Develop Reference

Improving the runtime of a pandas loop [duplicate]

This question already has an answer here:
Pandas - Explanation on apply function being slow
(1 answer)
Closed 7 months ago.
I am actively running some Python code in jupyter on a df consisting of about 84k rows. I'm estimating this is going to take somewhere in the neighborhood of 9 hours at this rate. My code is below, I have read that ideally one would vectorize for max speed but being sort of new to Python and coding in general, I'm not sure how I can go about changing the below code to vectorize it. The goal is to look at the value in the first column of the dataframe and add that value to the end of a url. I then check the first line in the url and compare it to some predetermined values to find out if there is a match. Any advice would be greatly appreciated!
#Python 3
import pandas as pd
import urllib
no_res = "Item Not Found"
error = "Does Not Exist"
for i in df1.index:
path = 'http://xxxx/xxx/xxx.pl?part=' + str(df1['ITEM_ID'][i])
parsed_path = path.replace(' ','%20')
f = urllib.request.urlopen(parsed_path)
raw = str(f.read().decode("utf-8"))
lines = raw.split('\n')
r = lines[0]
if r == no_res:
sap = 'NO'
elif r == error:
sap = 'ERROR'
else:
sap = 'YES'
df1["item exists"][i] = sap
df1["Path"][i] = path
df1["URL return value"][i] = r
Edit adding test code below
import concurrent.futures
import pandas as pd
import urllib
import numpy as np
def my_func(df_row):
no_res = "random"
error = "entered"
path = "http://www.google.com"
parsed_path = path.replace(' ','%20')
f = urllib.request.urlopen(parsed_path)
raw = str(f.read().decode("utf-8"))
lines = raw.split('\n')
r = df_row['0']
if r == no_res:
sap = "NO"
elif r == error:
sap = "ERROR"
else:
sap = "YES"
df_row['4'] = sap
df_row['5'] = lines[0]
df_row['6'] = r
n = 1000
my_df = pd.DataFrame(np.random.choice(['random','words','entered'], size=(n,3)))
my_df['4'] = ""
my_df['5'] = ""
my_df['6'] = ""
my_df = my_df.apply(lambda col: col.astype('category'))
executor = concurrent.futures.ProcessPoolExecutor(8)
futures = [executor.submit(my_func, row) for _,row in my_df.iterrows()]
concurrent.futures.wait(futures)
This is throwing the following error (shortened):
DoneAndNotDoneFutures(done={<Future at 0x1cfe4938040 state=finished raised BrokenProcessPool>, <Future at 0x1cfe48b8040 state=finished raised BrokenProcessPool>,

Since you are doing some outside operation with a URL, I do not think vectorization is a solution (let possible).
The bottleneck of your operation is the following line
f = urllib.request.urlopen(parsed_path)
This line waits for the response and is blocking, as mentioned your operation is I/O bound. The CPU can start other jobs while waiting for the response. The solution to address this is using concurrency.
Edit: My original answer was using python built-in multi threading which was problematic. The best way to do multiprocessing/threading with pandas data frame is using "dask" library.
The following code is tested with the dummy data set on my PC and on average speeds up the naive for loop by ~ 12 times.
#%%
import time
import urllib.request
import pandas as pd
import numpy as np
import dask.dataframe as dd
def my_func(df_row):
df_row = df_row.copy()
no_res = "random"
error = "entered"
path = "http://www.google.com"
parsed_path = path.replace(' ','%20')
f = urllib.request.urlopen(parsed_path)
# I had to change the encoding on my machine.
raw = str(f.read().decode("'windows-1252"))
lines = raw.split('\n')
r = df_row[0]
if r == no_res:
sap = "NO"
elif r == error:
sap = "ERROR"
else:
sap = "YES"
df_row['4'] = sap
df_row['5'] = lines[0]
df_row['6'] = r
return df_row
def run():
print("started.")
n = 1000
my_df = pd.DataFrame(np.random.choice(['random','words','entered'], size=(n,3)))
my_df = my_df.apply(lambda col: col.astype('category'))
my_df['4'] = ""
my_df['5'] = ""
my_df['6'] = ""
# Literally dask partitions the original dataframe into
# npartitions chunks and use them in apply function
# in parallel.
my_ddf = dd.from_pandas(my_df, npartitions=15)
start = time.time()
q = my_ddf.apply(my_func, axis= 1, meta=my_ddf)
# num_workers is number of threads used,
print(q.compute(num_workers= 50))
time_end = time.time()
print(f"Elapsed: {time_end - start:10.2f}")
if __name__ == "__main__":
run()
dask provides many other tools and options to facilitate concurrent processing and it would be a good idea to take a look at its documentation to investigate other options.
P.S. : if you run the above code too many times on google you will receive "HTTP Error 429: Too Many Requests". This happens to prevent something like DDoS attack on a public server. So, if for your real job you are querying a public website, you may end up receiving the same 429 response, if you try 84K queries in a short time.

Issues getting Python Multiprocessing Library Working when using the Process Function

I'm trying to build a list of parent/comment pairs from the publicly available Reddit data set.
I have a CSV file which I load into a Pandas dataframe which contains rows of the comments with the parent and child id, as well as the child comment. The data is loaded using the following block of code:
import os
import multiprocessing as mp
import numpy as np
import pandas as pd
sourcePATH = r'C:\'
workingFILE = r'\output-pt1.csv'
# filepaths
input_file = sourcePATH + workingFILE
data_df = pd.read_csv(input_file,header=None,names=['PostIDX','ParentIDX','Comment','Score','Controversiality'])
The aim is to scan through each row in the dataframe and using the parent id to search through the rest of the dataframe to see if their is a parent comment present. If it is I then store the child and parent comments in a tuple with some other information. This will then be added to a list which will then be written out to a csv file at the end. To do this I use the following code:
def checkChildParent(ParentIDX_curr, ChildIDX_curr,ChildComment_curr,ChildScore_curr,ChildCont_curr):
idx = data_df.loc[data_df['PostIDX'] == ParentIDX_curr]
if idx.empty is False:
ParentComment = idx.iloc[0,2]
ParentScore = idx.iloc[0,3]
ParentCont = idx.iloc[0,4]
outPut.put([ParentIDX_curr[0], ParentComment,ParentScore,ParentCont,ChildIDX_curr[0], ChildComment_curr[0],ChildScore_curr[0],ChildCont_curr[0]])
if __name__ == '__main__':
print('Process started')
t_start_init = time.time()
t_start = time.time()
noCores = 1
#pool = mp.Pool(processes=noCores)
update_freq = 100
n = 1000
#n = round(len(data_df)/8)
flag_create = 0
flag_run = 0
i = 0
outPut = mp.Queue()
#parent_child_df = pd.DataFrame()
#parent_child_df.coumns = ['PostIDX','ParentIDX']
while i < n:
#print(i)
procs = []
ParentIDX = []
ParentComment = []
ParentScore = []
ParentCont = []
ChildIDX = []
ChildComment = []
ChildScore = []
ChildCont = []
for worker in range(0,noCores):
ParentIDX.append(data_df.iloc[i,1])
ChildIDX.append(data_df.iloc[i,0])
ChildComment.append(data_df.iloc[i,2])
ChildScore.append(data_df.iloc[i,3])
ChildCont.append(data_df.iloc[i,4])
i = i + 1
#when I call the function this way it returns the expected matches
#checkChildParent(ParentIDX,ChildIDX,ChildComment,
# ChildScore,ChildCont)
#when I call the function with Process function nothing appears to be happening
for proc in range(0,noCores):
p = mp.Process(target = checkChildParent, args=(ParentIDX[proc],ChildIDX[proc],ChildComment[proc],ChildScore[proc],ChildCont[proc]))
procs.append(p)
p.start()
#for p in procs:
# p.join()
if outPut.empty() is False:
print(outPut.get())
At the top of the file is a function which scans the dataframe for a given row and returns the tuple of the matched parent and child comment if it was found. If I call this function normally then it works fine, however when I call the function using the Process function it doesn't match anything!. I'm guessing its the form the arguments that are being passed to the function is being passed to the function that is causing the issue, but I have been trying to debug this all afternoon and have failed so far. If anyone has any suggestions then please let me know!
Thanks!

Python script use while loop to keep updating job scripts and multiprocess the tasks in queue

I am trying to write a python script scanning a folder and collect updated SQL script, and then automatically pull data for the SQL script. In the code, a while loop is scanning new SQL file, and send to data pull function. I am having trouble to understand how to make a dynamic queue with while loop, but also have multiprocess to run the tasks in the queue.
The following code has a problem that the while loop iteration will work on a long job before it moves to next iteration and collects other jobs to fill the vacant processor.
Update:
Thanks to #pbacterio for catching the bug, and now the error message is gone. After changing the code, the python code can take all the job scripts during one iteration, and distribute the scripts to four processors. However, it will get hang by a long job to go to next iteration, scanning and submitting the newly added job scripts. Any idea how to reconstruct the code?
I finally figured out the solution see answer below. It turned out what I was looking for is
the_queue = Queue()
the_pool = Pool(4, worker_main,(the_queue,))
For those stumble on the similar idea, following is the whole architecture of this automation script converting a shared drive to a 'server for SQL pulling' or any other job queue 'server'.
a. The python script auto_data_pull.py as shown in the answer. You need to add your own job function.
b. A 'batch script' with following:
start C:\Anaconda2\python.exe C:\Users\bin\auto_data_pull.py
c. Add a task triggered by start computer, run the 'batch script'
That's all. It works.
Python Code:
from glob import glob
import os, time
import sys
import CSV
import re
import subprocess
import pandas as PD
import pypyodbc
from multiprocessing import Process, Queue, current_process, freeze_support
#
# Function run by worker processes
#
def worker(input, output):
for func, args in iter(input.get, 'STOP'):
result = compute(func, args)
output.put(result)
#
# Function used to compute result
#
def compute(func, args):
result = func(args)
return '%s says that %s%s = %s' % \
(current_process().name, func.__name__, args, result)
def query_sql(sql_file): #test func
#jsl file processing and SQL querying, data table will be saved to csv.
fo_name = os.path.splitext(sql_file)[0] + '.csv'
fo = open(fo_name, 'w')
print sql_file
fo.write("sql_file {0} is done\n".format(sql_file))
return "Query is done for \n".format(sql_file)
def check_files(path):
"""
arguments -- root path to monitor
returns -- dictionary of {file: timestamp, ...}
"""
sql_query_dirs = glob(path + "/*/IDABox/")
files_dict = {}
for sql_query_dir in sql_query_dirs:
for root, dirs, filenames in os.walk(sql_query_dir):
[files_dict.update({(root + filename): os.path.getmtime(root + filename)}) for
filename in filenames if filename.endswith('.jsl')]
return files_dict
##### working in single thread
def single_thread():
path = "Y:/"
before = check_files(path)
sql_queue = []
while True:
time.sleep(3)
after = check_files(path)
added = [f for f in after if not f in before]
deleted = [f for f in before if not f in after]
overlapped = list(set(list(after)) & set(list(before)))
updated = [f for f in overlapped if before[f] < after[f]]
before = after
sql_queue = added + updated
# print sql_queue
for sql_file in sql_queue:
try:
query_sql(sql_file)
except:
pass
##### not working in queue
def multiple_thread():
NUMBER_OF_PROCESSES = 4
path = "Y:/"
sql_queue = []
before = check_files(path) # get the current dictionary of sql_files
task_queue = Queue()
done_queue = Queue()
while True: #while loop to check the changes of the files
time.sleep(5)
after = check_files(path)
added = [f for f in after if not f in before]
deleted = [f for f in before if not f in after]
overlapped = list(set(list(after)) & set(list(before)))
updated = [f for f in overlapped if before[f] < after[f]]
before = after
sql_queue = added + updated
TASKS = [(query_sql, sql_file) for sql_file in sql_queue]
# Create queues
#submit task
for task in TASKS:
task_queue.put(task)
for i in range(NUMBER_OF_PROCESSES):
p = Process(target=worker, args=(task_queue, done_queue)).start()
# try:
# p = Process(target=worker, args=(task_queue))
# p.start()
# except:
# pass
# Get and print results
print 'Unordered results:'
for i in range(len(TASKS)):
print '\t', done_queue.get()
# Tell child processes to stop
for i in range(NUMBER_OF_PROCESSES):
task_queue.put('STOP')
# single_thread()
if __name__ == '__main__':
# freeze_support()
multiple_thread()
Reference:
monitor file changes with python script: http://timgolden.me.uk/python/win32_how_do_i/watch_directory_for_changes.html
Multiprocessing:
https://docs.python.org/2/library/multiprocessing.html

Where did you define sql_file in multiple_thread() in
multiprocessing.Process(target=query_sql, args=(sql_file)).start()
You have not defined sql_file in the method and moreover you have used that variable in a for loop. The variable's scope is only confined to the for loop.

Try replacing this:
result = func(*args)
by this:
result = func(args)

I have figured this out. Thank your for the response inspired the thought.
Now the script can run a while loop to monitor the folder for new updated/added SQL script, and then distribute the data pulling to multiple threads. The solution comes from the queue.get(), and queue.put(). I assume the queue object takes care of the communication by itself.
This is the final code --
from glob import glob
import os, time
import sys
import pypyodbc
from multiprocessing import Process, Queue, Event, Pool, current_process, freeze_support
def query_sql(sql_file): #test func
#jsl file processing and SQL querying, data table will be saved to csv.
fo_name = os.path.splitext(sql_file)[0] + '.csv'
fo = open(fo_name, 'w')
print sql_file
fo.write("sql_file {0} is done\n".format(sql_file))
return "Query is done for \n".format(sql_file)
def check_files(path):
"""
arguments -- root path to monitor
returns -- dictionary of {file: timestamp, ...}
"""
sql_query_dirs = glob(path + "/*/IDABox/")
files_dict = {}
try:
for sql_query_dir in sql_query_dirs:
for root, dirs, filenames in os.walk(sql_query_dir):
[files_dict.update({(root + filename): os.path.getmtime(root + filename)}) for
filename in filenames if filename.endswith('.jsl')]
except:
pass
return files_dict
def worker_main(queue):
print os.getpid(),"working"
while True:
item = queue.get(True)
query_sql(item)
def main():
the_queue = Queue()
the_pool = Pool(4, worker_main,(the_queue,))
path = "Y:/"
before = check_files(path) # get the current dictionary of sql_files
while True: #while loop to check the changes of the files
time.sleep(5)
sql_queue = []
after = check_files(path)
added = [f for f in after if not f in before]
deleted = [f for f in before if not f in after]
overlapped = list(set(list(after)) & set(list(before)))
updated = [f for f in overlapped if before[f] < after[f]]
before = after
sql_queue = added + updated
if sql_queue:
for jsl_file in sql_queue:
try:
the_queue.put(jsl_file)
except:
print "{0} failed with error {1}. \n".format(jsl_file, str(sys.exc_info()[0]))
pass
else:
pass
if __name__ == "__main__":
main()

Manage Python Multiprocessing with MongoDB

I'm trying to run my code with a multiprocessing function but mongo keep returning
"MongoClient opened before fork. Create MongoClient with
connect=False, or create client after forking."
I really doesn't understand how i can adapt my code to this.
Basically the structure is:
db = MongoClient().database
db.authenticate('user', 'password', mechanism='SCRAM-SHA-1')
collectionW = db['words']
collectionT = db['sinMemo']
collectionL = db['sinLogic']
def findW(word):
rows = collectionw.find({"word": word})
ind = 0
for row in rows:
ind += 1
id = row["_id"]
if ind == 0:
a = ind
else:
a = id
return a
def trainAI(stri):
...
if findW(word) == 0:
_id = db['words'].insert(
{"_id": getNextSequence(db.counters, "nodeid"), "word": word})
story = _id
else:
story = findW(word)
...
def train(index):
# searching progress
progFile = "./train/progress{0}.txt".format(index)
trainFile = "./train/small_file_{0}".format(index)
if os.path.exists(progFile):
f = open(progFile, "r")
ind = f.read().strip()
if ind != "":
pprint(ind)
i = int(ind)
else:
pprint("No progress saved or progress lost!")
i = 0
f.close()
else:
i = 0
#get the number of line of the file
rangeC = rawbigcount(trainFile)
#fix unicode
non_bmp_map = dict.fromkeys(range(0x10000, sys.maxunicode + 1), 0xfffd)
files = io.open(trainFile, "r", encoding="utf8")
str1 = ""
str2 = ""
filex = open(progFile, "w")
with progressbar.ProgressBar(max_value=rangeC) as bar:
for line in files:
line = line.replace("\n", "")
if i % 2 == 0:
str1 = line.translate(non_bmp_map)
else:
str2 = line.translate(non_bmp_map)
bar.update(i)
trainAI(str1 + " " + str2)
filex.seek(0)
filex.truncate()
filex.write(str(i))
i += 1
#multiprocessing function
maxProcess = 3
def f(l, i):
l.acquire()
train(i + 1)
l.release()
if __name__ == '__main__':
lock = Lock()
for num in range(maxProcess):
pprint("start " + str(num))
Process(target=f, args=(lock, num)).start()
This code is made for reading 4 different file in 4 different process and at the same time insert the data in the database.
I copied only part of the code for make you understand the structure of it.
I've tried to add connect=False to this code but nothing...
db = MongoClient(connect=False).database
db.authenticate('user', 'password', mechanism='SCRAM-SHA-1')
collectionW = db['words']
collectionT = db['sinMemo']
collectionL = db['sinLogic']
then i've tried to move it in the f function (right before train() but what i get is that the program doesn't find collectionW,collectionT and collectionL.
I'm not very expert of python or mongodb so i hope that this is not a silly question.
The code is running under Ubuntu 16.04.2 with python 2.7.12

db.authenticate will have to connect to mongo server and it will try to make a connection. So, even though connect=False is being used, db.authenticate will require a connection to be open.
Why don't you create the mongo client instance after fork? That's look like the easiest solution.

Since db.authenticate must open the MongoClient and connect to the server, it creates connections which won't work in the forked subprocess. Hence, the error message. Try this instead:
db = MongoClient('mongodb://user:password#localhost', connect=False).database
Also, delete the Lock l. Acquiring a lock in one subprocess has no effect on other subprocesses.

Here is how I did it for my problem:
import pathos.pools as pp
import time
import db_access
class MultiprocessingTest(object):
def __init__(self):
pass
def test_mp(self):
data = [[form,'form_number','client_id'] for form in range(5000)]
pool = pp.ProcessPool(4)
pool.map(db_access.insertData, data)
if __name__ == '__main__':
time_i = time.time()
mp = MultiprocessingTest()
mp.test_mp()
time_f = time.time()
print 'Time Taken: ', time_f - time_i
Here is db_access.py:
from pymongo import MongoClient
def insertData(form):
client = MongoClient()
db = client['TEST_001']
db.initialization.insert({
"form": form[0],
"form_number": form[1],
"client_id": form[2]
})
This is happening to your code because you are initiating MongoCLient() once for all the sub-processes. MongoClient is not fork safe. So, initiating inside each function works and let me know if there are other solutions.

Successive multiprocessing

I am filtering huge text files using multiprocessing.py. The code basically opens the text files, works on it, then closes it.
Thing is, I'd like to be able to launch it successively on multiple text files. Hence, I tried to add a loop, but for some reason it doesn't work (while the code works on each file). I believe this is an issue with:
if __name__ == '__main__':
However, I am looking for something else. I tried to create a Launcher and a LauncherCount files like this:
LauncherCount.py:
def setLauncherCount(n):
global LauncherCount
LauncherCount = n
and,
Launcher.py:
import os
import LauncherCount
LauncherCount.setLauncherCount(0)
os.system("OrientedFilterNoLoop.py")
LauncherCount.setLauncherCount(1)
os.system("OrientedFilterNoLoop.py")
...
I import LauncherCount.py, and use LauncherCount.LauncherCount as my loop index.
Of course, this doesn't work too as it edits the variable LauncherCount.LauncherCount locally, so it won't be edited in the imported version of LauncherCount.
Is there any way to edit globally a variable in an imported file? Or, is there any way to do this in any other way? What I need is running a code multiple times, in changing one value, and without using any loop apparently.
Thanks!
Edit: Here is my main code if necessary. Sorry for the bad style ...
import multiprocessing
import config
import time
import LauncherCount
class Filter:
""" Filtering methods """
def __init__(self):
print("launching methods")
# Return the list: [Latitude,Longitude] (elements are floating point numbers)
def LatLong(self,line):
comaCount = []
comaCount.append(line.find(','))
comaCount.append(line.find(',',comaCount[0] + 1))
comaCount.append(line.find(',',comaCount[1] + 1))
Lat = line[comaCount[0] + 1 : comaCount[1]]
Long = line[comaCount[1] + 1 : comaCount[2]]
try:
return [float(Lat) , float(Long)]
except ValueError:
return [0,0]
# Return a boolean:
# - True if the Lat/Long is within the Lat/Long rectangle defined by:
# tupleFilter = (minLat,maxLat,minLong,maxLong)
# - False if not
def LatLongFilter(self,LatLongList , tupleFilter) :
if tupleFilter[0] <= LatLongList[0] <= tupleFilter[1] and
tupleFilter[2] <= LatLongList[1] <= tupleFilter[3]:
return True
else:
return False
def writeLine(self,key,line):
filterDico[key][1].write(line)
def filteringProcess(dico):
myFilter = Filter()
while True:
try:
currentLine = readFile.readline()
except ValueError:
break
if len(currentLine) ==0: # Breaks at the end of the file
break
if len(currentLine) < 35: # Deletes wrong lines (too short)
continue
LatLongList = myFilter.LatLong(currentLine)
for key in dico:
if myFilter.LatLongFilter(LatLongList,dico[key][0]):
myFilter.writeLine(key,currentLine)
###########################################################################
# Main
###########################################################################
# Open read files:
readFile = open(config.readFileList[LauncherCount.LauncherCount][1], 'r')
# Generate writing files:
pathDico = {}
filterDico = config.filterDico
# Create outputs
for key in filterDico:
output_Name = config.readFileList[LauncherCount.LauncherCount][0][:-4]
+ '_' + key +'.log'
pathDico[output_Name] = config.writingFolder + output_Name
filterDico[key] = [filterDico[key],open(pathDico[output_Name],'w')]
p = []
CPUCount = multiprocessing.cpu_count()
CPURange = range(CPUCount)
startingTime = time.localtime()
if __name__ == '__main__':
### Create and start processes:
for i in CPURange:
p.append(multiprocessing.Process(target = filteringProcess ,
args = (filterDico,)))
p[i].start()
### Kill processes:
while True:
if [p[i].is_alive() for i in CPURange] == [False for i in CPURange]:
readFile.close()
for key in config.filterDico:
config.filterDico[key][1].close()
print(key,"is Done!")
endTime = time.localtime()
break
print("Process started at:",startingTime)
print("And ended at:",endTime)

To process groups of files in sequence while working on files within a group in parallel:
#!/usr/bin/env python
from multiprocessing import Pool
def work_on(args):
"""Process a single file."""
i, filename = args
print("working on %s" % (filename,))
return i
def files():
"""Generate input filenames to work on."""
#NOTE: you could read the file list from a file, get it using glob.glob, etc
yield "inputfile1"
yield "inputfile2"
def process_files(pool, filenames):
"""Process filenames using pool of processes.
Wait for results.
"""
for result in pool.imap_unordered(work_on, enumerate(filenames)):
#NOTE: in general the files won't be processed in the original order
print(result)
def main():
p = Pool()
# to do "successive" multiprocessing
for filenames in [files(), ['other', 'bunch', 'of', 'files']]:
process_files(p, filenames)
if __name__=="__main__":
main()
Each process_file() is called in sequence after the previous one has been complete i.e., the files from different calls to process_files() are not processed in parallel.