tl;dr

No need to initialize the CoordinateSequence when we construct a new one because we never used to do that.

The new code (d463bcb)

In the code above, we see that if the CoordinateSequence is generated by calling with two arguments, sz and dim, then initialize is always called. It calls fillVector which calls std::fill and loops through the size of the new vector to initialize all elements to 0.

I think that this is not necessary.

The old code (60edf0e)

Looking at commit 60edf0e we see that creating the CoordinateSequence directly like this is instead of calling getGeometrySequenceFactory()->create(sz, dim).

60edf0e#diff-95f3875df536135fe955deecfaeab0af5f6af52008d0fd5a56a3ed1b6638aca1L2427-R2429

getCoordinateSequenceFactory is a method of the GeometryFactory class:

It returns a CoordinateSequenceFactory, which has a create method. It's here in the old code:

We see that it is a pure virtual function so it must be implemented by inheritors of this class.

There are only two classes that inherit from this class, the DefaultCoordinateSequenceFactory and the CoordinateArraySequenceFactory:

Let's look at them one-by-one

CoordinateArraySequenceFactory

Here's the implementation of create(size_t, size_t):

It's creating a CoordinateArraySequence using the constructor and returning a unique_ptr. Here's that constuctor:

It is resizing the vector to make it the right size but not initializing anything.

DefaultCoordinateSequenceFactory

Here's the implementation of create(size_t, size_t):

It's making a FixedSizeCoordinateSequence,templated on the number of dimensions. It's like a fixed size list, like a tuple! Let's look at that constuctor:

We see that it is just coping the dimension. It is not initializing anything.

Conclusion

For all cases of create(size_t, size_t), the code before 60edf0e did not initialize the coordinates. The new code could also skip the initialization to maintain the same behavior. This would also improve performance because we'd eliminate the slow call to std::fill.