cmd.py

#!/usr/bin/python3
# line encoding :unix: LF
import rawMatGen
import sparseMatGen
import powerIter
import time
import metric


norm = metric.metric.get1Norm
beta = 0.8
N = 100
rankBlock = 2
tranMatBlock = 2
epsilon = 0.01
topK = 10

# note: block: 一块中含有的pagelink数目,比如block=2,则一块含两条记录.特别的,如果block=1,则代表不分块
# note: maybe the best practise: block=2

# if use mock,we will load&store data from memory,instead of file
# we will always store a copy of data in memory for legacy problem,
# but whether to load in-memory data or in-file data will be redecided by var "mock"
sparseMatGen.mock = False
powerIter.mock = False


def getMethodStr() -> str:
    if rankBlock != 1 and tranMatBlock != 1:
        return "block-strip"
    elif rankBlock != 1 and tranMatBlock == 1:
        return "block-based"
    elif rankBlock == 1 and tranMatBlock == 1:
        return "unblock"
    else:
        raise "not supported method"


def printSuperParam(delim=""):
    print("\033[1;31m"+delim)
    print(
        f"method:{getMethodStr()} norm:{norm.__name__} epsilon:{epsilon} beta:{beta}")
    print(f"N:{N} rankBlock:{rankBlock} transferMatBlock:{tranMatBlock} topK:{topK}")
    print(f"in-memory mock: mat:{sparseMatGen.mock} rank:{powerIter.mock}")
    print(delim, "\033[0m")


printSuperParam("~x~ "*10)


@metric.printTimeElapsed
def pageRank(version=3):
    spraseMat, pr = None, None
    if version == 3 or version == 2:
        sparseMat = sparseMatGen.sparseMat(N, tranMatBlock, version=version)
        pr = powerIter.pageRank(sparseMat, beta, rankBlock)
    elif version == 1:
        rawMat = rawMatGen.rawMat(N)
        sparseMat = sparseMatGen.sparseMat(
            rawMat, tranMatBlock, version=version)
        pr = powerIter.pageRank(sparseMat, beta, rankBlock)
    else:
        raise "not supported version"

    cnt = 0
    loss = 0.
    beg = time.time()

    while (True):
        cnt += 1
        # use rankBlock to iter, if rankBlock != tranMatBlock, it should be a block-based powerIter
        pr.iter(rankBlock)
        ok, loss = pr.isConvergence(epsilon, norm)
        print(f"\033[1;36miter:{cnt} loss:{loss}\033[0m")
        if ok:
            break

    end = time.time()

    print(
        f"\033[1;32m[done] iter:{cnt} , loss={loss} , time:{end-beg}s \033[0m")
    pages = pr.getkBest(topK)
    printSuperParam("=*"*20)
    print(f"topK:{topK}", pages)


pageRank(version=2)