Method doExec has a Cognitive Complexity of 172 (exceeds 5 allowed). Consider refactoring.

Open

    public ExecutionResult doExec(DifferentialFunction op,
                                  OpContext opContext,
                                  FrameIter outputFrameIter,
                                  Set<VarId> opInputs, Set<VarId> allIterInputs,
                                  Set<String> constAndPhInputs,

• Read up
  Read up

Method getOutputsHelperTensorArrayOps has a Cognitive Complexity of 158 (exceeds 5 allowed). Consider refactoring.

Open

    public ExecutionResult getOutputsHelperTensorArrayOps(DifferentialFunction op, FrameIter outputFrameIter, Set<VarId> opInputs, Set<VarId> allIterInputs, Map<String, SDValue> otherPlaceHolders) {
        /*
        TODO: TensorArray memory management note: For now, we'll close any INDArrays stored in the TensorArray at the end of
        graph execution. This uses more memory than necessary for an earlier close strategy, but simplifies memory management.
        This should be revisited and optimized later

• Read up
  Read up

Method getOutputs has a Cognitive Complexity of 152 (exceeds 5 allowed). Consider refactoring.

Open

    @Override
    public ExecutionResult getOutputs(Pair<SameDiffOp, OpContext> opPair,
                                      FrameIter outputFrameIter,
                                      Set<VarId> opInputs,
                                      Set<VarId> allIterInputs,

• Read up
  Read up

File InferenceSession.java has 1172 lines of code (exceeds 250 allowed). Consider refactoring.

Open

/*
 *  ******************************************************************************
 *  *
 *  *
 *  * This program and the accompanying materials are made available under the

Method getAndParameterizeOp has a Cognitive Complexity of 105 (exceeds 5 allowed). Consider refactoring.

Open

    @Override
    public Pair<SameDiffOp,OpContext> getAndParameterizeOp(String opName, FrameIter frameIter, Set<VarId> opInputs, Set<VarId> allIterInputs,
                                                           Set<String> constAndPhInputs, Map<String, INDArray> placeholderValues, Set<String> allReqVariables, Map<String, SDValue> otherPlaceholders) {
        SameDiffOp sdo = sameDiff.getOps().get(opName);
        DifferentialFunction df = sdo.getOp();

• Read up
  Read up

Method doExec has 299 lines of code (exceeds 25 allowed). Consider refactoring.

Open

    public ExecutionResult doExec(DifferentialFunction op,
                                  OpContext opContext,
                                  FrameIter outputFrameIter,
                                  Set<VarId> opInputs, Set<VarId> allIterInputs,
                                  Set<String> constAndPhInputs,

Method getOutputsHelperTensorArrayOps has 276 lines of code (exceeds 25 allowed). Consider refactoring.

Open

    public ExecutionResult getOutputsHelperTensorArrayOps(DifferentialFunction op, FrameIter outputFrameIter, Set<VarId> opInputs, Set<VarId> allIterInputs, Map<String, SDValue> otherPlaceHolders) {
        /*
        TODO: TensorArray memory management note: For now, we'll close any INDArrays stored in the TensorArray at the end of
        graph execution. This uses more memory than necessary for an earlier close strategy, but simplifies memory management.
        This should be revisited and optimized later

Method getAndParameterizeOp has 159 lines of code (exceeds 25 allowed). Consider refactoring.

Open

    @Override
    public Pair<SameDiffOp,OpContext> getAndParameterizeOp(String opName, FrameIter frameIter, Set<VarId> opInputs, Set<VarId> allIterInputs,
                                                           Set<String> constAndPhInputs, Map<String, INDArray> placeholderValues, Set<String> allReqVariables, Map<String, SDValue> otherPlaceholders) {
        SameDiffOp sdo = sameDiff.getOps().get(opName);
        DifferentialFunction df = sdo.getOp();

Method getOutputs has 157 lines of code (exceeds 25 allowed). Consider refactoring.

Open

    @Override
    public ExecutionResult getOutputs(Pair<SameDiffOp, OpContext> opPair,
                                      FrameIter outputFrameIter,
                                      Set<VarId> opInputs,
                                      Set<VarId> allIterInputs,

Method preprocessPlaceholders has a Cognitive Complexity of 37 (exceeds 5 allowed). Consider refactoring.

Open

    @Override
    protected Map<String, INDArray> preprocessPlaceholders(Map<String, INDArray> placeholders, At at) {
        arrayUseTracker.clear();

        //We'll also use this method as a "pre execution" hook-in, to mark variables as something we should never deallocate

• Read up
  Read up

Method postProcessOutputValues has a Cognitive Complexity of 27 (exceeds 5 allowed). Consider refactoring.

Open

    @Override
    protected Map<String, SDValue> postProcessOutputValues(Map<String, SDValue> output) {
        //For any queued (not yet processed) ops - mark them as satisfied, so we can deallocate any arrays
        // that are waiting on them
        if (dt.hasNewAllSatisfied()) {

• Read up
  Read up

Method preprocessPlaceholders has 58 lines of code (exceeds 25 allowed). Consider refactoring.

Open

    @Override
    protected Map<String, INDArray> preprocessPlaceholders(Map<String, INDArray> placeholders, At at) {
        arrayUseTracker.clear();

        //We'll also use this method as a "pre execution" hook-in, to mark variables as something we should never deallocate

Consider simplifying this complex logical expression.

Open

        if (df instanceof LoopCond || df instanceof Enter || df instanceof Exit || df instanceof NextIteration ||
                df instanceof Merge || df instanceof Switch || df instanceof BaseTensorOp || df instanceof Invoke) {
            //Control dependencies and tensor ops (like TensorArray, TensorArrayRead etc) don't need inputs set, execution is a special case
            return new Pair<>(sdo, null);
        }

Method getOutputs has 10 arguments (exceeds 4 allowed). Consider refactoring.

Open

    public ExecutionResult getOutputs(Pair<SameDiffOp, OpContext> opPair,
                                      FrameIter outputFrameIter,
                                      Set<VarId> opInputs,
                                      Set<VarId> allIterInputs,
                                      Set<String> constAndPhInputs,

Method postProcessOutputValues has 31 lines of code (exceeds 25 allowed). Consider refactoring.

Open

    @Override
    protected Map<String, SDValue> postProcessOutputValues(Map<String, SDValue> output) {
        //For any queued (not yet processed) ops - mark them as satisfied, so we can deallocate any arrays
        // that are waiting on them
        if (dt.hasNewAllSatisfied()) {

Method getAndParameterizeOp has 8 arguments (exceeds 4 allowed). Consider refactoring.

Open

    public Pair<SameDiffOp,OpContext> getAndParameterizeOp(String opName, FrameIter frameIter, Set<VarId> opInputs, Set<VarId> allIterInputs,
                                                           Set<String> constAndPhInputs, Map<String, INDArray> placeholderValues, Set<String> allReqVariables, Map<String, SDValue> otherPlaceholders) {

Method doExec has 7 arguments (exceeds 4 allowed). Consider refactoring.

Open

    public ExecutionResult doExec(DifferentialFunction op,
                                  OpContext opContext,
                                  FrameIter outputFrameIter,
                                  Set<VarId> opInputs, Set<VarId> allIterInputs,
                                  Set<String> constAndPhInputs,

Avoid deeply nested control flow statements.

Open

                                        if(getValue.getListValue().get(j) !=  null) {
                                            args[i] = getValue.getListValue().get(j);
                                            break;
                                        }

Avoid deeply nested control flow statements.

Open

                        if (e.isConstant()) {
                        /*
                        Constant enter case: Need to keep this array around for the entire duration of the frame, including
                        any nested frames, and all iterations.
                        Unfortunately, we don't know exactly when we're done with a frame for good

Avoid deeply nested control flow statements.

Open

                            if(arr != null && !freedArrays.contains(arr.getId()) && sameDiff.isEnableCache()) {
                                mmgr.release(arr);
                                freedArrays.add(arr.getId());
                            }

Avoid deeply nested control flow statements.

Open

                            if(!freedArrays.contains(value.getTensorValue().getId()) &&
                                    sameDiff.isEnableCache() && !containsOutput) {
                                mmgr.release(value.getTensorValue());
                                freedArrays.add(value.getTensorValue().getId());
                            }

Avoid deeply nested control flow statements.

Open

                            for(INDArray arr : value.getListValue())
                                if(arr != null && !freedArrays.contains(arr.getId()) && sameDiff.isEnableCache() && !containsOutput) {
                                    mmgr.release(arr);
                                    freedArrays.add(arr.getId());
                                }

Avoid deeply nested control flow statements.

Open

                        for (int i = 0; i < out.numResults(); i++)
                            namedOutsBuilder.put(op.outputsOfOp.get(i), out.resultAt(i));

Method getOutputsHelperTensorArrayOps has 5 arguments (exceeds 4 allowed). Consider refactoring.

Open

    public ExecutionResult getOutputsHelperTensorArrayOps(DifferentialFunction op, FrameIter outputFrameIter, Set<VarId> opInputs, Set<VarId> allIterInputs, Map<String, SDValue> otherPlaceHolders) {

Avoid too many return statements within this method.

Open

                return ExecutionResult.createFrom(Arrays.asList(vid.getVariable()),new INDArray[]{inArr});

Avoid too many return statements within this method.

Open

            return ExecutionResult.createFrom(v.getVariable(),out);

Avoid too many return statements within this method.

Open

            return ExecutionResult.createValue(op.outputVariablesNames()[0], sdValue);

Avoid too many return statements within this method.

Open

            return ExecutionResult.createFrom(tensorArray.getVar().name(),scalar);

Avoid too many return statements within this method.

Open

                return ExecutionResult.createValue(vid.getVariable(),nodeValueOutputs.get(vid));

Avoid too many return statements within this method.

Open

            return ExecutionResult.createValue(op.outputVariablesNames()[0], sdValue);

Avoid too many return statements within this method.

Open

                return ExecutionResult.createFrom(Arrays.asList(in.getVariable()),new INDArray[]{inArr});

Avoid too many return statements within this method.

Open

            return ExecutionResult.createFrom(tArr.getVariable(),out);

Avoid too many return statements within this method.

Open

            return ExecutionResult.createFrom(Arrays.asList("gradientbackwardsmarker"), new INDArray[]{out});

Avoid too many return statements within this method.

Open

                    return ExecutionResult.createValue(in.getVariable(),value);

Avoid too many return statements within this method.

Open

            return ExecutionResult.createValue(vid.getVariable(),orig);

Avoid too many return statements within this method.

Open

            return ExecutionResult.createFrom(Arrays.asList(n), new INDArray[]{out});

Avoid too many return statements within this method.

Open

                        return ExecutionResult.createFrom(c,opContext);

Avoid too many return statements within this method.

Open

            return ExecutionResult.createValue(vid.getVariable(), getValue);

Avoid too many return statements within this method.

Open

            return ExecutionResult.createFrom(op.outputVariablesNames()[0], from);

Avoid too many return statements within this method.

Open

                    return ExecutionResult.createValue(inputVarId.getVariable(),
                            value);

Avoid too many return statements within this method.

Open

            return getOutputsHelperTensorArrayOps(op, outputFrameIter, opInputs, allIterInputs, otherPlaceHolders);

Avoid too many return statements within this method.

Open

            return ExecutionResult.createValue(inputVarId.getVariable(), sdValue);

Avoid too many return statements within this method.

Open

                return ExecutionResult.createFrom(Arrays.asList(inputVarId.getVariable()),new INDArray[]{inArr});

Avoid too many return statements within this method.

Open

            return ExecutionResult.createValue(op.outputVariable().name(),sdValue1);

Avoid too many return statements within this method.

Open

                    return ExecutionResult.createFrom(Arrays.asList(in.getVariable()),new INDArray[]{inArr});

Avoid too many return statements within this method.

Open

                    return ExecutionResult.createFrom(Arrays.asList(inputVarId.getVariable()),new INDArray[]{inArr});

Avoid too many return statements within this method.

Open

            return executionResultBuilder.build();

Avoid too many return statements within this method.

Open

                return ExecutionResult.createFrom(tArr.getVariable(),out);

Avoid too many return statements within this method.

Open

                return ExecutionResult.createValue(vid.getVariable(), getSdValue(vid));

Avoid too many return statements within this method.

Open

            return ExecutionResult.createValue(vid.getVariable(),listValue);

Identical blocks of code found in 2 locations. Consider refactoring.

Open

            if(nodeValueOutputs.containsKey(in) && getSdValue(in) != null) {
                SDValue value = getSdValue(in);
                if(value != null && value.getSdValueType() == SDValueType.LIST) {
                    return ExecutionResult.createValue(in.getVariable(),value);
                } else if(value != null && value.getSdValueType() == SDValueType.TENSOR) {

• Read up
  Read up

Identical blocks of code found in 2 locations. Consider refactoring.

Open

            if(nodeValueOutputs.containsKey(inputVarId)) {
                SDValue value = getSdValue(inputVarId);
                if(value != null && value.getSdValueType() == SDValueType.LIST) {
                    return ExecutionResult.createValue(inputVarId.getVariable(),
                            value);

• Read up
  Read up

Identical blocks of code found in 2 locations. Consider refactoring.

Open

            if (constPhInput) {
                //Constant or placeholder
                inputVarId = new VarId(constAndPhInputs.iterator().next(), OUTER_FRAME, 0, null);
            } else if (allIterInputs != null && allIterInputs.size() > 0) {
                inputVarId = allIterInputs.iterator().next();

• Read up
  Read up

Identical blocks of code found in 2 locations. Consider refactoring.

Open

            if (constPhInput) {
                //Constant or placeholder
                inputVarId = new VarId(constAndPhInputs.iterator().next(), OUTER_FRAME, 0, null);
            } else if (allIterInputs != null && allIterInputs.size() > 0) {
                inputVarId = allIterInputs.iterator().next();

• Read up
  Read up

Identical blocks of code found in 2 locations. Consider refactoring.

Open

            while (sameDiff.getVariableOutputOp(inTensorArray.name()) instanceof Enter) {
                //Handle the Enter case: this is like TensorArray -> Enter -> TensorArrayWrite
                //TODO also TensorArrayScatter, etc??
                inTensorArray = sameDiff.getVariableOutputOp(inTensorArray.name()).arg();
                tArr = tArr.getParentFrame().toVarId(inTensorArray.name());

• Read up
  Read up

Identical blocks of code found in 2 locations. Consider refactoring.

Open

            while (sameDiff.getVariableOutputOp(inTensorArray.name()) instanceof Enter) {
                //Handle the Enter case: this is like TensorArray -> Enter -> TensorArrayWrite
                //TODO also TensorArrayScatter, etc??
                inTensorArray = sameDiff.getVariableOutputOp(inTensorArray.name()).arg();
                tArr = tArr.getParentFrame().toVarId(inTensorArray.name());

• Read up
  Read up

Similar blocks of code found in 2 locations. Consider refactoring.

Open

                log.trace("Scattering item at index " + i + " for list " + tArr + " with value " + get + " from whole list of " + l + " from values array " + valuesArr.toStringFull() + " named " + valuesSDV.name());

• Read up
  Read up