Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Support for Aggregate CAS Multiplier #25

Open
wants to merge 23 commits into
base: main
Choose a base branch
from
Open
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
23 commits
Select commit Hold shift + click to select a range
1232506
framework: initial implementation of FlowController and related classes
mac-op Jul 19, 2024
50ddfb0
framework: allow CAS to keep track of owner and release itself
mac-op Jul 25, 2024
5d4f657
test: fixed implementation name in DaveDetector
mac-op Jul 25, 2024
280b444
framework: allow aggregate CAS Multiplier
mac-op Jul 25, 2024
6d30b20
Merge branch 'apache:main' into feature/agg-cas-multiplier
mac-op Jul 25, 2024
8aa5919
framework: fixed small hack in AnnotatorManager
mac-op Jul 25, 2024
d579898
framework: added function and class documentation
mac-op Jul 25, 2024
0d635ca
framework: added documentation and comments
mac-op Jul 26, 2024
0adb05c
framework: fixed memory leak in AnnotatorManager
mac-op Jul 26, 2024
3d4086f
bug: fixed undefined behavior in Step union
mac-op Aug 6, 2024
5f7ae16
bug: fixed another undefined behavior in Step union
mac-op Aug 7, 2024
f5d045b
framework: introduce owner variable for CASes in Pool
mac-op Aug 13, 2024
e3f1b33
clean up for CAS and CASPool
mac-op Aug 18, 2024
2034217
framework: Added check for invalid call to CAS::release
mac-op Aug 20, 2024
d49a58e
framework: formatting and documentation
mac-op Aug 20, 2024
88ba14a
framework: correctly setting the CAS Owner in CAS Pool
mac-op Aug 20, 2024
f89eb88
fix: small bug from last commit in releaseCAS
mac-op Aug 20, 2024
5966b78
test: fix incorrect test case for Aggregate CAS combiner
mac-op Aug 20, 2024
df7bf0c
test: added test case for Step class
mac-op Aug 22, 2024
ce9cd6c
fix: better exception message in annotator_mgr.cpp
mac-op Aug 22, 2024
7cd9ab9
fix: added test case in test_engine
mac-op Aug 22, 2024
e07c09f
fix: remove CAS from activeCASes in AnnotatorManager
mac-op Aug 22, 2024
d4588ba
framework: removed old constructor for CASPool
mac-op Aug 23, 2024
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 22 additions & 0 deletions src/cas/cas.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -179,6 +179,7 @@ namespace uima {
iv_indexRepository(NULL),
iv_filterBuilder(NULL),
iv_componentInfo(NULL),
iv_owner(NULL),
iv_utDocumentType(uima::lowlevel::TypeSystem::INVALID_TYPE),
iv_utDocumentLangAsIntFeat(uima::lowlevel::TypeSystem::INVALID_FEATURE),
iv_utDocumentLangAsStrFeat(uima::lowlevel::TypeSystem::INVALID_FEATURE),
Expand Down Expand Up @@ -221,6 +222,7 @@ namespace uima {
iv_typeSystem = inCas->iv_typeSystem;
iv_heap = inCas->iv_heap;
iv_componentInfo = inCas->iv_componentInfo;
iv_owner = inCas->iv_owner;
iv_utDocumentLangAsIntFeat = uima::lowlevel::TypeSystem::INVALID_FEATURE;
iv_utDocumentLangAsStrFeat = uima::lowlevel::TypeSystem::INVALID_FEATURE;
refreshCachedTypes();
Expand Down Expand Up @@ -262,6 +264,7 @@ namespace uima {
iv_sofaCount(0),
initialSofaCreated(false),
iv_initialView(NULL),
iv_owner(NULL),
iv_indexRepository(NULL),
iv_filterBuilder(NULL),
iv_componentInfo(NULL),
Expand Down Expand Up @@ -676,6 +679,10 @@ namespace uima {
);
}

void CAS::setOwner(AnnotatorContext *owner) {
iv_baseCas->iv_owner = owner;
}

// deprecated version
void CAS::setDocumentText(UChar const * cpDocument, size_t uiLength, bool bCopyToCAS ) {
if (cpDocument == NULL) {
Expand Down Expand Up @@ -809,6 +816,21 @@ namespace uima {
}
}

void CAS::release() {
if (iv_baseCas->iv_owner) {
iv_baseCas->iv_owner->releaseCAS(*this);
} else {
ErrorMessage msg(UIMA_MSG_ID_EXC_INVALID_CAS_RELEASE);
msg.addParam("This CAS does not have any owner");
UIMA_EXC_THROW_NEW(CASException,
UIMA_ERR_CAS_RELEASE,
msg,
ErrorMessage(UIMA_MSG_ID_EXCON_UNKNOWN_CONTEXT),
ErrorInfo::recoverable
);
}
}

ANIndex CAS::getAnnotationIndex(Type const & crType) {
if (isbaseCas) {
assertWithMsg(false, "Annotation Index does not exist in Base CAS!");
Expand Down
10 changes: 10 additions & 0 deletions src/cas/uima/cas.hpp
Original file line number Diff line number Diff line change
Expand Up @@ -191,6 +191,8 @@ namespace uima {
void bumpSofaCount();
void invalidBaseCasMethod();

/** Set the owner of the base CAS */
void setOwner(AnnotatorContext* owner);
void registerView(SofaFS);
void updateDocumentAnnotation( );
void copyDocumentString(UnicodeStringRef);
Expand Down Expand Up @@ -231,6 +233,7 @@ namespace uima {
bool initialSofaCreated;
bool isDeletingViews; //set this flag to true when destroying CAS
AnnotatorContext *iv_componentInfo;
AnnotatorContext *iv_owner;

uima::lowlevel::TyFSType iv_utDocumentType;
uima::lowlevel::TyFSFeature iv_utDocumentLangAsIntFeat;
Expand Down Expand Up @@ -772,6 +775,13 @@ namespace uima {
icu::UnicodeString getAnnotationIndexID() const {
return CAS::INDEXID_ANNOTATION;
}

/**
* When called this CAS will release itself by calling releaseCas on the AnnotatorContext that owns it.
* NOTE: This only works for CASes that have an owner, ie. belong to a CASPool.
*/
void release();

/** @} */

/** @defgroup PreDefTypes Predefined Types
Expand Down
4 changes: 3 additions & 1 deletion src/framework/Makefile.am
Original file line number Diff line number Diff line change
Expand Up @@ -78,7 +78,9 @@ libuima_la_SOURCES += consoleui.cpp
libuima_la_SOURCES += cp2ucnvrt.cpp
libuima_la_SOURCES += dottypesystemwriter.cpp
libuima_la_SOURCES += engine.cpp
libuima_la_SOURCES += exceptions.cpp
libuima_la_SOURCES += exceptions.cpp
libuima_la_SOURCES += flow_controller.cpp
libuima_la_SOURCES += flow.cpp
libuima_la_SOURCES += ftools.cpp
libuima_la_SOURCES += internal_aggregate_engine.cpp
libuima_la_SOURCES += internal_capability_container.cpp
Expand Down
8 changes: 6 additions & 2 deletions src/framework/annotator_context.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -165,7 +165,11 @@ namespace uima {
}

TyErrorId AnnotatorContext::defineCASPool(size_t numInstances) {
iv_pCasPool = new CASPool(getTaeSpecifier(),numInstances);
if (iv_pParentAnC) { // If this is a delegate
iv_pCasPool = new CASPool(this, iv_pParentAnC->getTaeSpecifier(),numInstances);
} else {
iv_pCasPool = new CASPool(this, getTaeSpecifier(), numInstances);
}
if (iv_pCasPool == NULL) {
return UIMA_ERR_USER_ANNOTATOR_OUT_OF_MEMORY;
}
Expand Down Expand Up @@ -500,7 +504,7 @@ namespace uima {
NameValuePair const * AnnotatorContext::findNameValuePair(const icu::UnicodeString & paramName,
const icu::UnicodeString & ancKey) const {
/* return findNameValuePair(getGroupNameWhenNotSpec(), paramName, iv_pTaeSpecifier->getSearchStrategy()); */
NameValuePair const * pValueLocal = pValueLocal = iv_pTaeSpecifier->getNameValuePair(paramName, ancKey);
NameValuePair const * pValueLocal = iv_pTaeSpecifier->getNameValuePair(paramName, ancKey);

// the request was invalid we got an exception we leave to others to catch

Expand Down
234 changes: 203 additions & 31 deletions src/framework/annotator_mgr.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -103,6 +103,8 @@ namespace uima {
launchDeInit();
}
assert( iv_vecEntries.empty() );
if (iv_pFlowController)
delete iv_pFlowController;

}

Expand Down Expand Up @@ -152,12 +154,21 @@ namespace uima {
assert(iv_vecEntries.empty());

AnnotatorContext & rANC = iv_pEngine->getAnnotatorContext();

AnalysisEngineDescription const & crTAESpecifier = rANC.getTaeSpecifier(); // this method must be added

const AnalysisEngineMetaData* pEngineMetadata = crTAESpecifier.getAnalysisEngineMetaData();
assert( ! crTAESpecifier.isPrimitive() );
//BSIvector < icu::UnicodeString > const & crVecEngineNames = crTAESpecifier.getAnalysisEngineMetaData()->getFixedFlow()->getNodes();
vector < icu::UnicodeString > const & crVecEngineNames = crTAESpecifier.getAnalysisEngineMetaData()->getFlowConstraints()->getNodes();

// FIXME: This shouldn't have been necessary since FlowContrainst::getFlowContraintsType
DrDub marked this conversation as resolved.
Show resolved Hide resolved
// should have been const in the first place
auto flowContraints = CONST_CAST(FlowConstraints *, pEngineMetadata->getFlowConstraints());
if (flowContraints->getFlowConstraintsType() == FlowConstraints::FIXED) {
iv_pFlowController = new FixedFlowController;
iv_pFlowController->initialize(rANC);
}

if (const OperationalProperties* operationalProps = pEngineMetadata->getOperationalProperties())
iv_bOutputNewCases = operationalProps->getOutputsNewCASes();
vector < icu::UnicodeString > const & crVecEngineNames = flowContraints->getNodes();

// for all engines in the flow
size_t ui;
Expand Down Expand Up @@ -431,11 +442,7 @@ namespace uima {
return bHasTOF;
}





TyErrorId AnnotatorManager::launchProcessDocument(CAS & cas, ResultSpecification const & crResultSpec) {
TyErrorId AnnotatorManager::processCapabilityLanguageFlow(CAS &cas, ResultSpecification const &crResultSpec) {
/*
This works as follows:
The passes result spec is copied and for each delegate AE, it is determined
Expand Down Expand Up @@ -486,27 +493,13 @@ namespace uima {
bool requiresTCas=true;

if (cas.isBackwardCompatibleCas()) {
tcas = &cas;
}
//this populates the tofsToBeRemoved vector so always call it
callEngine = shouldEngineBeCalled(*pCapContainer,
resSpec,
cas.getDocumentAnnotation().getLanguage(),
tofsToBeRemoved);
//check the FlowConstraintType specified in the aggregate engine
//if CapabilityLanguageFlow whether engine is called is
//determined by shouldEngineBeCalled()
AnnotatorContext & rANC = iv_pEngine->getAnnotatorContext();
AnalysisEngineDescription const & crTAESpecifier = rANC.getTaeSpecifier();
FlowConstraints const * pFlow = crTAESpecifier.getAnalysisEngineMetaData()->getFlowConstraints();
FlowConstraints * flow = CONST_CAST(FlowConstraints *, pFlow);
FlowConstraints::EnFlowType flowType = flow->getFlowConstraintsType();

//if FixedFlow specified all engines are always called so reset callEngine is true
if (flowType == FlowConstraints::FIXED) {
callEngine=true;
}

tcas = &cas;
}
//this populates the tofsToBeRemoved vector so always call it
callEngine = shouldEngineBeCalled(*pCapContainer,
resSpec,
cas.getDocumentAnnotation().getLanguage(),
tofsToBeRemoved);

if ( callEngine ) {

Expand Down Expand Up @@ -583,7 +576,186 @@ namespace uima {
}

UIMA_ANNOTATOR_TIMING(iv_clTimerLaunchProcess.stop());
return(utRetVal);
return utRetVal;
}

CAS *AnnotatorManager::processUntilNextOutputCas() {
unique_ptr<Flow> flow{};
while (true) {
CAS *currentCas = nullptr;
Step nextStep;
flow = nullptr;

// get a cas from the stack
if (casIterStack.empty())
return nullptr;

StackFrame &frame = casIterStack.top();
try {
if (frame.casMultiplier && frame.casMultiplier->hasNext()) {
currentCas = &frame.casMultiplier->next();
// compute flow for newly produced CAS
flow = frame.originalFlow->newCasProduced(*currentCas, frame.lastEngineKey);
}
} catch (Exception &exception) {
if (!frame.originalFlow->continueOnFailure(frame.lastEngineKey /* ,exception */))
throw;
}

if (!currentCas) {
// if there is no more output CASes from the stack, take the original CAS that was processed by
// the CAS Multiplier and continue with its flow
currentCas = frame.originalCas;
flow = std::move(frame.originalFlow);
currentCas->setCurrentComponentInfo(nullptr);
casIterStack.pop();
}

activeCASes.insert(currentCas);
mac-op marked this conversation as resolved.
Show resolved Hide resolved

if (nextStep.getType() == Step::StepType::UNSPECIFIED) {
nextStep = flow->next(); // get the next step for the current flow
}

while (nextStep.getType() != Step::StepType::FINALSTEP) {
if (nextStep.getType() == Step::StepType::SIMPLESTEP) {
// find the AE specified by the step
const icu::UnicodeString &nextAEKey = nextStep.getSimpleStep()->getEngineName();
auto it = std::find_if(iv_vecEntries.begin(), iv_vecEntries.end(),
[&, nextAEKey](const EngineEntry &entry) {
return entry.iv_pEngine->getAnnotatorContext().iv_AnCKey == nextAEKey;
});

if (it != iv_vecEntries.end()) {
AnalysisEngine *nextAE = it->iv_pEngine;
CAS *outputCas = nullptr;

// call process one the AE and see if it has produced a new CAS
try {
CASIterator casIter = nextAE->processAndOutputNewCASes(*currentCas);
if (casIter.hasNext())
outputCas = &casIter.next();
} catch (Exception &e) {
if (!flow->continueOnFailure(nextAEKey))
throw;
}

if (outputCas) {
// new CAS is output so put the current components on the stack so we can process
// the other output CASes and original CASes later
std::unique_ptr<Flow> nextFlow = flow->newCasProduced(*outputCas, nextAEKey);
casIterStack.push({nextAE, currentCas, std::move(flow), nextAEKey});
flow = std::move(nextFlow);
currentCas = outputCas;
activeCASes.insert(currentCas);
} else {
// No new CASes are output, this CAS is done being processed by the current engine.
currentCas->setCurrentComponentInfo(nullptr);
}
} else {
UIMA_EXC_THROW_NEW(EngineProcessingException,
UIMA_ERR_USER_ANNOTATOR_COULD_NOT_PROCESS,
UIMA_MSG_ID_EXCON_PROCESSING_CAS,
ErrorMessage(UIMA_MSG_ID_LITERAL_STRING, "Unknown Delegate Key " + nextAEKey),
ErrorInfo::unrecoverable);
}
} else if (nextStep.getType() == Step::StepType::PARALLELSTEP) {
// TODO: ParallelStep not supported yet
UIMA_EXC_THROW_NEW(NotYetImplementedException,
UIMA_ERR_NOT_YET_IMPLEMENTED,
UIMA_MSG_ID_EXC_NOT_YET_IMPLEMENTED,
ErrorMessage(UIMA_MSG_ID_LITERAL_STRING, "Parallel Step not supported yet"),
ErrorInfo::unrecoverable
);
} else {
UIMA_EXC_THROW_NEW(EngineProcessingException,
UIMA_ERR_USER_ANNOTATOR_COULD_NOT_PROCESS,
UIMA_MSG_ID_EXCON_PROCESSING_CAS,
ErrorMessage(UIMA_MSG_ID_LITERAL_STRING, "Unknown Step Type"),
ErrorInfo::unrecoverable);
}

nextStep = flow->next();
}

const FinalStep *finalStep = nextStep.getFinalStep();
activeCASes.erase(currentCas);
if (currentCas == inputCas) {
if (finalStep->getForceDropCAS()) {
// Not allowed to drop the input CAS so something must have gone wrong
UIMA_EXC_THROW_NEW(EngineProcessingException,
UIMA_ERR_USER_ANNOTATOR_COULD_NOT_PROCESS,
UIMA_MSG_ID_EXCON_PROCESSING_CAS,
ErrorMessage(UIMA_MSG_ID_LITERAL_STRING, "Illegal CAS drop"),
ErrorInfo::unrecoverable);
}
return nullptr;
}

if (iv_bOutputNewCases && !finalStep->getForceDropCAS())
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just to double check I'm understanding the code correctly, this is where the DROP_IF_NEW_CAS_PRODUCED functionality is implemented?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes the finalStep will determine if that CAS is dropped or not. In this instance that finalStep is constructed by the FixedFlowController and FixedFlowObject, and the default action in the FixedFlowController is DROP_IF_NEW_CAS_PRODUCED

return currentCas;
currentCas->release();
}
}

bool AnnotatorManager::hasNext() {
if (!nextCas)
nextCas = processUntilNextOutputCas();
return nextCas != nullptr;
}

CAS & AnnotatorManager::next() {
CAS* result = nextCas;
if (!result)
result = processUntilNextOutputCas();
if (!result) {
UIMA_EXC_THROW_NEW(Exception,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

shouldn't this call also release()?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No when it is returned to the caller/user they will call release themselves. Or do you mean something else?

UIMA_ERR_USER_ANNOTATOR_COULD_NOT_PROCESS,
UIMA_MSG_ID_EXCON_PROCESSING_CAS,
ErrorMessage(UIMA_MSG_ID_LITERAL_STRING, "There is not next() available."),
ErrorInfo::unrecoverable);
}
nextCas = nullptr;
return *result;
}

void AnnotatorManager::release() {
while (!casIterStack.empty()) {
StackFrame& frame = casIterStack.top();
frame.originalFlow->aborted();
casIterStack.pop();
}
for (CAS *cas : activeCASes) {
if (cas != inputCas)
cas->release();
}

activeCASes.clear();
}


TyErrorId AnnotatorManager::launchProcessDocument(CAS &cas, ResultSpecification const &crResultSpec) {
//if engine uses Capability Language Flow
// TODO: Turn this logic and processCapabilityLanguageFlow into a separate CapabilityLanguageFlowController class that inherits from FlowController
AnnotatorContext &rANC = iv_pEngine->getAnnotatorContext();
AnalysisEngineDescription const &crTAESpecifier = rANC.getTaeSpecifier();
FlowConstraints const *pFlow = crTAESpecifier.getAnalysisEngineMetaData()->getFlowConstraints();
FlowConstraints *flow = CONST_CAST(FlowConstraints *, pFlow);

// Process according to capability language specifications
if (flow->getFlowConstraintsType() == FlowConstraints::CAPABILITYLANGUAGE)
return processCapabilityLanguageFlow(cas, crResultSpec);

inputCas = &cas;

casIterStack.push({nullptr, inputCas, iv_pFlowController->computeFlow(*inputCas), {}});
try {
nextCas = processUntilNextOutputCas();
} catch (...) {
release();
throw;
}
return UIMA_ERR_NONE;
}

#ifdef UIMA_DEBUG_ANNOTATOR_TIMING
Expand Down
Loading