)),
515 reduce_cont(Reducer), Result).
516
517 reduce_cont(Reducer,s(NOk,Failed-[],Erroneous-[],S), R) :-
518 ( NOk>0
519 -> simple_task(reducing(Reducer),call(Reducer,S,R1)),
520 put_coverage(NOk,Failed,Erroneous,R1,R)
521 ; put_coverage(NOk,Failed,Erroneous,_{status:'no successfully mapped items'},D),
522 throw(dml_error(20, D))
523 ).
524
525 put_coverage(NOk,Failed,Erroneous,R1,R) :-
526 put_dict(coverage,R1,_{ok_count:NOk, failed:Failed, errors:Erroneous},R).
527
528 safe_call(Mapper,X,Z) :-
529 ( catch((call(Mapper,X,Y), Z=ok(X,Y)), Ex,
530 ( Ex=abort(_) -> throw(Ex)
531 ; Z=error(X,Ex))), !
532 ; Z=fail(X)
533 ).
534
535 safe_fold(Folder,Items,s(NOk1,FH-FT1,EH-ET1,S1),s(NOk2,FH-FT2,EH-ET2,S2)) :-
536 seqmap(partition,Items,s(NOk1,OkH,FT1,ET1),s(NOk2,OkT,FT2,ET2)),
537 call(Folder,OkH-OkT,S1,S2).
538
539 partition(ok(_,X),s(N,[X|O],F,E),s(M,O,F,E)) :- M is N+1.
540 partition(fail(X),s(N,O,[X|F],E),s(N,O,F,E)).
541 partition(error(X,Ex),s(N,O,F,[_{item:X, error:Msg}|E]),s(N,O,F,E)) :- message_to_string(Ex,Msg).
542