2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2010,2011,2012,2013,2014 by the GROMACS development team.
5 * Copyright (c) 2015,2018,2019,2020,2021, by the GROMACS development team, led by
6 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
7 * and including many others, as listed in the AUTHORS file in the
8 * top-level source directory and at http://www.gromacs.org.
10 * GROMACS is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public License
12 * as published by the Free Software Foundation; either version 2.1
13 * of the License, or (at your option) any later version.
15 * GROMACS is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with GROMACS; if not, see
22 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
23 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
25 * If you want to redistribute modifications to GROMACS, please
26 * consider that scientific software is very special. Version
27 * control is crucial - bugs must be traceable. We will be happy to
28 * consider code for inclusion in the official distribution, but
29 * derived work must not be called official GROMACS. Details are found
30 * in the README & COPYING files - if they are missing, get the
31 * official version at http://www.gromacs.org.
33 * To help us fund GROMACS development, we humbly ask that you cite
34 * the research papers on the package. Check out http://www.gromacs.org.
38 * Declares gmx::TrajectoryAnalysisModule and
39 * gmx::TrajectoryAnalysisModuleData.
41 * \author Teemu Murtola <teemu.murtola@gmail.com>
43 * \ingroup module_trajectoryanalysis
45 #ifndef GMX_TRAJECTORYANALYSIS_ANALYSISMODULE_H
46 #define GMX_TRAJECTORYANALYSIS_ANALYSISMODULE_H
52 #include "gromacs/selection/selection.h" // For gmx::SelectionList
60 class AbstractAnalysisData;
62 class AnalysisDataHandle;
63 class AnalysisDataParallelOptions;
64 class IOptionsContainer;
66 class SelectionCollection;
67 class TopologyInformation;
68 class TrajectoryAnalysisModule;
69 class TrajectoryAnalysisSettings;
72 * Base class for thread-local data storage during trajectory analysis.
74 * Thread-local storage of data handles and selections is implemented in this
75 * class; TrajectoryAnalysisModule instances can access the thread-local values
76 * in their TrajectoryAnalysisModule::analyzeFrame() method using dataHandle()
77 * and parallelSelection().
79 * \see TrajectoryAnalysisModule::startFrames()
80 * \see TrajectoryAnalysisModule::analyzeFrame()
81 * \see TrajectoryAnalysisModule::finishFrames()
84 * \ingroup module_trajectoryanalysis
86 class TrajectoryAnalysisModuleData
89 virtual ~TrajectoryAnalysisModuleData();
92 * Performs any finishing actions after all frames have been processed.
94 * \throws unspecified Implementation may throw exceptions to indicate
97 * This function is called immediately before the destructor, after
98 * TrajectoryAnalysisModule::finishFrames().
99 * Derived classes should implement any final operations that need to
100 * be done after successful analysis.
101 * All implementations should call finishDataHandles().
103 virtual void finish() = 0;
106 * Returns a data handle for a given dataset.
108 * \param[in] data Analysis data object.
109 * \returns Data handle for \p data stored in this thread-local data.
111 * \p data should have previously been registered with
112 * TrajectoryAnalysisModule::registerAnalysisDataset().
113 * If \p data has zero columns in all data sets, the returned data
118 AnalysisDataHandle dataHandle(const AnalysisData& data);
120 * Returns a selection that corresponds to the given selection.
122 * \param[in] selection Global selection object.
123 * \returns Selection object corresponding to this thread-local data.
125 * \p selection is the selection object that was obtained from
126 * SelectionOption. The return value is the corresponding selection
127 * in the selection collection with which this data object was
132 Selection parallelSelection(const Selection& selection);
134 * Returns a set of selection that corresponds to the given selections.
136 * \throws std::bad_alloc if out of memory.
138 * Works as parallelSelection(), but for a list of selections at once.
140 * \see parallelSelection()
142 SelectionList parallelSelections(const SelectionList& selections);
146 * Initializes thread-local storage for data handles and selections.
148 * \param[in] module Analysis module to use for data objects.
149 * \param[in] opt Data parallelization options.
150 * \param[in] selections Thread-local selection collection.
151 * \throws std::bad_alloc if out of memory.
152 * \throws unspecified Can throw any exception thrown by
153 * AnalysisData::startData().
155 * Calls AnalysisData::startData() on all data objects registered with
156 * TrajectoryAnalysisModule::registerAnalysisDataset() in \p module.
157 * The handles are accessible through dataHandle().
159 TrajectoryAnalysisModuleData(TrajectoryAnalysisModule* module,
160 const AnalysisDataParallelOptions& opt,
161 const SelectionCollection& selections);
164 * Calls finishData() on all data handles.
166 * \throws unspecified Can throw any exception thrown by
167 * AnalysisDataHandle::finishData().
169 * This function should be called from the implementation of finish()
172 void finishDataHandles();
177 std::unique_ptr<Impl> impl_;
180 //! Smart pointer to manage a TrajectoryAnalysisModuleData object.
181 typedef std::unique_ptr<TrajectoryAnalysisModuleData> TrajectoryAnalysisModuleDataPointer;
184 * Base class for trajectory analysis modules.
186 * Trajectory analysis methods should derive from this class and override the
187 * necessary virtual methods to implement initialization (initOptions(),
188 * optionsFinished(), initAnalysis(), initAfterFirstFrame()), per-frame analysis
189 * (analyzeFrame()), and final processing (finishFrames(), finishAnalysis(),
192 * For parallel analysis using threads, only a single object is constructed,
193 * but the methods startFrames(), analyzeFrame() and finishFrames() are called
194 * in each thread. Frame-local data should be initialized in startFrames() and
195 * stored in a class derived from TrajectoryAnalysisModuleData that is passed
196 * to the other methods. The default implementation of startFrames() can be
197 * used if only data handles and selections need to be thread-local.
199 * To get the full benefit from this class,
200 * \ref module_analysisdata "analysis data objects" and
201 * \ref module_selection "selections" should be used in the implementation.
202 * See the corresponding modules' documentation for details of how they work.
204 * Typical way of using AnalysisData in derived classes is to have the
205 * AnalysisData object as a member variable and register it using
206 * registerAnalysisDataset(). Analysis modules are initialized in
207 * initAnalysis() and the processing chain is initialized. If any of the
208 * modules is required, e.g., for post-processing in finishAnalysis(), it can
209 * be stored in a member variable. To add data to the data object in
210 * analyzeFrame(), a data handle is obtained using
211 * TrajectoryAnalysisModuleData::dataHandle().
213 * Typical way of using selections in derived classes is to have the required
214 * \ref Selection objects (or ::SelectionList objects) as member variables, and
215 * add the required selection options in initOptions(). These member variables
216 * can be accessed in initAnalysis() to get general information about the
217 * selections. In analyzeFrame(), these selection objects should not be used
218 * directly, but instead TrajectoryAnalysisModuleData::parallelSelection()
219 * should be used to obtain a selection object that works correctly also for
222 * Derived classes should use exceptions to indicate errors in the virtual
226 * \ingroup module_trajectoryanalysis
228 class TrajectoryAnalysisModule
231 virtual ~TrajectoryAnalysisModule();
234 * Initializes options understood by the module.
236 * \param[in,out] options Options object to add the options to.
237 * \param[in,out] settings Settings to pass to and from the module.
239 * This method is called first after the constructor, and it should
240 * add options understood by the module to \p options. Output values
241 * from options (including selections) should be stored in member
244 * In addition to initializing the options, this method can also
245 * provide information about the module's requirements using the
246 * \p settings object; see TrajectoryAnalysisSettings for more details.
248 * If settings depend on the option values provided by the user, see
251 virtual void initOptions(IOptionsContainer* options, TrajectoryAnalysisSettings* settings) = 0;
253 * Called after all option values have been set.
255 * \param[in,out] settings Settings to pass to and from the module.
257 * This method is called after option values have been assigned (but
258 * interactive selection input has not yet been performed).
260 * If the module needs to change settings that affect topology loading
261 * (can be done using the \p settings object) or selection
262 * initialization (can be done using SelectionOptionInfo) based on
263 * option values, this method has to be overridden.
265 * The default implementation does nothing.
267 virtual void optionsFinished(TrajectoryAnalysisSettings* settings);
269 * Initializes the analysis.
271 * \param[in] settings Settings to pass to and from the module.
272 * \param[in] top Topology information.
274 * When this function is called, selections have been initialized based
275 * on user input, and a topology has been loaded if provided by the
276 * user. For dynamic selections, the selections have been evaluated to
277 * the largest possible selection, i.e., the selections passed to
278 * analyzeFrame() are always a subset of the selections provided here.
280 virtual void initAnalysis(const TrajectoryAnalysisSettings& settings,
281 const TopologyInformation& top) = 0;
283 * Performs additional initialization after reading the first frame.
285 * When this function is called, selections are the same as in
286 * initAnalysis(), i.e., they have not been evaluated for the first
289 * It is necessary to override this method only if the module needs to
290 * do initialization for which it requires data from the first frame.
292 * The default implementation does nothing.
294 virtual void initAfterFirstFrame(const TrajectoryAnalysisSettings& settings, const t_trxframe& fr);
297 * Starts the analysis of frames.
299 * \param[in] opt Parallel options
300 * \param[in] selections Frame-local selection collection object.
301 * \returns Data structure for thread-local data.
303 * This function is necessary only for threaded parallelization.
304 * It is called once for each thread and should initialize a class that
305 * contains any required frame-local data in the returned value.
306 * The default implementation creates a basic data structure that holds
307 * thread-local data handles for all data objects registered with
308 * registerAnalysisDataset(), as well as the thread-local selection
309 * collection. These can be accessed in analyzeFrame() using the
310 * methods in TrajectoryAnalysisModuleData.
311 * If other thread-local data is needed, this function should be
312 * overridden and it should create an instance of a class derived from
313 * TrajectoryAnalysisModuleData.
315 * \see TrajectoryAnalysisModuleData
317 virtual TrajectoryAnalysisModuleDataPointer startFrames(const AnalysisDataParallelOptions& opt,
318 const SelectionCollection& selections);
320 * Analyzes a single frame.
322 * \param[in] frnr Frame number, a zero-based index that
323 * uniquely identifies the frame.
324 * \param[in] fr Current frame.
325 * \param[in] pbc Periodic boundary conditions for \p fr.
326 * \param[in,out] pdata Data structure for frame-local data.
328 * This method is called once for each frame to be analyzed, and should
329 * analyze the positions provided in the selections. Data handles and
330 * selections should be obtained from the \p pdata structure.
332 * For threaded analysis, this method is called asynchronously in
333 * different threads to analyze different frames. The \p pdata
334 * structure is one of the structures created with startFrames(),
335 * but no assumptions should be made about which of these data
336 * structures is used. It is guaranteed that two instances of
337 * analyzeFrame() are not running concurrently with the same \p pdata
339 * Any access to data structures not stored in \p pdata should be
340 * designed to be thread-safe.
342 virtual void analyzeFrame(int frnr, const t_trxframe& fr, t_pbc* pbc, TrajectoryAnalysisModuleData* pdata) = 0;
344 * Finishes the analysis of frames.
346 * \param[in] pdata Data structure for thread-local data.
348 * This method is called once for each call of startFrames(), with the
349 * data structure returned by the corresponding startFrames().
350 * The \p pdata object should be destroyed by the caller after this
351 * function has been called.
353 * You only need to override this method if you need custom
354 * operations to combine data from the frame-local data structures
355 * to get the final result. In such cases, the data should be
356 * aggregated in this function and stored in a member attribute.
358 * The default implementation does nothing.
362 virtual void finishFrames(TrajectoryAnalysisModuleData* pdata);
365 * Postprocesses data after frames have been read.
367 * \param[in] nframes Total number of frames processed.
369 * This function is called after all finishFrames() calls have been
371 * \p nframes will equal the number of calls to analyzeFrame() that
374 virtual void finishAnalysis(int nframes) = 0;
376 * Writes output into files and/or standard output/error.
378 * All output from the module, excluding data written out for each
379 * frame during analyzeFrame(), should be confined into this function.
380 * This function is guaranteed to be called only after
383 virtual void writeOutput() = 0;
386 * Returns the number of datasets provided by the module.
390 int datasetCount() const;
392 * Returns a vector with the names of datasets provided by the module.
396 const std::vector<std::string>& datasetNames() const;
398 * Returns a pointer to the data set \p index.
400 * \param[in] index Data set to query for.
401 * \returns Reference to the requested data set.
402 * \throws APIError if \p index is not valid.
404 * \p index should be >= 0 and < datasetCount().
406 * The return value is not const to allow callers to add modules to the
407 * data sets. However, the AbstractAnalysisData interface does not
408 * provide any means to alter the data, so the module does not need to
409 * care about external modifications.
411 AbstractAnalysisData& datasetFromIndex(int index) const;
413 * Returns a pointer to the data set with name \p name
415 * \param[in] name Data set to query for.
416 * \returns Reference to the requested data set.
417 * \throws APIError if \p name is not valid.
419 * \p name should be one of the names returned by datasetNames().
421 * The return value is not const to allow callers to add modules to the
422 * data sets. However, the AbstractAnalysisData interface does not
423 * provide any means to alter the data, so the module does not need to
424 * care about external modifications.
426 AbstractAnalysisData& datasetFromName(const char* name) const;
428 * Processes data in AnalysisData objects in serial for each frame.
430 * \param[in] frameIndex Index of the frame that has been finished.
432 * This method is called by the framework in order for each frame,
433 * after the analysis for that frame has been finished. These calls
434 * always execute in serial and in sequential frame order, even during
435 * parallel analysis where multiple analyzeFrame() calls may be
436 * executing concurrently.
438 * \see AnalysisData::finishFrameSerial()
440 void finishFrameSerial(int frameIndex);
444 * Initializes the dataset registration mechanism.
446 * \throws std::bad_alloc if out of memory.
448 TrajectoryAnalysisModule();
451 * Registers a dataset that exports data.
453 * \param data Data object to register.
454 * \param[in] name Name to register the dataset with.
455 * \throws std::bad_alloc if out of memory.
457 * Registers \p data as a dataset that provides output from the
458 * analysis module. Callers for the module can access the dataset
459 * with datasetFromName() using \p name as an AbstractAnalysisData
460 * object. This allows them to add their own data modules to do extra
463 * \p name must be unique across all calls within the same
464 * TrajectoryAnalysisModule instance.
466 void registerBasicDataset(AbstractAnalysisData* data, const char* name);
468 * Registers a parallelized dataset that exports data.
470 * \param data AnalysisData object to register.
471 * \param[in] name Name to register the dataset with.
472 * \throws std::bad_alloc if out of memory.
474 * This method works as registerBasicDataset(), but additionally allows
475 * data handles for \p data to be accessed using
476 * TrajectoryAnalysisData.
478 * \see registerBasicDataset()
480 void registerAnalysisDataset(AnalysisData* data, const char* name);
485 std::unique_ptr<Impl> impl_;
488 * Needed to access the registered analysis data sets.
490 friend class TrajectoryAnalysisModuleData;
493 //! Smart pointer to manage a TrajectoryAnalysisModule.
494 typedef std::unique_ptr<TrajectoryAnalysisModule> TrajectoryAnalysisModulePointer;