commented out code

This library can detect both inline or block commented out code.

It sends rpc to the server which uses LSTM model to predict commented out code. The server runs on my school data center, and will be migrated to a private machine.

Only support c/c++ now. More language support is about to come

How to use

sample code

text =  '''
    void DropoutLayer::updateB(){

        int num = this->numUnit;
        //cout<<"Error!!!!!!!!!!!!!!!!in DropoutLayer!!!!!!!"<<endl;
        if (this->z == NULL){ // embeddings
            // gradB += dE_dy, because y = b
            iXpY( num , this->dE_dy, gradBiases + bidx );

            return;
        }

        if (fprime != dummy){

            // dy_dz = f', evaluated at y
            ( * this->fprime)(this->y, this->dy_dz, num);
            // dE_dz = dE_dy .* dy_dz

            pointwise_dot(this->dE_dy, this->dy_dz, this->dE_dz, num);
            //cout<<"dropout backward"<<endl;
            for( int i = 0; i < num; ++i){

                this->dE_dz[i] *= this->indicator[i];

            }

        }// else if fprime == softmaxprime{
            // do nothing, because we assume dE_dz is given by softmax

        //}

    //		ReLUPrime(float * y, float * dy_dz, int n);
    }
'''
print(client.search(text, 'cpp')) # return OrderedDict object, key is line_number, value is the commented code itself

# output
# OrderedDict([(4, 'cout<<"Error!!!!!!!!!!!!!!!!in DropoutLayer!!!!!!!"<<endl;'),
#             (19, 'cout<<"dropout backward"<<endl;'),
#             (26, 'else if fprime == softmaxprime{'),
#             (29, '}'),
#             (31, 'ReLUPrime(float * y, float * dy_dz, int n);')])