Hi Borek, is this loop only supposed to copy data over? In that case

example11 simplifications about mupif HOT 4 CLOSED

mupif commented on May 28, 2024

example11 simplifications

from mupif.

Comments (4)

eudoxos commented on May 28, 2024

I added the simplified options as default branches to the code; feel free to delete the low-level variants if you don't see them as necessary to keep

mupif/examples/Example11/model2.py

Lines 75 to 152 in b69bdf2

 def solveStep(self, tstep, stageID=0, runInBackground=False): 

 # (1) read source grain state, into new state, then (2) replace a molecule with dopant (different molecule) 

 print(self.inputGrainState) 

 # 

 # (1) copy old state into a new one 

 # 

 if 1: 

 # transfer via file copy of the underlying storage, very fast (local access only!) 

 fd,h5=tempfile.mkstemp(prefix='mupif-example11-',suffix='.h5') 

 shutil.copy(self.inputGrainState.h5path,h5) 

 self.outputGrainState=mp.HeavyDataHandle(h5path=h5,id=mp.dataid.MiscID.ID_GrainState) 

 inGrains = self.inputGrainState.getData(mode='readonly') 

 outGrains=self.outputGrainState.getData(mode='readwrite') 

 elif 1: 

 # transfer via inject (serializes into RAM, network-transparent) 

 inGrains = self.inputGrainState.getData(mode='readonly') 

 self.outputGrainState=mp.HeavyDataHandle(id=mp.dataid.MiscID.ID_GrainState) 

 log.warning(f'Created temporary {self.outputGrainState.h5path}') 

 outGrains = self.outputGrainState.getData(mode='create',schemaName='org.mupif.sample.grain',schemasJson=mp.heavydata.sampleSchemas_json) 

 outGrains.inject(inGrains) 

 else: 

 # transfer via explicit loop over data 

 inGrains = self.inputGrainState.getData(mode='readonly') 

 self.outputGrainState=mp.HeavyDataHandle(id=mp.dataid.MiscID.ID_GrainState) 

 log.warning(f'Created temporary {self.outputGrainState.h5path}') 

 outGrains = self.outputGrainState.getData(mode='create',schemaName='org.mupif.sample.grain',schemasJson=mp.heavydata.sampleSchemas_json) 

 t0=time.time() 

 atomCounter=0 

 outGrains.resize(size=len(inGrains)) 

 #todo: copy inGrains to outGrains (check for more elegant way) 

 for igNum,ig in enumerate(inGrains): 

 outGrains[igNum].getMolecules().resize(size=len(ig.getMolecules())) 

 for imNum, im in enumerate(ig.getMolecules()): 

 om = outGrains[igNum].getMolecules()[imNum] 

 om.getIdentity().setMolecularWeight(im.getIdentity().getMolecularWeight()) 

 om.getAtoms().resize(size=len(im.getAtoms())) 

 for iaNum, ia in enumerate(im.getAtoms()): 

 oa = om.getAtoms()[iaNum] 

 oa.getIdentity().setElement(ia.getIdentity().getElement()) 

 oa.getProperties().getTopology().setPosition(ia.getProperties().getTopology().getPosition()) 

 oa.getProperties().getTopology().setVelocity(ia.getProperties().getTopology().getVelocity()) 

 oa.getProperties().getTopology().setStructure(ia.getProperties().getTopology().getStructure()) 

 atomCounter+=1 

 t1=time.time() 

 print(f'{atomCounter} atoms created in {t1-t0:g} sec ({atomCounter/(t1-t0):g}/sec).') 

 # 

 # replace one molecule in outGrains with a different molecule 

 # 

 if 1: 

 # use inject to replace outGrains/1/molecules/2 with inGrains/0/molecules/3 

 outGrains[1].getMolecules()[2].inject(inGrains[0].getMolecules()[3]) 

 else: 

 # explicit loop over data, random molecule choice 

 # select random grain and molecule 

 t0=time.time() 

 atomCounter = 0 

 rgNum = random.randint(0,len(outGrains)-1) 

 rmNum = random.randint(0,len(outGrains[rgNum].getMolecules())-1) 

 repMol = outGrains[rgNum].getMolecules()[rmNum] 

 # replace this molecule 

 repMol.getIdentity().setMolecularWeight(random.randint(1,10)*u.yg) 

 if (1): 

 #print(repMol.getAtoms()[0]) # call _T_assertDataset() 

 #print (repMol.getAtoms()) 

 #print("Deleting "+repMol.getAtoms().ctx.h5group.name+'/'+repMol.getAtoms()[0].datasetName) 

 ##todo: make a method to solve this 

 #del self.outputGrainState._h5obj[repMol.getAtoms().ctx.h5group.name+'/'+repMol.getAtoms()[0].datasetName] 

 repMol.getAtoms().resize(size=random.randint(30,60),reset=True) 

 #print (repMol.getAtoms()[0]) 

 for a in repMol.getAtoms(): 

 a.getIdentity().setElement(random.choice(['H','N','Cl','Na','Fe'])) 

 a.getProperties().getTopology().setPosition((1,2,3)*u.nm) 

 a.getProperties().getTopology().setVelocity((24,5,77)*u.m/u.s) 

 struct=np.array([random.randint(1,20) for i in range(random.randint(5,20))],dtype='l') 

 a.getProperties().getTopology().setStructure(struct) 

 atomCounter+=1 

 t1=time.time() 

 print(f'{atomCounter} atoms replaced in {t1-t0:g} sec ({atomCounter/(t1-t0):g}/sec).')

from mupif.

bpatzak commented on May 28, 2024

Hi, I think the first option (copy old state into a new one) should be avoided, as the use needs knowledge about internal representation (needs to copy file). The code at this level should abstract from this and rely solely on methods provided by a datatype.
The second option is ok. But what is the difference between inject and set method?

from mupif.

eudoxos commented on May 28, 2024

Hi, I think the first option (copy old state into a new one) should be avoided, as the use needs knowledge about internal representation (needs to copy file).

The file copy could be hidden by a method or perhaps by an additional data access mode, e.g. getData(mode='clone',...). It is by far the most efficient way of cloning the data.

But what is the difference between inject and set method?

inject takes another context as parameter (the data pass through serialization/deserizalization in RAM) and replaces the whole subtree. set* methods are only defined for leaf nodes (such as setMolecularWeight(…)) , not for a whole hierarchy of data (there is no setMolecules(…), for instance).

from mupif.

bpatzak commented on May 28, 2024

If abstracted by method (like copy()) then the first approach is ok. My objections were mainly to the direct file copying in workflow.
Related to the second, from a perspective of the user, who has no knowledge about the internals, the meaning of inject and set methods is quite close. Can we merge them together? So there will be jut set method defined for leaf nodes as well as for sub trees.

from mupif.

example11 simplifications about mupif HOT 4 CLOSED

Comments (4)

Related Issues (20)

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent

	def solveStep(self, tstep, stageID=0, runInBackground=False):
	# (1) read source grain state, into new state, then (2) replace a molecule with dopant (different molecule)
	print(self.inputGrainState)
	#
	# (1) copy old state into a new one
	#
	if 1:
	# transfer via file copy of the underlying storage, very fast (local access only!)
	fd,h5=tempfile.mkstemp(prefix='mupif-example11-',suffix='.h5')
	shutil.copy(self.inputGrainState.h5path,h5)
	self.outputGrainState=mp.HeavyDataHandle(h5path=h5,id=mp.dataid.MiscID.ID_GrainState)
	inGrains = self.inputGrainState.getData(mode='readonly')
	outGrains=self.outputGrainState.getData(mode='readwrite')
	elif 1:
	# transfer via inject (serializes into RAM, network-transparent)
	inGrains = self.inputGrainState.getData(mode='readonly')
	self.outputGrainState=mp.HeavyDataHandle(id=mp.dataid.MiscID.ID_GrainState)
	log.warning(f'Created temporary {self.outputGrainState.h5path}')
	outGrains = self.outputGrainState.getData(mode='create',schemaName='org.mupif.sample.grain',schemasJson=mp.heavydata.sampleSchemas_json)
	outGrains.inject(inGrains)
	else:
	# transfer via explicit loop over data
	inGrains = self.inputGrainState.getData(mode='readonly')
	self.outputGrainState=mp.HeavyDataHandle(id=mp.dataid.MiscID.ID_GrainState)
	log.warning(f'Created temporary {self.outputGrainState.h5path}')
	outGrains = self.outputGrainState.getData(mode='create',schemaName='org.mupif.sample.grain',schemasJson=mp.heavydata.sampleSchemas_json)
	t0=time.time()
	atomCounter=0
	outGrains.resize(size=len(inGrains))
	#todo: copy inGrains to outGrains (check for more elegant way)
	for igNum,ig in enumerate(inGrains):
	outGrains[igNum].getMolecules().resize(size=len(ig.getMolecules()))
	for imNum, im in enumerate(ig.getMolecules()):
	om = outGrains[igNum].getMolecules()[imNum]
	om.getIdentity().setMolecularWeight(im.getIdentity().getMolecularWeight())
	om.getAtoms().resize(size=len(im.getAtoms()))
	for iaNum, ia in enumerate(im.getAtoms()):
	oa = om.getAtoms()[iaNum]
	oa.getIdentity().setElement(ia.getIdentity().getElement())
	oa.getProperties().getTopology().setPosition(ia.getProperties().getTopology().getPosition())
	oa.getProperties().getTopology().setVelocity(ia.getProperties().getTopology().getVelocity())
	oa.getProperties().getTopology().setStructure(ia.getProperties().getTopology().getStructure())
	atomCounter+=1
	t1=time.time()
	print(f'{atomCounter} atoms created in {t1-t0:g} sec ({atomCounter/(t1-t0):g}/sec).')
	#
	# replace one molecule in outGrains with a different molecule
	#
	if 1:
	# use inject to replace outGrains/1/molecules/2 with inGrains/0/molecules/3
	outGrains[1].getMolecules()[2].inject(inGrains[0].getMolecules()[3])
	else:
	# explicit loop over data, random molecule choice
	# select random grain and molecule
	t0=time.time()
	atomCounter = 0
	rgNum = random.randint(0,len(outGrains)-1)
	rmNum = random.randint(0,len(outGrains[rgNum].getMolecules())-1)
	repMol = outGrains[rgNum].getMolecules()[rmNum]
	# replace this molecule
	repMol.getIdentity().setMolecularWeight(random.randint(1,10)*u.yg)
	if (1):
	#print(repMol.getAtoms()[0]) # call _T_assertDataset()
	#print (repMol.getAtoms())
	#print("Deleting "+repMol.getAtoms().ctx.h5group.name+'/'+repMol.getAtoms()[0].datasetName)
	##todo: make a method to solve this
	#del self.outputGrainState._h5obj[repMol.getAtoms().ctx.h5group.name+'/'+repMol.getAtoms()[0].datasetName]
	repMol.getAtoms().resize(size=random.randint(30,60),reset=True)
	#print (repMol.getAtoms()[0])
	for a in repMol.getAtoms():
	a.getIdentity().setElement(random.choice(['H','N','Cl','Na','Fe']))
	a.getProperties().getTopology().setPosition((1,2,3)*u.nm)
	a.getProperties().getTopology().setVelocity((24,5,77)*u.m/u.s)
	struct=np.array([random.randint(1,20) for i in range(random.randint(5,20))],dtype='l')
	a.getProperties().getTopology().setStructure(struct)
	atomCounter+=1
	t1=time.time()
	print(f'{atomCounter} atoms replaced in {t1-t0:g} sec ({atomCounter/(t1-t0):g}/sec).')