Issue / bug with initialization of information variable (H) in SingleCounter and MultiCounter classes?

[abhimat]

I noticed the following error when using the resume_from_similar_file() function in ultranest:

Traceback (most recent call last):
  File "/u/home/a/abhimat/research/2023_S2-36_analysis/phitter_fits/202411_3/fit_un_2_Kbb_HeII_only_RVs/run_un_fit.py", line 391, in <module>
    sampler = ultranest.ReactiveNestedSampler(
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/u/home/a/abhimat/.conda/envs/phoebe_py38/lib/python3.12/site-packages/ultranest/integrator.py", line 1219, in __init__
    resume_from_similar_file(
  File "/u/home/a/abhimat/.conda/envs/phoebe_py38/lib/python3.12/site-packages/ultranest/integrator.py", line 342, in resume_from_similar_file
    main_iterator2.passing_node(node2, active_nodes2)
  File "/u/home/a/abhimat/.conda/envs/phoebe_py38/lib/python3.12/site-packages/ultranest/netiter.py", line 536, in passing_node
    self.H = exp(wi - logZnew) * Li + exp(self.logZ - logZnew) * (self.H + self.logZ) - logZnew
                                                                  ~~~~~~~^~~~~~~~~~~
TypeError: unsupported operand type(s) for +: 'NoneType' and 'float'

Upon investigating this issue, It appears to stem from this particular line

UltraNest/ultranest/netiter.py

Line 536 in fbde672

self.H = exp(wi - logZnew) * Li + exp(self.logZ - logZnew) * (self.H + self.logZ) - logZnew

and how the information variable (self.H) appears to be getting initialized in both the SingleCounter and MultiCounter classes, as None in SingleCounter and an array of NaNs in MultiCounter:

UltraNest/ultranest/netiter.py

Lines 483 to 492 in fbde672

	def reset(self):
	"""Reset counters and integration."""
	self.logweights = []
	self.H = None
	self.logZ = -np.inf
	self.logZerr = np.inf
	self.logVolremaining = 0
	self.i = 0
	self.fraction_remaining = np.inf
	self.Lmax = -np.inf

UltraNest/ultranest/netiter.py

Lines 630 to 655 in fbde672

	def reset(self, nentries):
	"""Reset counters/integrator.
	Parameters
	----------
	nentries: int
	number of iterators
	"""
	self.logweights = []
	self.istail = []
	self.logZ = -np.inf
	self.logZerr = np.inf
	self.all_H = -np.nan * np.ones(nentries)
	self.all_logZ = -np.inf * np.ones(nentries)
	self.all_logVolremaining = np.zeros(nentries)
	self.logVolremaining = 0.0
	self.Lmax = -np.inf
	self.all_logZremain = np.inf * np.ones(nentries)
	self.logZremainMax = np.inf
	self.logZremain = np.inf
	self.remainder_ratio = 1.0
	self.remainder_fraction = 1.0
	self.insertion_order_accumulator.reset()
	self.insertion_order_runs = []

Am I correct in initializing self.H = 0? Upon trying it for my use case, it appears to fix the issue.

Just wanted to double check about the initial value before submitting a pull request with the fix, thank you!

[JohannesBuchner]

Sorry, I missed this issue somehow. Your solution seems reasonable, please open a PR, ideally with a unit test that fails without the change.