Update default value of particlefile chunks to length of pset

[erikvansebille]

This fixes a bug where zarr output file writing can be very slow when only one particle is released on the first day of simulation, since the default chunks was set to the number of particles to write the first time, instead of the length of the ParticleSet.

Note that the new code also aligns with the information box on output chunking at https://docs.oceanparcels.org/en/latest/examples/tutorial_parcels_structure.html#4.-Execution-and-output

[VeckoTheGecko]

Looks good. Only comment is that if someone uses repeatdt this fix isn't 100% going to work (from what I can tell) since a new pset is created and then combined with the first.

Parcels/parcels/particleset.py

Lines 1216 to 1232 in 7542e4c

	if abs(time - next_prelease) < tol:
	pset_new = self.__class__(
	fieldset=self.fieldset,
	time=time,
	lon=self._repeatlon,
	lat=self._repeatlat,
	depth=self._repeatdepth,
	pclass=self._repeatpclass,
	lonlatdepth_dtype=self.particledata.lonlatdepth_dtype,
	partition_function=False,
	pid_orig=self._repeatpid,
	**self._repeatkwargs,
	)
	for p in pset_new:
	p.dt = dt
	self.add(pset_new)
	next_prelease += self.repeatdt * np.sign(dt)

[erikvansebille]

Looks good. Only comment is that if someone uses repeatdt this fix isn't 100% going to work (from what I can tell) since a new pset is created and then combined with the first.

Yes agree, but that's exactly what the note in https://docs.oceanparcels.org/en/latest/examples/tutorial_parcels_structure.html#4.-Execution-and-output is about; it even explicitly mentions this. Users will just have to change chunks themselves (which is a good strategy anyways to improve Parcels performance)

[VeckoTheGecko]

it even explicitly mentions this

ah gotcha, stopped reading 1 sentence too early 😅