Difference between create_state_from_gsd and create_state_from_snapshot?

[tylah25]

What is the difference between the create_state_from_gsd function and create_state_from_snapshot? Besides the obvious differences in what they do (create the simulation state from a GSD file or a snapshot), I've encountered instances where the two functions give me different results.

For example, if I create an initial configuration file with just one bond type that is not yet present in the system (I will be adding the bonds later with an MD/MC scheme), create_state_from_gsd does not seem to read in this bond type, while create_state_from_snapshot does.

[joaander]

There should be no functional difference beyond the obvious. You should find it takes less time to call create_state_from_gsd than it does to implement the equivalent snapshot creation in Python.

Regarding the bond types behavior you describe, feel free to open a bug with a minimal test case that shows the problem (i.e. it only has to create a Simulation, call create_state_from_gsd then print simulation.state.bond_types). I would welcome a pull request that fixes the bug. I would guess the root cause is that the types are read inside the N > 0 check:

hoomd-blue/hoomd/GSDReader.cc

Lines 251 to 288 in 37f6348

	readChunk(&N, m_frame, "bonds/N", 4);
	if (N > 0)
	{
	m_snapshot->bond_data.resize(N);
	m_snapshot->bond_data.type_mapping = readTypes(m_frame, "bonds/types");
	readChunk(&m_snapshot->bond_data.type_id[0], m_frame, "bonds/typeid", N * 4, N);
	readChunk(&m_snapshot->bond_data.groups[0], m_frame, "bonds/group", N * 8, N);
	}
	N = 0;
	readChunk(&N, m_frame, "angles/N", 4);
	if (N > 0)
	{
	m_snapshot->angle_data.resize(N);
	m_snapshot->angle_data.type_mapping = readTypes(m_frame, "angles/types");
	readChunk(&m_snapshot->angle_data.type_id[0], m_frame, "angles/typeid", N * 4, N);
	readChunk(&m_snapshot->angle_data.groups[0], m_frame, "angles/group", N * 12, N);
	}
	N = 0;
	readChunk(&N, m_frame, "dihedrals/N", 4);
	if (N > 0)
	{
	m_snapshot->dihedral_data.resize(N);
	m_snapshot->dihedral_data.type_mapping = readTypes(m_frame, "dihedrals/types");
	readChunk(&m_snapshot->dihedral_data.type_id[0], m_frame, "dihedrals/typeid", N * 4, N);
	readChunk(&m_snapshot->dihedral_data.groups[0], m_frame, "dihedrals/group", N * 16, N);
	}
	N = 0;
	readChunk(&N, m_frame, "impropers/N", 4);
	if (N > 0)
	{
	m_snapshot->improper_data.resize(N);
	m_snapshot->improper_data.type_mapping = readTypes(m_frame, "impropers/types");
	readChunk(&m_snapshot->improper_data.type_id[0], m_frame, "impropers/typeid", N * 4, N);
	readChunk(&m_snapshot->improper_data.groups[0], m_frame, "impropers/group", N * 16, N);
	}

[joaander]

This bug will be fixed in the next release: #1729