2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2010,2011,2012,2013,2014,2015, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
37 * Declares gmx::TrajectoryAnalysisModule and
38 * gmx::TrajectoryAnalysisModuleData.
40 * \author Teemu Murtola <teemu.murtola@gmail.com>
42 * \ingroup module_trajectoryanalysis
44 #ifndef GMX_TRAJECTORYANALYSIS_ANALYSISMODULE_H
45 #define GMX_TRAJECTORYANALYSIS_ANALYSISMODULE_H
50 #include <boost/shared_ptr.hpp>
52 #include "gromacs/selection/selection.h" // For gmx::SelectionList
53 #include "gromacs/utility/classhelpers.h"
61 class AbstractAnalysisData;
63 class AnalysisDataHandle;
64 class AnalysisDataParallelOptions;
65 class IOptionsContainer;
67 class SelectionCollection;
68 class TopologyInformation;
69 class TrajectoryAnalysisModule;
70 class TrajectoryAnalysisSettings;
73 * Base class for thread-local data storage during trajectory analysis.
75 * Thread-local storage of data handles and selections is implemented in this
76 * class; TrajectoryAnalysisModule instances can access the thread-local values
77 * in their TrajectoryAnalysisModule::analyzeFrame() method using dataHandle()
78 * and parallelSelection().
80 * \see TrajectoryAnalysisModule::startFrames()
81 * \see TrajectoryAnalysisModule::analyzeFrame()
82 * \see TrajectoryAnalysisModule::finishFrames()
85 * \ingroup module_trajectoryanalysis
87 class TrajectoryAnalysisModuleData
90 virtual ~TrajectoryAnalysisModuleData();
93 * Performs any finishing actions after all frames have been processed.
95 * \throws unspecified Implementation may throw exceptions to indicate
98 * This function is called immediately before the destructor, after
99 * TrajectoryAnalysisModule::finishFrames().
100 * Derived classes should implement any final operations that need to
101 * be done after successful analysis.
102 * All implementations should call finishDataHandles().
104 virtual void finish() = 0;
107 * Returns a data handle for a given dataset.
109 * \param[in] data Analysis data object.
110 * \returns Data handle for \p data stored in this thread-local data.
112 * \p data should have previously been registered with
113 * TrajectoryAnalysisModule::registerAnalysisDataset().
114 * If \p data has zero columns in all data sets, the returned data
119 AnalysisDataHandle dataHandle(const AnalysisData &data);
121 * Returns a selection that corresponds to the given selection.
123 * \param[in] selection Global selection object.
124 * \returns Selection object corresponding to this thread-local data.
126 * \p selection is the selection object that was obtained from
127 * SelectionOption. The return value is the corresponding selection
128 * in the selection collection with which this data object was
133 Selection parallelSelection(const Selection &selection);
135 * Returns a set of selection that corresponds to the given selections.
137 * \throws std::bad_alloc if out of memory.
139 * Works as parallelSelection(), but for a list of selections at once.
141 * \see parallelSelection()
143 SelectionList parallelSelections(const SelectionList &selections);
147 * Initializes thread-local storage for data handles and selections.
149 * \param[in] module Analysis module to use for data objects.
150 * \param[in] opt Data parallelization options.
151 * \param[in] selections Thread-local selection collection.
152 * \throws std::bad_alloc if out of memory.
153 * \throws unspecified Can throw any exception thrown by
154 * AnalysisData::startData().
156 * Calls AnalysisData::startData() on all data objects registered with
157 * TrajectoryAnalysisModule::registerAnalysisDataset() in \p module.
158 * The handles are accessible through dataHandle().
160 TrajectoryAnalysisModuleData(TrajectoryAnalysisModule *module,
161 const AnalysisDataParallelOptions &opt,
162 const SelectionCollection &selections);
165 * Calls finishData() on all data handles.
167 * \throws unspecified Can throw any exception thrown by
168 * AnalysisDataHandle::finishData().
170 * This function should be called from the implementation of finish()
173 void finishDataHandles();
178 PrivateImplPointer<Impl> impl_;
181 //! Smart pointer to manage a TrajectoryAnalysisModuleData object.
182 typedef boost::shared_ptr<TrajectoryAnalysisModuleData>
183 TrajectoryAnalysisModuleDataPointer;
186 * Base class for trajectory analysis modules.
188 * Trajectory analysis methods should derive from this class and override the
189 * necessary virtual methods to implement initialization (initOptions(),
190 * optionsFinished(), initAnalysis(), initAfterFirstFrame()), per-frame analysis
191 * (analyzeFrame()), and final processing (finishFrames(), finishAnalysis(),
194 * For parallel analysis using threads, only a single object is constructed,
195 * but the methods startFrames(), analyzeFrame() and finishFrames() are called
196 * in each thread. Frame-local data should be initialized in startFrames() and
197 * stored in a class derived from TrajectoryAnalysisModuleData that is passed
198 * to the other methods. The default implementation of startFrames() can be
199 * used if only data handles and selections need to be thread-local.
201 * To get the full benefit from this class,
202 * \ref module_analysisdata "analysis data objects" and
203 * \ref module_selection "selections" should be used in the implementation.
204 * See the corresponding modules' documentation for details of how they work.
206 * Typical way of using AnalysisData in derived classes is to have the
207 * AnalysisData object as a member variable and register it using
208 * registerAnalysisDataset(). Analysis modules are initialized in
209 * initAnalysis() and the processing chain is initialized. If any of the
210 * modules is required, e.g., for post-processing in finishAnalysis(), it can
211 * be stored in a member variable. To add data to the data object in
212 * analyzeFrame(), a data handle is obtained using
213 * TrajectoryAnalysisModuleData::dataHandle().
215 * Typical way of using selections in derived classes is to have the required
216 * \ref Selection objects (or ::SelectionList objects) as member variables, and
217 * add the required selection options in initOptions(). These member variables
218 * can be accessed in initAnalysis() to get general information about the
219 * selections. In analyzeFrame(), these selection objects should not be used
220 * directly, but instead TrajectoryAnalysisModuleData::parallelSelection()
221 * should be used to obtain a selection object that works correctly also for
224 * Derived classes should use exceptions to indicate errors in the virtual
228 * \ingroup module_trajectoryanalysis
230 class TrajectoryAnalysisModule
233 virtual ~TrajectoryAnalysisModule();
236 * Initializes options understood by the module.
238 * \param[in,out] options Options object to add the options to.
239 * \param[in,out] settings Settings to pass to and from the module.
241 * This method is called first after the constructor, and it should
242 * add options understood by the module to \p options. Output values
243 * from options (including selections) should be stored in member
246 * In addition to initializing the options, this method can also
247 * provide information about the module's requirements using the
248 * \p settings object; see TrajectoryAnalysisSettings for more details.
250 * If settings depend on the option values provided by the user, see
253 virtual void initOptions(IOptionsContainer *options,
254 TrajectoryAnalysisSettings *settings) = 0;
256 * Called after all option values have been set.
258 * \param[in,out] settings Settings to pass to and from the module.
260 * This method is called after option values have been assigned (but
261 * interactive selection input has not yet been performed).
263 * If the module needs to change settings that affect topology loading
264 * (can be done using the \p settings object) or selection
265 * initialization (can be done using SelectionOptionInfo) based on
266 * option values, this method has to be overridden.
268 * The default implementation does nothing.
270 virtual void optionsFinished(TrajectoryAnalysisSettings *settings);
272 * Initializes the analysis.
274 * \param[in] settings Settings to pass to and from the module.
275 * \param[in] top Topology information.
277 * When this function is called, selections have been initialized based
278 * on user input, and a topology has been loaded if provided by the
279 * user. For dynamic selections, the selections have been evaluated to
280 * the largest possible selection, i.e., the selections passed to
281 * analyzeFrame() are always a subset of the selections provided here.
283 virtual void initAnalysis(const TrajectoryAnalysisSettings &settings,
284 const TopologyInformation &top) = 0;
286 * Performs additional initialization after reading the first frame.
288 * When this function is called, selections are the same as in
289 * initAnalysis(), i.e., they have not been evaluated for the first
292 * It is necessary to override this method only if the module needs to
293 * do initialization for which it requires data from the first frame.
295 * The default implementation does nothing.
297 virtual void initAfterFirstFrame(const TrajectoryAnalysisSettings &settings,
298 const t_trxframe &fr);
301 * Starts the analysis of frames.
304 * \param[in] selections Frame-local selection collection object.
305 * \returns Data structure for thread-local data.
307 * This function is necessary only for threaded parallelization.
308 * It is called once for each thread and should initialize a class that
309 * contains any required frame-local data in the returned value.
310 * The default implementation creates a basic data structure that holds
311 * thread-local data handles for all data objects registered with
312 * registerAnalysisDataset(), as well as the thread-local selection
313 * collection. These can be accessed in analyzeFrame() using the
314 * methods in TrajectoryAnalysisModuleData.
315 * If other thread-local data is needed, this function should be
316 * overridden and it should create an instance of a class derived from
317 * TrajectoryAnalysisModuleData.
319 * \see TrajectoryAnalysisModuleData
321 virtual TrajectoryAnalysisModuleDataPointer startFrames(
322 const AnalysisDataParallelOptions &opt,
323 const SelectionCollection &selections);
325 * Analyzes a single frame.
327 * \param[in] frnr Frame number, a zero-based index that
328 * uniquely identifies the frame.
329 * \param[in] fr Current frame.
330 * \param[in] pbc Periodic boundary conditions for \p fr.
331 * \param[in,out] pdata Data structure for frame-local data.
333 * This method is called once for each frame to be analyzed, and should
334 * analyze the positions provided in the selections. Data handles and
335 * selections should be obtained from the \p pdata structure.
337 * For threaded analysis, this method is called asynchronously in
338 * different threads to analyze different frames. The \p pdata
339 * structure is one of the structures created with startFrames(),
340 * but no assumptions should be made about which of these data
341 * structures is used. It is guaranteed that two instances of
342 * analyzeFrame() are not running concurrently with the same \p pdata
344 * Any access to data structures not stored in \p pdata should be
345 * designed to be thread-safe.
347 virtual void analyzeFrame(int frnr, const t_trxframe &fr, t_pbc *pbc,
348 TrajectoryAnalysisModuleData *pdata) = 0;
350 * Finishes the analysis of frames.
352 * \param[in] pdata Data structure for thread-local data.
354 * This method is called once for each call of startFrames(), with the
355 * data structure returned by the corresponding startFrames().
356 * The \p pdata object should be destroyed by the caller after this
357 * function has been called.
359 * You only need to override this method if you need custom
360 * operations to combine data from the frame-local data structures
361 * to get the final result. In such cases, the data should be
362 * aggregated in this function and stored in a member attribute.
364 * The default implementation does nothing.
368 virtual void finishFrames(TrajectoryAnalysisModuleData *pdata);
371 * Postprocesses data after frames have been read.
373 * \param[in] nframes Total number of frames processed.
375 * This function is called after all finishFrames() calls have been
377 * \p nframes will equal the number of calls to analyzeFrame() that
380 virtual void finishAnalysis(int nframes) = 0;
382 * Writes output into files and/or standard output/error.
384 * All output from the module, excluding data written out for each
385 * frame during analyzeFrame(), should be confined into this function.
386 * This function is guaranteed to be called only after
389 virtual void writeOutput() = 0;
391 typedef std::vector<TrajectoryAnalysisModule*> Batch;
392 virtual const Batch getBatch() { return Batch(); }
394 virtual const std::vector<char*> getArgv(int) { return std::vector<char*>(); }
397 * Returns the name of the analysis module.
401 const char *name() const;
403 * Returns short description for the analysis module.
407 const char *description() const;
409 * Returns the number of datasets provided by the module.
413 int datasetCount() const;
415 * Returns a vector with the names of datasets provided by the module.
419 const std::vector<std::string> &datasetNames() const;
421 * Returns a pointer to the data set \p index.
423 * \param[in] index Data set to query for.
424 * \returns Reference to the requested data set.
425 * \throws APIError if \p index is not valid.
427 * \p index should be >= 0 and < datasetCount().
429 * The return value is not const to allow callers to add modules to the
430 * data sets. However, the AbstractAnalysisData interface does not
431 * provide any means to alter the data, so the module does not need to
432 * care about external modifications.
434 AbstractAnalysisData &datasetFromIndex(int index) const;
436 * Returns a pointer to the data set with name \p name
438 * \param[in] name Data set to query for.
439 * \returns Reference to the requested data set.
440 * \throws APIError if \p name is not valid.
442 * \p name should be one of the names returned by datasetNames().
444 * The return value is not const to allow callers to add modules to the
445 * data sets. However, the AbstractAnalysisData interface does not
446 * provide any means to alter the data, so the module does not need to
447 * care about external modifications.
449 AbstractAnalysisData &datasetFromName(const char *name) const;
451 * Processes data in AnalysisData objects in serial for each frame.
453 * \param[in] frameIndex Index of the frame that has been finished.
455 * This method is called by the framework in order for each frame,
456 * after the analysis for that frame has been finished. These calls
457 * always execute in serial and in sequential frame order, even during
458 * parallel analysis where multiple analyzeFrame() calls may be
459 * executing concurrently.
461 * \see AnalysisData::finishFrameSerial()
463 void finishFrameSerial(int frameIndex);
467 * Initializes the dataset registration mechanism.
469 * \param[in] name Name for the module.
470 * \param[in] description One-line description for the module.
471 * \throws std::bad_alloc if out of memory.
473 TrajectoryAnalysisModule(const char *name, const char *description);
476 * Registers a dataset that exports data.
478 * \param data Data object to register.
479 * \param[in] name Name to register the dataset with.
480 * \throws std::bad_alloc if out of memory.
482 * Registers \p data as a dataset that provides output from the
483 * analysis module. Callers for the module can access the dataset
484 * with datasetFromName() using \p name as an AbstractAnalysisData
485 * object. This allows them to add their own data modules to do extra
488 * \p name must be unique across all calls within the same
489 * TrajectoryAnalysisModule instance.
491 void registerBasicDataset(AbstractAnalysisData *data, const char *name);
493 * Registers a parallelized dataset that exports data.
495 * \param data AnalysisData object to register.
496 * \param[in] name Name to register the dataset with.
497 * \throws std::bad_alloc if out of memory.
499 * This method works as registerBasicDataset(), but additionally allows
500 * data handles for \p data to be accessed using
501 * TrajectoryAnalysisData.
503 * \see registerBasicDataset()
505 void registerAnalysisDataset(AnalysisData *data, const char *name);
510 PrivateImplPointer<Impl> impl_;
513 * Needed to access the registered analysis data sets.
515 friend class TrajectoryAnalysisModuleData;
518 //! Smart pointer to manage a TrajectoryAnalysisModule.
519 typedef boost::shared_ptr<TrajectoryAnalysisModule>
520 TrajectoryAnalysisModulePointer;