Follow RTD deprecation

[jinningwang]

Follow RTD's deprecation of Sphinx context injection at build time

[cuihantao]

This function is meant to change self.t in place. The new implementation creates a new np.array for self.t. I can't recall other usages of self.t that assume a persistent memory. If that exists, the new implementation will fail: even though t is being updated here, the previous memory is being referred elsewhere.

Can you check: 1) if this function is called elsewhere? 2) any other implementation that changes t in place? I know numpy scalars are immutable; that's why itemset was used.

[jinningwang]

itemset is only used in the reset() method:

andes/andes/variables/dae.py

Lines 458 to 476 in 5ab784b

	def reset(self):
	"""
	Reset array sizes to zero and clear all arrays.
	"""
	self.set_t(0.0)
	self.m = 0
	self.n = 0
	self.o = 0
	self.resize_arrays()
	self.clear_ijv()
	self.clear_ts()
	def set_t(self, t):
	"""
	Helper function for setting time in-place.
	"""
	self.t.itemset(t)

I changed to in-place assignment, and the test shows below:

print(f"Memory address before: {ss.dae.t.ctypes.data}")
ss.dae.reset()
print(f"Memory address after: {ss.dae.t.ctypes.data}")

Memory address before: 4393601808
Memory address after: 4393601808

[cuihantao]

Thanks. Looks good! I didn't know triple dots do the trick!

[jinningwang]

I learnt this from GitHub Copilot just right now, a positive AGI datapoint!