qubo_lab/util/script.py

import configparser
import os
import argparse
import pymongo
import ssl
import mysql.connector
import networkx as nx
from . import queries
from . import graph
from . import SAT2QUBO
import minorminer
from tqdm import tqdm
import numpy as np
import random
import sys


def readConfig(configFilePath):
    config = configparser.ConfigParser()
    
    if os.path.isfile(configFilePath):
        config.read(configFilePath)

    return config

class ArgParser:
    def __init__(self):
        self.__flags = {}
        self.__parser = argparse.ArgumentParser()
        self.__instanceDirArgSet = False
        self.__config = None
        self.__parsedArgs = {}
        
    def addArg(self, alias,
                     shortFlag,
                     longFlag,
                     help,
                     type,
                     default=None,
                     ignoreDatabaseConfig=False):
                         
        self.__flags[alias] = {"longFlag": longFlag,
                               "hasDefault": False,
                               "ignoreDatabaseConfig": ignoreDatabaseConfig,
                               "type": type}
        
        if default != None:
            self.__flags[alias]["hasDefault"] = True
            
        self.__parser.add_argument("-%s" % shortFlag,
                                   "--%s" % longFlag,
                                   help=help,
                                   type=type,
                                   default=default)
    
    def addInstanceDirArg(self):
        self.__instanceDirArgSet = True
        
        self.addArg(alias="datasetDir", shortFlag="d", longFlag="dataset_dir",
                    help="the base direcotry of the dataset; if this flag is given the others can be omitted",
                    type=str, ignoreDatabaseConfig=True)
        
    
    def parse(self):
        self.__parsedArgs = {}
        args = vars(self.__parser.parse_args())
        
        if self.__instanceDirArgSet:
            self.__config = readConfig(os.path.join(args["dataset_dir"],
                                                    "dataset.config"))
            self.__parseDatasetConfig()
        
        for alias, flag in self.__flags.items():
            self.__parsedArgs[alias] = self.__processFlag(args, flag)
        
        self.__config = None
        
        return self.__parsedArgs
    
    def __parseDatasetConfig(self):
        for flag, value in self.__config["STRUCTURE"].items():
            self.__parsedArgs[flag] = value
    
    def __processFlag(self, args, flag):
        longFlag = flag["longFlag"]
        
        tmpValue = self.__parsedArgs[longFlag] if longFlag in self.__parsedArgs else None
        
        if flag["ignoreDatabaseConfig"] == True:
            tmpValue = None
        
        if args[longFlag]:
            tmpValue = args[longFlag]
        
        if tmpValue == None:
            tmpValue = flag["type"](input("pass arguement %s: " % longFlag))

        
        return tmpValue
    
def getDBContext(dbConfigPath):
    dbContext = {}
    
    dbContext["client"] = connect_to_instance_pool(dbConfigPath)
    dbContext["db"] = dbContext["client"]["experiments"]
    dbContext["instances"] = dbContext["db"]["instances"]
    dbContext["experimentScopes"] = dbContext["db"]["experiment_scopes"]

    return dbContext

def connect_to_instance_pool(dbConfigPath = "database.config"):
    dbConf = readConfig(dbConfigPath)
    
    client = pymongo.MongoClient(
            "mongodb://%s:%s@%s:%s/%s"
            % ( dbConf["INSTANCE_POOL"]["user"],
                dbConf["INSTANCE_POOL"]["pw"],
                dbConf["INSTANCE_POOL"]["url"],
                dbConf["INSTANCE_POOL"]["port"],
                dbConf["INSTANCE_POOL"]["database"]), 
            ssl=True, 
            ssl_cert_reqs=ssl.CERT_NONE)
        
    return client[dbConf["INSTANCE_POOL"]["database"]]
            
def connect_to_experimetns_db(dbConfigPath = "database.config"):
    dbConfig = readConfig(dbConfigPath)
    
    return mysql.connector.connect(
        host=dbConfig["EXPERIMENT_DB"]["url"],
        port=dbConfig["EXPERIMENT_DB"]["port"],
        user=dbConfig["EXPERIMENT_DB"]["user"],
        password=dbConfig["EXPERIMENT_DB"]["pw"],
        database=dbConfig["EXPERIMENT_DB"]["database"]
    )

def frange(start, stop, steps):
    while start < stop:
        yield start
        start += steps
        
def create_experiment_scope(db, description, name):
    experimentScope = {}
    experimentScope["instances"] = []
    
    experimentScope["description"] = description
    
    experimentScope["_id"] = name.strip()
    
    db["experiment_scopes"].insert_one(experimentScope)
    
def write_instance_to_pool_db(db, instance):
    instance_document = instance.writeJSONLike()
    
    result = db["instances"].insert_one(instance_document)
    
    return  result.inserted_id

def add_instance_to_experiment_scope(db, scope_name, instance_id):
    db["experiment_scopes"].update_one(
        {"_id": scope_name},
        {"$push": {"instances": instance_id}}
    )
    
def write_qubo_to_pool_db(collection, qubo, sat_instance_id):
    doc = {}
        
    doc["instance"] = sat_instance_id
    doc["description"] = {"<qubo>": "<entrys>",
                          "<entrys>": "<entry><entrys> | <entry> | \"\"",
                          "<entry>": "<coupler><energy>",
                          "<energy>": "<real_number>",
                          "<coupler>": "<node><node>",
                          "<node>": "<clause><literal>",
                          "<clause>": "<natural_number>",
                          "<literal>": "<integer>"}
    
    doc["qubo"] = __qubo_to_JSON(qubo)
    
    collection.insert_one(doc)
    
def create_wmis_qubos_for_scope(db, scope):
    instances = queries.Instance_scope_query(db)
    instances.query(scope)
    
    for instance, instance_id in instances:
        qubo = SAT2QUBO.WMISdictQUBO(instance)
        
        write_qubo_to_pool_db(db["wmis_qubos"], qubo, instance_id)
    
def __qubo_to_JSON(qubo):
    quboJSON = []
    
    for coupler, value in qubo.items():
        quboJSON.append([coupler, float(value)])
        
    return quboJSON

def write_wmis_embedding_to_pool_db(collection, qubo_id, solver_graph_id, seed, embedding):
    if not __embedding_entry_exists(collection, qubo_id, solver_graph_id):
        __prepare_new_wmis_embedding_entry(collection, qubo_id, solver_graph_id)
        
    collection.update_one(
        {"qubo": qubo_id, "solver_graph": solver_graph_id},
        {
            "$push": 
            {
                "embeddings":
                {
                    "embedding": __embedding_to_array(embedding),
                    "seed": seed
                }
            }
        }
    )
    
def __embedding_entry_exists(collection, qubo_id, solver_graph_id):
    filter = {"qubo": qubo_id, "solver_graph": solver_graph_id}
    
    if collection.count_documents(filter) > 0:
        return True
    
    return False

def __prepare_new_wmis_embedding_entry(collection, qubo_id, solver_graph_id):
    doc = {}
    
    doc["qubo"] = qubo_id
    doc["solver_graph"] = solver_graph_id
    doc["description"] = {"<embedding>": "<chains>",
                          "<chains>": "<chain><chains> | \"\"",
                          "<chain>" : "<original_node><chimera_nodes>",
                          "<chimera_nodes>": "<chimera_node><chimera_nodes> | \"\""}
    doc["embeddings"] = []
    
    collection.insert_one(doc)
    
def __embedding_to_array(embedding):
    emb_arr = []
    
    for node, chain in embedding.items():
        emb_arr.append([node, chain])
    
    return emb_arr

def write_solver_graph_to_pool_db(collection, graph):
    data = nx.node_link_data(graph)
    
    id = queries.get_id_of_solver_graph(collection, data)
    
    if id != None:
        return id
    
    doc = {}
    doc["data"] = data
    
    return collection.insert_one(doc).inserted_id

def find_wmis_embeddings_for_scope(db, scope, solver_graph):
    solver_graph_id = write_solver_graph_to_pool_db(db["solver_graphs"], 
                                                    solver_graph)
    
    qubos = queries.WMIS_scope_query(db)
    qubos.query(scope)
    
    new_embeddings_found = 0
    already_found = 0
    total_count = 0
    for qubo, qubo_id in tqdm(qubos):
        total_count += 1
        
        max_no_improvement = 10
        for i in range(5):
            if __embedding_entry_exists(db["embeddings"], qubo_id, solver_graph_id):
                if i == 0:
                    already_found += 1
                break;
            else: 
                nx_qubo = graph.qubo_to_nx_graph(qubo)
                
                seed = random.randint(0, sys.maxsize)
                
                emb = minorminer.find_embedding(nx_qubo.edges(),
                                                solver_graph.edges(),
                                                return_overlap=True,
                                                max_no_improvement=max_no_improvement,
                                                random_seed=seed)
                
                if emb[1] == 1:
                    write_wmis_embedding_to_pool_db(db["embeddings"],
                                                    qubo_id,
                                                    solver_graph_id,
                                                    seed,
                                                    emb[0])
                    new_embeddings_found += 1
                    
            max_no_improvement *= 1.5
    
    percentage = 0
    
    if total_count > 0:
        percentage = round(((new_embeddings_found + already_found) / total_count) * 100)
    
    print("found {} of {} embeddigns ({}%)".format(new_embeddings_found + already_found,
                                                   total_count,
                                                   percentage))
    print("{} new embeddigns found".format(new_embeddings_found))

def save_sample_set(collection, result, solver_input, emb_list_index, run):
    doc = {}
    
    doc["data"] = result.to_serializable()
    doc["instance"] = solver_input["instance_id"]
    doc["embedding"] = {
        "embedding_id": solver_input["embeddings_id"],
        "list_index": emb_list_index
    }
    doc["run"] = run
    
    collection.insert_one(doc)
    
def save_qpu_result():
    doc = {}
    
def analyze_wmis_sample(sample):
    data = {}
    
    data["number_of_assignments"] = np.count_nonzero(list(sample.sample.values()))
    data["chain_break_fraction"] = sample.chain_break_fraction
    data["num_occurrences"] = sample.num_occurrences
    data["energy"] = sample.energy
    
    return data

def analyde_minisat_run(run_document):
    data = {}
    
    data["satisfiable"] = run_document["satisfiable"]
    
    return data

def majority_vote_sample(sample):
    assignments = {}
    
    for coupler, energy in sample.items():
        
        var = abs(coupler[1])
        
        if var not in assignments:
            assignments[var] = {"all": []}
        
        if energy == 1:
            assignments[var]["all"].append(1 if coupler[1] > 0 else 0)
            
    for var, a in assignments.items():
        assignments[var]["majority"] = 1 if __true_percentage(a["all"]) >= 0.5 else 0
        
    assignment = [0 for i in range(len(assignments))]
    
    for var, a in assignments.items():
        assignment[var - 1] = a["majority"]
        
    return assignment

def __true_percentage(a):
    if len(a) == 0:
        return 0
    
    return np.count_nonzero(a) / len(a)