def predict(self,**kw):
from MGT.Taxa import loadTaxaTree
opt = self.opt
prOpt = opt.prOpt
store = self.store
clOpt = prOpt.clOpt.copy()
assert len(clOpt.thresh) == 1
clOpt.runMode = "inproc" #we would need to split this into two methods for "batchDep"
featStoreTrain = self.getFeatStore(opt.featName)
clOpt.modelRoot = pjoin(featStoreTrain.getPath(),"train/1",clOpt.modelRoot)
sampStorePred = SampStore.open(path=opt.sampStorePred)
featStore = sampStorePred.loadStore(name=opt.featName)
featStore.predict(opt=clOpt,name=prOpt.runName)
#nameIL = store.getIdLabsName()+'.tr'
nameIL = None
idLab = store.loadIdLabs(name=nameIL)
labToName = idLab.getLabToName()
predStore = featStore.predictStore(name=prOpt.runName)
pred = predStore.loadObj("pred")
if prOpt.get("outExportDir",None) is not None:
outStore = DirStore.open(path=prOpt.outExportDir,mode="w") #"c" can zap cwd of the user
else:
outStore = predStore
idToName = {}
for id,lab in zip(pred.idPred["id"],pred.labPred[0]):
idToName[id] = labToName[lab]
taxaTree = loadTaxaTree()
idToLin = {}
for id,labn in idToName.items():
if not isinstance(labn,str):
idToLin[id] = taxaTree.getNode(labn).lineageStr()
else:
idToLin[id] = labn
out = open(outStore.getFilePath("idlin.csv"),'w')
for (id,lin) in sorted(idToLin.items()):
out.write("%s\t%s\n" % (id,lin))
out.close()
linCnt = "\n".join(["%s\t%s" % y for y in sorted(binCount(idToLin.values()).items(),key=lambda x:-x[1])])
out = open(outStore.getFilePath("lincnt.csv"),'w')
out.write(linCnt)
out.write("\n")
out.close()
def getTaxaTree(self):
if self.taxaTree is None:
self.taxaTree = loadTaxaTree()
return self.taxaTree
