1 #include <petsc/private/regressorimpl.h> 2 3 PetscBool PetscRegressorRegisterAllCalled = PETSC_FALSE; 4 PetscFunctionList PetscRegressorList = NULL; 5 6 PetscClassId PETSCREGRESSOR_CLASSID; 7 8 /* Logging support */ 9 PetscLogEvent PetscRegressor_SetUp, PetscRegressor_Fit, PetscRegressor_Predict; 10 11 /*@C 12 PetscRegressorRegister - Adds a method to the `PetscRegressor` package. 13 14 Not collective 15 16 Input Parameters: 17 + sname - name of a new user-defined regressor 18 - function - routine to create method context 19 20 Notes: 21 `PetscRegressorRegister()` may be called multiple times to add several user-defined regressors. 22 23 Example Usage: 24 .vb 25 PetscRegressorRegister("my_regressor",MyRegressorCreate); 26 .ve 27 28 Then, your regressor can be chosen with the procedural interface via 29 .vb 30 PetscRegressorSetType(regressor,"my_regressor") 31 .ve 32 or at runtime via the option 33 .vb 34 -regressor_type my_regressor 35 .ve 36 37 Level: advanced 38 39 .seealso: `PetscRegressorRegisterAll()` 40 @*/ 41 PetscErrorCode PetscRegressorRegister(const char sname[], PetscErrorCode (*function)(PetscRegressor)) 42 { 43 PetscFunctionBegin; 44 PetscCall(PetscRegressorInitializePackage()); 45 PetscCall(PetscFunctionListAdd(&PetscRegressorList, sname, function)); 46 PetscFunctionReturn(PETSC_SUCCESS); 47 } 48 49 /*@ 50 PetscRegressorCreate - Creates a `PetscRegressor` object. 51 52 Collective 53 54 Input Parameter: 55 . comm - the MPI communicator that will share the `PetscRegressor` object 56 57 Output Parameter: 58 . newregressor - the new `PetscRegressor` object 59 60 Level: beginner 61 62 .seealso: `PetscRegressorFit()`, `PetscRegressorPredict()`, `PetscRegressor` 63 @*/ 64 PetscErrorCode PetscRegressorCreate(MPI_Comm comm, PetscRegressor *newregressor) 65 { 66 PetscRegressor regressor; 67 68 PetscFunctionBegin; 69 PetscAssertPointer(newregressor, 2); 70 *newregressor = NULL; 71 PetscCall(PetscRegressorInitializePackage()); 72 73 PetscCall(PetscHeaderCreate(regressor, PETSCREGRESSOR_CLASSID, "PetscRegressor", "Regressor", "PetscRegressor", comm, PetscRegressorDestroy, PetscRegressorView)); 74 75 regressor->setupcalled = PETSC_FALSE; 76 regressor->fitcalled = PETSC_FALSE; 77 regressor->data = NULL; 78 regressor->training = NULL; 79 regressor->target = NULL; 80 PetscObjectParameterSetDefault(regressor, regularizer_weight, 1.0); // Default to regularizer weight of 1.0, usually the default in SciKit-learn 81 82 *newregressor = regressor; 83 PetscFunctionReturn(PETSC_SUCCESS); 84 } 85 86 /*@ 87 PetscRegressorView - Prints information about the `PetscRegressor` object 88 89 Collective 90 91 Input Parameters: 92 + regressor - the `PetscRegressor` context 93 - viewer - a `PetscViewer` context 94 95 Options Database Key: 96 . -regressor_view - Calls `PetscRegressorView()` at the end of `PetscRegressorFit()` 97 98 Level: beginner 99 100 Notes: 101 The available visualization contexts include 102 + `PETSC_VIEWER_STDOUT_SELF` - standard output (default) 103 - `PETSC_VIEWER_STDOUT_WORLD` - synchronized standard 104 output where only the first processor opens 105 the file. All other processors send their 106 data to the first processor to print. 107 108 .seealso: [](ch_regressor), `PetscRegressor`, `PetscViewerASCIIOpen()` 109 @*/ 110 PetscErrorCode PetscRegressorView(PetscRegressor regressor, PetscViewer viewer) 111 { 112 PetscBool isascii, isstring; 113 PetscRegressorType type; 114 115 PetscFunctionBegin; 116 PetscValidHeaderSpecific(regressor, PETSCREGRESSOR_CLASSID, 1); 117 if (!viewer) PetscCall(PetscViewerASCIIGetStdout(((PetscObject)regressor)->comm, &viewer)); 118 PetscValidHeaderSpecific(viewer, PETSC_VIEWER_CLASSID, 2); 119 PetscCheckSameComm(regressor, 1, viewer, 2); 120 121 PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERASCII, &isascii)); 122 PetscCall(PetscObjectTypeCompare((PetscObject)viewer, PETSCVIEWERSTRING, &isstring)); 123 if (isascii) { 124 PetscCall(PetscObjectPrintClassNamePrefixType((PetscObject)regressor, viewer)); 125 126 PetscCall(PetscViewerASCIIPushTab(viewer)); 127 PetscTryTypeMethod(regressor, view, viewer); 128 if (regressor->tao) PetscCall(TaoView(regressor->tao, viewer)); 129 PetscCall(PetscViewerASCIIPopTab(viewer)); 130 } else if (isstring) { 131 PetscCall(PetscRegressorGetType(regressor, &type)); 132 PetscCall(PetscViewerStringSPrintf(viewer, " PetscRegressorType: %-7.7s", type)); 133 } 134 PetscFunctionReturn(PETSC_SUCCESS); 135 } 136 137 /*@ 138 PetscRegressorViewFromOptions - View a `PetscRegressor` object based on values in the options database 139 140 Collective 141 142 Input Parameters: 143 + A - the `PetscRegressor` context 144 . obj - Optional object that provides the prefix for the options database 145 - name - command line option 146 147 Level: intermediate 148 149 .seealso: [](ch_regressor), `PetscRegressor`, `PetscRegressorView`, `PetscObjectViewFromOptions()`, `PetscRegressorCreate()` 150 @*/ 151 PetscErrorCode PetscRegressorViewFromOptions(PetscRegressor A, PetscObject obj, const char name[]) 152 { 153 PetscFunctionBegin; 154 PetscValidHeaderSpecific(A, PETSCREGRESSOR_CLASSID, 1); 155 PetscCall(PetscObjectViewFromOptions((PetscObject)A, obj, name)); 156 PetscFunctionReturn(PETSC_SUCCESS); 157 } 158 159 /*@ 160 PetscRegressorSetFromOptions - Sets `PetscRegressor` options from the options database. 161 162 Collective 163 164 Input Parameter: 165 . regressor - the `PetscRegressor` context 166 167 Options Database Keys: 168 . -regressor_type <type> - the particular type of regressor to be used; see `PetscRegressorType` for complete list 169 170 Level: beginner 171 172 Note: 173 This routine must be called before `PetscRegressorSetUp()` (or `PetscRegressorFit()`, which calls 174 the former) if the user is to be allowed to set the regressor type. 175 176 .seealso: `PetscRegressor`, `PetscRegressorCreate()` 177 @*/ 178 PetscErrorCode PetscRegressorSetFromOptions(PetscRegressor regressor) 179 { 180 PetscBool flg; 181 PetscRegressorType default_type = PETSCREGRESSORLINEAR; 182 char type[256]; 183 184 PetscFunctionBegin; 185 PetscValidHeaderSpecific(regressor, PETSCREGRESSOR_CLASSID, 1); 186 if (((PetscObject)regressor)->type_name) default_type = ((PetscObject)regressor)->type_name; 187 PetscObjectOptionsBegin((PetscObject)regressor); 188 /* Check for type from options */ 189 PetscCall(PetscOptionsFList("-regressor_type", "PetscRegressor type", "PetscRegressorSetType", PetscRegressorList, default_type, type, 256, &flg)); 190 if (flg) { 191 PetscCall(PetscRegressorSetType(regressor, type)); 192 } else if (!((PetscObject)regressor)->type_name) { 193 PetscCall(PetscRegressorSetType(regressor, default_type)); 194 } 195 PetscCall(PetscOptionsReal("-regressor_regularizer_weight", "Weight for the regularizer", "PetscRegressorSetRegularizerWeight", regressor->regularizer_weight, ®ressor->regularizer_weight, &flg)); 196 if (flg) PetscCall(PetscRegressorSetRegularizerWeight(regressor, regressor->regularizer_weight)); 197 // The above is a little superfluous, because we have already set regressor->regularizer_weight above, but we also need to set the flag indicating that the user has set the weight! 198 PetscTryTypeMethod(regressor, setfromoptions, PetscOptionsObject); 199 PetscOptionsEnd(); 200 PetscFunctionReturn(PETSC_SUCCESS); 201 } 202 203 /*@ 204 PetscRegressorSetUp - Sets up the internal data structures for the later use of a regressor. 205 206 Collective 207 208 Input Parameter: 209 . regressor - the `PetscRegressor` context 210 211 Notes: 212 For basic use of the `PetscRegressor` solvers the user need not to explicitly call 213 `PetscRegressorSetUp()`, since these actions will automatically occur during 214 the call to `PetscRegressorFit()`. However, if one wishes to control this 215 phase separately, `PetscRegressorSetUp()` should be called after `PetscRegressorCreate()`, 216 `PetscRegressorSetUp()`, and optional routines of the form `PetscRegressorSetXXX()`, 217 but before `PetscRegressorFit()`. 218 219 Level: advanced 220 221 .seealso: `PetscRegressorCreate()`, `PetscRegressorFit()`, `PetscRegressorDestroy()` 222 @*/ 223 PetscErrorCode PetscRegressorSetUp(PetscRegressor regressor) 224 { 225 PetscFunctionBegin; 226 PetscValidHeaderSpecific(regressor, PETSCREGRESSOR_CLASSID, 1); 227 if (regressor->setupcalled) PetscFunctionReturn(PETSC_SUCCESS); 228 PetscCall(PetscLogEventBegin(PetscRegressor_SetUp, regressor, 0, 0, 0)); 229 //TODO is there some mat vec etc that must be set, like TaoSolution? 230 PetscTryTypeMethod(regressor, setup); 231 regressor->setupcalled = PETSC_TRUE; 232 PetscCall(PetscLogEventEnd(PetscRegressor_SetUp, regressor, 0, 0, 0)); 233 PetscFunctionReturn(PETSC_SUCCESS); 234 } 235 236 /* NOTE: I've decided to make this take X and y, like the Scikit-learn Fit routines do. 237 * Am I overlooking some reason that X should be set in a separate function call, a la KSPSetOperators()?. */ 238 /*@ 239 PetscRegressorFit - Fit, or train, a regressor from a training dataset 240 241 Collective 242 243 Input Parameters: 244 + regressor - the `PetscRegressor` context 245 . X - matrix of training data (of dimension [number of samples] x [number of features]) 246 - y - vector of target values from the training dataset 247 248 Level: beginner 249 250 .seealso: `PetscRegressorCreate()`, `PetscRegressorSetUp()`, `PetscRegressorDestroy()`, `PetscRegressorPredict()` 251 @*/ 252 PetscErrorCode PetscRegressorFit(PetscRegressor regressor, Mat X, Vec y) 253 { 254 PetscFunctionBegin; 255 PetscValidHeaderSpecific(regressor, PETSCREGRESSOR_CLASSID, 1); 256 if (X) PetscValidHeaderSpecific(X, MAT_CLASSID, 2); 257 if (y) PetscValidHeaderSpecific(y, VEC_CLASSID, 3); 258 259 if (X) { 260 PetscCall(PetscObjectReference((PetscObject)X)); 261 PetscCall(MatDestroy(®ressor->training)); 262 regressor->training = X; 263 } 264 if (y) { 265 PetscCall(PetscObjectReference((PetscObject)y)); 266 PetscCall(VecDestroy(®ressor->target)); 267 regressor->target = y; 268 } 269 PetscCall(PetscRegressorSetUp(regressor)); 270 271 PetscCall(PetscLogEventBegin(PetscRegressor_Fit, regressor, X, y, 0)); 272 PetscUseTypeMethod(regressor, fit); 273 PetscCall(PetscLogEventEnd(PetscRegressor_Fit, regressor, X, y, 0)); 274 //TODO print convergence data 275 PetscCall(PetscRegressorViewFromOptions(regressor, NULL, "-regressor_view")); 276 regressor->fitcalled = PETSC_TRUE; 277 PetscFunctionReturn(PETSC_SUCCESS); 278 } 279 280 /*@ 281 PetscRegressorPredict - Compute predictions (that is, perform inference) using a fitted regression model. 282 283 Collective 284 285 Input Parameters: 286 + regressor - the `PetscRegressor` context (for which `PetscRegressorFit()` must have been called) 287 - X - data matrix of unlabeled observations 288 289 Output Parameter: 290 . y - vector of predicted labels 291 292 Level: beginner 293 294 .seealso: `PetscRegressorFit()`, `PetscRegressorDestroy()` 295 @*/ 296 PetscErrorCode PetscRegressorPredict(PetscRegressor regressor, Mat X, Vec y) 297 { 298 PetscFunctionBegin; 299 PetscValidHeaderSpecific(regressor, PETSCREGRESSOR_CLASSID, 1); 300 if (X) PetscValidHeaderSpecific(X, MAT_CLASSID, 2); 301 if (y) PetscValidHeaderSpecific(y, VEC_CLASSID, 3); 302 PetscCheck(regressor->fitcalled == PETSC_TRUE, ((PetscObject)regressor)->comm, PETSC_ERR_ARG_WRONGSTATE, "PetscRegressorFit() must be called before PetscRegressorPredict()"); 303 PetscCall(PetscLogEventBegin(PetscRegressor_Predict, regressor, X, y, 0)); 304 PetscTryTypeMethod(regressor, predict, X, y); 305 PetscCall(PetscLogEventEnd(PetscRegressor_Predict, regressor, X, y, 0)); 306 PetscFunctionReturn(PETSC_SUCCESS); 307 } 308 309 /*@ 310 PetscRegressorReset - Resets a `PetscRegressor` context by removing any allocated `Vec` and `Mat`. Any options set in the object remain. 311 312 Collective 313 314 Input Parameter: 315 . regressor - context obtained from `PetscRegressorCreate()` 316 317 Level: intermediate 318 319 .seealso: `PetscRegressorCreate()`, `PetscRegressorSetUp()`, `PetscRegressorFit()`, `PetscRegressorPredict()`, `PetscRegressorDestroy()` 320 @*/ 321 PetscErrorCode PetscRegressorReset(PetscRegressor regressor) 322 { 323 PetscFunctionBegin; 324 PetscValidHeaderSpecific(regressor, PETSCREGRESSOR_CLASSID, 1); 325 if (regressor->ops->reset) PetscTryTypeMethod(regressor, reset); 326 PetscCall(MatDestroy(®ressor->training)); 327 PetscCall(VecDestroy(®ressor->target)); 328 PetscCall(TaoDestroy(®ressor->tao)); 329 regressor->setupcalled = PETSC_FALSE; 330 regressor->fitcalled = PETSC_FALSE; 331 PetscFunctionReturn(PETSC_SUCCESS); 332 } 333 334 /*@C 335 PetscRegressorDestroy - Destroys the regressor context that was created with `PetscRegressorCreate()`. 336 337 Collective 338 339 Input Parameter: 340 . regressor - the `PetscRegressor` context 341 342 Level: beginner 343 344 .seealso: `PetscRegressorCreate()`, `PetscRegressorSetUp()`, `PetscRegressorReset()`, `PetscRegressor` 345 @*/ 346 PetscErrorCode PetscRegressorDestroy(PetscRegressor *regressor) 347 { 348 PetscFunctionBegin; 349 if (!*regressor) PetscFunctionReturn(PETSC_SUCCESS); 350 PetscValidHeaderSpecific((*regressor), PETSCREGRESSOR_CLASSID, 1); 351 if (--((PetscObject)(*regressor))->refct > 0) { 352 *regressor = NULL; 353 PetscFunctionReturn(PETSC_SUCCESS); 354 } 355 356 PetscCall(PetscRegressorReset(*regressor)); 357 PetscTryTypeMethod(*regressor, destroy); 358 359 PetscCall(PetscHeaderDestroy(regressor)); 360 PetscFunctionReturn(PETSC_SUCCESS); 361 } 362 363 /*@C 364 PetscRegressorSetType - Sets the type for the regressor. 365 366 Collective 367 368 Input Parameters: 369 + regressor - the `PetscRegressor` context 370 - type - a known regression method 371 372 Options Database Key: 373 . -regressor_type <type> - Sets the type of regressor; use -help for a list of available types 374 375 Level: intermediate 376 377 Notes: 378 See "include/petscregressor.h" for available methods (for instance) 379 . `PETSCREGRESSORLINEAR` - Regression model that is linear in its coefficients; supports ordinary least squares as well as regularized variants 380 381 Normally, it is best to use the `PetscRegressorSetFromOptions()` command and then 382 set the `PetscRegressor` type from the options database rather than by using 383 this routine, as this provides maximum flexibility. 384 The `PetscRegressorSetType()` routine is provided for those situations where it 385 is necessary to set the nonlinear solver independently of the command 386 line or options database. 387 388 .seealso: `PetscRegressorType` 389 @*/ 390 PetscErrorCode PetscRegressorSetType(PetscRegressor regressor, PetscRegressorType type) 391 { 392 PetscErrorCode (*r)(PetscRegressor); 393 PetscBool match; 394 395 PetscFunctionBegin; 396 PetscValidHeaderSpecific(regressor, PETSCREGRESSOR_CLASSID, 1); 397 PetscAssertPointer(type, 2); 398 399 PetscCall(PetscObjectTypeCompare((PetscObject)regressor, type, &match)); 400 if (match) PetscFunctionReturn(PETSC_SUCCESS); 401 402 PetscCall(PetscFunctionListFind(PetscRegressorList, type, &r)); 403 PetscCheck(r, PetscObjectComm((PetscObject)regressor), PETSC_ERR_ARG_UNKNOWN_TYPE, "Unable to find requested PetscRegressor type %s", type); 404 405 /* Destroy the existing solver information */ 406 PetscTryTypeMethod(regressor, destroy); 407 PetscCall(TaoDestroy(®ressor->tao)); 408 regressor->ops->setup = NULL; 409 regressor->ops->setfromoptions = NULL; 410 regressor->ops->settraining = NULL; 411 regressor->ops->fit = NULL; 412 regressor->ops->predict = NULL; 413 regressor->ops->destroy = NULL; 414 regressor->ops->reset = NULL; 415 regressor->ops->view = NULL; 416 417 /* Call the PetscRegressorCreate_XXX routine for this particular regressor */ 418 regressor->setupcalled = PETSC_FALSE; 419 PetscCall((*r)(regressor)); 420 PetscCall(PetscObjectChangeTypeName((PetscObject)regressor, type)); 421 PetscFunctionReturn(PETSC_SUCCESS); 422 } 423 424 /*@ 425 PetscRegressorGetType - Gets the current `PetscRegressorType` being used in the `PetscRegressor` object 426 427 Not Collective 428 429 Input Parameter: 430 . regressor - the `PetscRegressor` solver context 431 432 Output Parameter: 433 . type - the `PetscRegressorType` 434 435 Level: intermediate 436 437 .seealso: [](ch_regressor), `PetscRegressor`, `PetscRegressorType`, `PetscRegressorSetType()` 438 @*/ 439 PetscErrorCode PetscRegressorGetType(PetscRegressor regressor, PetscRegressorType *type) 440 { 441 PetscFunctionBegin; 442 PetscValidHeaderSpecific(regressor, PETSCREGRESSOR_CLASSID, 1); 443 PetscAssertPointer(type, 2); 444 *type = ((PetscObject)regressor)->type_name; 445 PetscFunctionReturn(PETSC_SUCCESS); 446 } 447 448 /*@ 449 PetscRegressorSetRegularizerWeight - Sets the weight to be used for the regularizer for a `PetscRegressor` context 450 451 Logically Collective 452 453 Input Parameters: 454 + regressor - the `PetscRegressor` context 455 - weight - the regularizer weight 456 457 Options Database Key: 458 . regressor_regularizer_weight <weight> - sets the regularizer's weight 459 460 Level: beginner 461 462 .seealso: `PetscRegressorSetType` 463 @*/ 464 PetscErrorCode PetscRegressorSetRegularizerWeight(PetscRegressor regressor, PetscReal weight) 465 { 466 PetscFunctionBegin; 467 PetscValidHeaderSpecific(regressor, PETSCREGRESSOR_CLASSID, 1); 468 PetscValidLogicalCollectiveReal(regressor, weight, 2); 469 regressor->regularizer_weight = weight; 470 PetscFunctionReturn(PETSC_SUCCESS); 471 } 472 473 /*@ 474 PetscRegressorGetTao - Returns the `Tao` context for a `PetscRegressor` object. 475 476 Not Collective, but if the `PetscRegressor` is parallel, then the `Tao` object is parallel 477 478 Input Parameter: 479 . regressor - the regressor context 480 481 Output Parameter: 482 . tao - the `Tao` context 483 484 Level: beginner 485 486 Notes: 487 The `Tao` object will be created if it does not yet exist. 488 489 The user can directly manipulate the `Tao` context to set various 490 options, etc. Likewise, the user can then extract and manipulate the 491 child contexts such as `KSP` or `TaoLineSearch`as well. 492 493 Depending on the type of the regressor and the options that are set, the regressor may use not use a `Tao` object. 494 495 .seealso: `PetscRegressorLinearGetKSP()` 496 @*/ 497 PetscErrorCode PetscRegressorGetTao(PetscRegressor regressor, Tao *tao) 498 { 499 PetscFunctionBegin; 500 PetscValidHeaderSpecific(regressor, PETSCREGRESSOR_CLASSID, 1); 501 PetscAssertPointer(tao, 2); 502 // Analogous to how SNESGetKSP() operates, this routine should create the Tao if it doesn't exist. 503 if (!regressor->tao) { 504 PetscCall(TaoCreate(PetscObjectComm((PetscObject)regressor), ®ressor->tao)); 505 PetscCall(PetscObjectIncrementTabLevel((PetscObject)regressor->tao, (PetscObject)regressor, 1)); 506 PetscCall(PetscObjectSetOptions((PetscObject)regressor->tao, ((PetscObject)regressor)->options)); 507 } 508 *tao = regressor->tao; 509 PetscFunctionReturn(PETSC_SUCCESS); 510 } 511 512 /*@ 513 PetscRegressorSetOptionsPrefix - Sets the prefix used for searching for all 514 PetscRegressor options in the database. 515 516 Logically Collective 517 518 Input Parameters: 519 + regressor - the `PetscRegressor` context 520 - p - the prefix string to prepend to all PetscRegressor option requests 521 522 Level: advanced 523 524 Notes: 525 A hyphen (-) must NOT be given at the beginning of the prefix name. 526 The first character of all runtime options is AUTOMATICALLY the hyphen. 527 528 For example, to distinguish between the runtime options for two 529 different PetscRegressor solvers, one could call 530 .vb 531 PetscRegressorSetOptionsPrefix(regressor1,"sys1_") 532 PetscRegressorSetOptionsPrefix(regressor2,"sys2_") 533 .ve 534 535 This would enable use of different options for each system, such as 536 .vb 537 -sys1_regressor_method linear -sys1_regressor_regularizer_weight 1.2 538 -sys2_regressor_method linear -sys2_regressor_regularizer_weight 1.1 539 .ve 540 541 .seealso: [](ch_regressor), `PetscRegressor`, `PetscRegressorSetFromOptions()`, `PetscRegressorAppendOptionsPrefix()`, `PetscRegressorGetOptionsPrefix()` 542 @*/ 543 PetscErrorCode PetscRegressorSetOptionsPrefix(PetscRegressor regressor, const char p[]) 544 { 545 PetscFunctionBegin; 546 PetscValidHeaderSpecific(regressor, PETSCREGRESSOR_CLASSID, 1); 547 PetscCall(PetscObjectSetOptionsPrefix((PetscObject)regressor, p)); 548 PetscFunctionReturn(PETSC_SUCCESS); 549 } 550 551 /*@ 552 PetscRegressorAppendOptionsPrefix - Appends to the prefix used for searching for all PetscRegressor options in the database. 553 554 Logically Collective 555 556 Input Parameters: 557 + regressor - the `PetscRegressor` solver context 558 - p - the prefix string to prepend to all `PetscRegressor` option requests 559 560 Level: advanced 561 562 Note: 563 A hyphen (-) must NOT be given at the beginning of the prefix name. 564 The first character of all runtime options is automatically the hyphen. 565 566 .seealso: [](ch_regressor), `PetscRegressor`, `PetscRegressorSetFromOptions()`, `PetscRegressorSetOptionsPrefix()`, `PetscRegressorGetOptionsPrefix()` 567 @*/ 568 PetscErrorCode PetscRegressorAppendOptionsPrefix(PetscRegressor regressor, const char p[]) 569 { 570 PetscFunctionBegin; 571 PetscValidHeaderSpecific(regressor, PETSCREGRESSOR_CLASSID, 1); 572 PetscCall(PetscObjectAppendOptionsPrefix((PetscObject)regressor, p)); 573 PetscFunctionReturn(PETSC_SUCCESS); 574 } 575 576 /*@ 577 PetscRegressorGetOptionsPrefix - Gets the prefix used for searching for all 578 PetscRegressor options in the database 579 580 Not Collective 581 582 Input Parameter: 583 . regressor - the `PetscRegressor` context 584 585 Output Parameter: 586 . p - pointer to the prefix string used is returned 587 588 Fortran Notes: 589 Pass in a string 'prefix' of sufficient length to hold the prefix. 590 591 Level: advanced 592 593 .seealso: [](ch_regressor), `PetscRegressor`, `PetscRegressorSetFromOptions()`, `PetscRegressorSetOptionsPrefix()`, `PetscRegressorAppendOptionsPrefix()` 594 @*/ 595 PetscErrorCode PetscRegressorGetOptionsPrefix(PetscRegressor regressor, const char *p[]) 596 { 597 PetscFunctionBegin; 598 PetscValidHeaderSpecific(regressor, PETSCREGRESSOR_CLASSID, 1); 599 PetscCall(PetscObjectGetOptionsPrefix((PetscObject)regressor, p)); 600 PetscFunctionReturn(PETSC_SUCCESS); 601 } 602