How to merge only the unique lines from file_a to file_b?

Question

This question has been asked here in one form or another but not quite the thing I'm looking for. So, this is the situation I shall be having: I already have one file, named file_a and I'm creating another file - file_b. file_a is always bigger than file_b in size. There will be a number of duplicate lines in file_b (hence, in file_a as well) but both the files will have some unique lines. What I want to do is: to copy/merge only the unique lines from file_a to file_b and then sort the line order, so that the file_b becomes the most up-to-date one with all the unique entries. Either of the original files shouldn't be more than 10MB in size. What's the most efficient (and fastest) way I can do that?

I was thinking something like that, which does the merging alright.

#!/usr/bin/env python

import os, time, sys

# Convert Date/time to epoch
def toEpoch(dt):
    dt_ptrn = '%d/%m/%y %H:%M:%S'
    return int(time.mktime(time.strptime(dt, dt_ptrn)))

# input files
o_file = "file_a"
c_file = "file_b"
n_file = [o_file,c_file]

m_file = "merged.file"

for x in range(len(n_file)):
    P = open(n_file[x],"r")
    output = P.readlines()
    P.close()

    # Sort the output, order by 2nd last field
    #sp_lines = [ line.split('\t') for line in output ]
    #sp_lines.sort( lambda a, b: cmp(toEpoch(a[-2]),toEpoch(b[-2])) )

    F = open(m_file,'w') 
    #for line in sp_lines:
    for line in output:
        if "group_" in line:
            F.write(line)
    F.close()

But, it's:

• not with only the unique lines
• not sorted (by next to last field)
• and introduces the 3rd file i.e. m_file

Just a side note (long story short): I can't use sorted() here as I'm using v2.3, unfortunately. The input files look like this:

On 23/03/11 00:40:03
JobID   Group.User          Ctime   Wtime   Status  QDate               CDate
===================================================================================
430792  group_atlas.pltatl16    0   32  4   02/03/11 21:52:38   02/03/11 22:02:15
430793  group_atlas.atlas084    30  472 4   02/03/11 21:57:43   02/03/11 22:09:35
430794  group_atlas.atlas084    12  181 4   02/03/11 22:02:37   02/03/11 22:05:42
430796  group_atlas.atlas084    8   185 4   02/03/11 22:02:38   02/03/11 22:05:46

I tried to use cmp() to sort by the 2nd last field but, I think, it doesn't work just because of the first 3 lines of the input files.

Can anyone please help? Cheers!!!

---

Update 1:

For the future reference, as suggested by Jakob, here is the complete script. It worked just fine.

#!/usr/bin/env python

import os, time, sys
from sets import Set as set

def toEpoch(dt):
    dt_ptrn = '%d/%m/%y %H:%M:%S'
    return int(time.mktime(time.strptime(dt, dt_ptrn)))

def yield_lines(fileobj):
    #I want to discard the headers
    for i in xrange(3):
        fileobj.readline()
    #
    for line in fileobj:
        yield line

def app(path1, path2):
    file1 = set(yield_lines(open(path1)))
    file2 = set(yield_lines(open(path2)))
    return file1.union(file2)

# Input files
o_file = "testScript/03"
c_file = "03.bak"
m_file = "finished.file"

print time.strftime('%H:%M:%S', time.localtime())

# Sorting the output, order by 2nd last field
sp_lines = [ line.split('\t') for line in app(o_file, c_file) ]
sp_lines.sort( lambda a, b: cmp(toEpoch(a[-2]),toEpoch(b[-2])) )

F = open(m_file,'w')
print "No. of lines: ",len(sp_lines)

for line in sp_lines:

    MF = '\t'.join(line)
    F.write(MF)
F.close()

It took about 2m:47s to finish for 145244 lines.

[testac1@serv07 ~]$ ./uniq-merge.py 
17:19:21
No. of lines:  145244
17:22:08

thanks!!

---

Update 2:

Hi eyquem, this is the Error message I get when I run your script(s).

From the first script:

[testac1@serv07 ~]$ ./uniq-merge_2.py 
  File "./uniq-merge_2.py", line 44
    fm.writelines( '\n'.join(v)+'\n' for k,v in output )
                                       ^
SyntaxError: invalid syntax

From the second script:

[testac1@serv07 ~]$ ./uniq-merge_3.py 
  File "./uniq-merge_3.py", line 24
    output = sett(line.rstrip() for line in fa)
                                  ^
SyntaxError: invalid syntax

Cheers!!

---

Update 3:

The previous one wasn't sorting the list at all. Thanks to eyquem to pointing that out. Well, it does now. This is a further modification to Jakob's version - I converted the set:app(path1, path2) to a list:myList() and then applied the sort( lambda ... ) to the myList to sort the merged file by the nest to last field. This is the final script.

#!/usr/bin/env python

import os, time, sys
from sets import Set as set

def toEpoch(dt):
    # Convert date/time to epoch
    dt_ptrn = '%d/%m/%y %H:%M:%S'
    return int(time.mktime(time.strptime(dt, dt_ptrn)))

def yield_lines(fileobj):
    # Discard the headers (1st 3 lines)
    for i in xrange(3):
        fileobj.readline()

    for line in fileobj:
        yield line

def app(path1, path2):
    # Remove duplicate lines
    file1 = set(yield_lines(open(path1)))
    file2 = set(yield_lines(open(path2)))
    return file1.union(file2)

print time.strftime('%H:%M:%S', time.localtime())

# I/O files
o_file = "testScript/03"
c_file = "03.bak"
m_file = "finished.file"

# Convert set into to list
myList = list(app(o_file, c_file))

# Sort the list by the date
sp_lines = [ line.split('\t') for line in myList ]
sp_lines.sort( lambda a, b: cmp(toEpoch(a[-2]),toEpoch(b[-2])) )

F = open(m_file,'w')
print "No. of lines: ",len(sp_lines)

# Finally write to the outFile
for line in sp_lines:
    MF = '\t'.join(line)
    F.write(MF)
F.close()

There is no speed boost at all, it took 2m:50s to process the same 145244 lines. Is anyone see any scope of improvement, please let me know. Thanks to Jakob and eyquem for their time. Cheers!!

---

Update 4:

Just for future reference, this is a modified version of eyguem, which works much better and faster then the previous ones.

#!/usr/bin/env python

import os, sys, re
from sets import Set as sett
from time import mktime, strptime, strftime

def sorting_merge(o_file, c_file, m_file ):

    # RegEx for Date/time filed
    pat = re.compile('[0123]\d/[01]\d/\d{2} [012]\d:[0-6]\d:[0-6]\d')

    def kl(lines,pat = pat):
        # match only the next to last field
        line = lines.split('\t')
        line = line[-2]
        return mktime(strptime((pat.search(line).group()),'%d/%m/%y %H:%M:%S'))

    output = sett()
    head = []

    # Separate the header & remove the duplicates
    def rmHead(f_n):
        f_n.readline()
        for line1 in f_n:
            if pat.search(line1):  break
            else:  head.append(line1) # line of the header
        for line in f_n:
            output.add(line.rstrip())
        output.add(line1.rstrip())
        f_n.close()

    fa = open(o_file, 'r')
    rmHead(fa)

    fb = open(c_file, 'r')
    rmHead(fb)

    # Sorting date-wise
    output = [ (kl(line),line.rstrip()) for line in output if line.rstrip() ]
    output.sort()

    fm = open(m_file,'w')
    # Write to the file & add the header
    fm.write(strftime('On %d/%m/%y %H:%M:%S\n')+(''.join(head[0]+head[1])))
    for t,line in output:
        fm.write(line + '\n')
    fm.close()


c_f = "03_a"
o_f = "03_b"

sorting_merge(o_f, c_f, 'outfile.txt')

This version is much faster - 6.99 sec. for 145244 lines compare to the 2m:47s - then the previous one using lambda a, b: cmp(). Thanks to eyquem for all his support. Cheers!!

Answer 1

Maybe something along these lines?

from sets import Set as set

def yield_lines(fileobj):
    #I want to discard the headers
    for i in xrange(3):
        fileobj.readline()

    for line in fileobj:
        yield line

def app(path1, path2):
    file1 = set(yield_lines(open(path1)))
    file2 = set(yield_lines(open(path2)))

    return file1.union(file2)

EDIT: Forgot about with :$

Answer 2

EDIT 2

My previous codes have problems with output = sett(line.rstrip() for line in fa) and output.sort(key=kl)

Moreover, they have some complications.

So I examined the choice of reading the files directly with a set() function taken by Jakob Bowyer in his code.

Congratulations Jakob ! (and Michal Chruszcz by the way) : set() is unbeatable, it's faster than a reading one line at a time.

Then , I abandonned my idea to read the files line after line.

.

But I kept my idea to avoid a sorting with the help of cmp() function because, as it is described in the doc:

s.sort([cmpfunc=None])

The sort() method takes an optional argument specifying a comparison function of two arguments (list items) (...) Note that this slows the sorting process down considerably

http://docs.python.org/release/2.3/lib/typesseq-mutable.html

Then, I managed to obtain a list of tuples (t,line) in which the t is

time.mktime(time.strptime(( 1st date-and-hour in line ,'%d/%m/%y %H:%M:%S'))

by the instruction

output = [ (kl(line),line.rstrip()) for line in output]

.

I tested 2 codes. The following one in which 1st date-and-hour in line is computed thanks to a regex:

def kl(line,pat = pat):
    return time.mktime(time.strptime((pat.search(line).group()),'%d/%m/%y %H:%M:%S'))

output = [ (kl(line),line.rstrip()) for line in output if line.rstrip()]

output.sort()

And a second code in which kl() is:

def kl(line,pat = pat):
    return time.mktime(time.strptime(line.split('\t')[-2],'%d/%m/%y %H:%M:%S'))

.

The results are

Times of execution:

0.03598 seconds for the first code with regex

0.03580 seconds for the second code with split('\t')

that is to say the same

This algorithm is faster than a code using a function cmp() :

a code in which the set of lines output isn't transformed in a list of tuples by

output = [ (kl(line),line.rstrip()) for line in output]

but is only transformed in a list of the lines (without duplicates, then) and sorted with a function mycmp() (see the doc):

def mycmp(a,b):
    return cmp(time.mktime(time.strptime(a.split('\t')[-2],'%d/%m/%y %H:%M:%S')),
               time.mktime(time.strptime(b.split('\t')[-2],'%d/%m/%y %H:%M:%S')))

output = [ line.rstrip() for line in output] # not list(output) , to avoid the problem of newline of the last line of each file
output.sort(mycmp)

for line in output:
    fm.write(line+'\n')

has an execution time of

0.11574 seconds

.

The code:

#!/usr/bin/env python

import os, time, sys, re
from sets import Set as sett

def sorting_merge(o_file , c_file, m_file ):

    pat = re.compile('[0123]\d/[01]\d/\d{2} [012]\d:[0-6]\d:[0-6]\d'
                     '(?=[ \t]+[0123]\d/[01]\d/\d{2} [012]\d:[0-6]\d:[0-6]\d)') 

    def kl(line,pat = pat):
        return time.mktime(time.strptime((pat.search(line).group()),'%d/%m/%y %H:%M:%S'))

    output = sett()
    head = []

    fa = open(o_file)
    fa.readline() # first line is skipped
    while True:
        line1 = fa.readline()
        mat1  = pat.search(line1)
        if not mat1: head.append(line1) # line1 is here a line of the header
        else: break # the loop ends on the first line1 not being a line of the heading
    output = sett( fa )
    fa.close()

    fb = open(c_file)
    while True:
        line1 = fb.readline()
        if pat.search(line1):  break
    output = output.union(sett( fb ))
    fb.close()

    output = [ (kl(line),line.rstrip()) for line in output]
    output.sort()

    fm = open(m_file,'w')
    fm.write(time.strftime('On %d/%m/%y %H:%M:%S\n')+(''.join(head)))
    for t,line in output:
        fm.write(line + '\n')
    fm.close()


te = time.clock()
sorting_merge('ytre.txt','tataye.txt','merged.file.txt')
print time.clock()-te

This time, I hope it will run correctly, and that the only thing to do is to wait the times of execution on real files much bigger than the ones on which I tested the codes

.

EDIT 3

pat = re.compile('[0123]\d/[01]\d/\d{2} [012]\d:[0-6]\d:[0-6]\d'
                 '(?=[ \t]+'
                 '[0123]\d/[01]\d/\d{2} [012]\d:[0-6]\d:[0-6]\d'
                 '|'
                 '[ \t]+aborted/deleted)')

.

EDIT 4

#!/usr/bin/env python

import os, time, sys, re
from sets import Set

def sorting_merge(o_file , c_file, m_file ):

    pat = re.compile('[0123]\d/[01]\d/\d{2} [012]\d:[0-6]\d:[0-6]\d'
                     '(?=[ \t]+'
                     '[0123]\d/[01]\d/\d{2} [012]\d:[0-6]\d:[0-6]\d'
                     '|'
                     '[ \t]+aborted/deleted)')

    def kl(line,pat = pat):
        return time.mktime(time.strptime((pat.search(line).group()),'%d/%m/%y %H:%M:%S'))

    head = []
    output = Set()

    fa = open(o_file)
    fa.readline() # first line is skipped
    for line1 in fa:
        if pat.search(line1):  break # first line after the heading
        else:  head.append(line1) # line of the header
    for line in fa:
        output.add(line.rstrip())
    output.add(line1.rstrip())
    fa.close()

    fb = open(c_file)
    for line1 in fb:
        if pat.search(line1):  break
    for line in fb:
        output.add(line.rstrip())
    output.add(line1.rstrip())
    fb.close()

    if '' in output:  output.remove('')
    output = [ (kl(line),line) for line in output]
    output.sort()

    fm = open(m_file,'w')
    fm.write(time.strftime('On %d/%m/%y %H:%M:%S\n')+(''.join(head)))
    for t,line in output:
        fm.write(line+'\n')
    fm.close()

te = time.clock()
sorting_merge('A.txt','B.txt','C.txt')
print time.clock()-te

Answer 3

I wrote this new code, with the ease of using a set. It is faster that my previous code. And, it seems, than your code

#!/usr/bin/env python

import os, time, sys, re
from sets import Set as sett

def sorting_merge(o_file , c_file, m_file ):

    # Convert Date/time to epoch
    def toEpoch(dt):
        dt_ptrn = '%d/%m/%y %H:%M:%S'
        return int(time.mktime(time.strptime(dt, dt_ptrn)))

    pat = re.compile('([0123]\d/[01]\d/\d{2} [012]\d:[0-6]\d:[0-6]\d)'
                     '[ \t]+[0123]\d/[01]\d/\d{2} [012]\d:[0-6]\d:[0-6]\d') 

    fa = open(o_file)
    head = []
    fa.readline()
    while True:
        line1 = fa.readline()
        mat1 = pat.search(line1)
        if not mat1:
            head.append(('',line1.rstrip()))
        else:
            break
    output = sett((toEpoch(pat.search(line).group(1)) , line.rstrip())
                 for line in fa)
    output.add((toEpoch(mat1.group(1)) , line1.rstrip()))
    fa.close()


    fb = open(c_file)
    while True:
        line1 = fb.readline()
        mat1 = pat.search(line1)
        if mat1:  break
    for line in fb:
        output.add((toEpoch(pat.search(line).group(1)) , line.rstrip()))
    output.add((toEpoch(mat1.group(1)) , line1.rstrip()))
    fb.close()

    output = list(output)
    output.sort()
    output[0:0] = head
    output[0:0] = [('',time.strftime('On %d/%m/%y %H:%M:%S'))]

    fm = open(m_file,'w')
    fm.writelines( line+'\n' for t,line in output)
    fm.close()



te = time.clock()
sorting_merge('ytr.txt','tatay.txt','merged.file.txt')
print time.clock()-te

Note that this code put a heading in the merged file

.

EDIT

Aaaaaah... I got it... :-))

Execution's time divided by 3 !

#!/usr/bin/env python

import os, time, sys, re
from sets import Set as sett

def sorting_merge(o_file , c_file, m_file ):

    pat = re.compile('[0123]\d/[01]\d/\d{2} [012]\d:[0-6]\d:[0-6]\d'
                     '(?=[ \t]+[0123]\d/[01]\d/\d{2} [012]\d:[0-6]\d:[0-6]\d)') 

    def kl(line,pat = pat):
        return time.mktime(time.strptime((pat.search(line).group()),'%d/%m/%y %H:%M:%S'))

    fa = open(o_file)
    head = []
    fa.readline()
    while True:
        line1 = fa.readline()
        mat1 = pat.search(line1)
        if not mat1:
            head.append(line1.rstrip())
        else:
            break
    output = sett(line.rstrip() for line in fa)
    output.add(line1.rstrip())
    fa.close()

    fb = open(c_file)
    while True:
        line1 = fb.readline()
        mat1 = pat.search(line1)
        if mat1:  break
    for line in fb:
        output.add(line.rstrip())
    output.add(line1.rstrip())
    fb.close()

    output = list(output)
    output.sort(key=kl)
    output[0:0] = [time.strftime('On %d/%m/%y %H:%M:%S')] + head

    fm = open(m_file,'w')
    fm.writelines( line+'\n' for line in output)
    fm.close()

te = time.clock()
sorting_merge('ytre.txt','tataye.txt','merged.file.txt')
print time.clock()-te

Answer 4

Last codes, I hope.

Because I found a killer code.

First , I created two files "xxA.txt" and "yyB.txt" of 30 lines having 30000 lines as

430559  group_atlas.atlas084    12  181 4       04/03/10 01:38:02   02/03/11 22:05:42
430502  group_atlas.atlas084    12  181 4       23/01/10 21:45:05   02/03/11 22:05:42
430544  group_atlas.atlas084    12  181 4       17/06/11 12:58:10   02/03/11 22:05:42
430566  group_atlas.atlas084    12  181 4       25/03/10 23:55:22   02/03/11 22:05:42

with the following code:

create AB.py

from random import choice

n = tuple( str(x) for x in xrange(500,600))
days = ('01','02','03','04','05','06','07','08','09','10','11','12','13','14','15','16',
        '17','18','19','20','21','22','23','24','25','26','27','28')
# not '29','30,'31' to avoid problems with strptime() on last days of february
months = days[0:12]
hours = days[0:23]
ms = ['00','01','02','03','04','05','06','07','09'] + [str(x) for x in xrange(10,60)]

repeat = 30000

with open('xxA.txt','w') as f:
    # 430794  group_atlas.atlas084    12  181 4     02/03/11 22:02:37   02/03/11 22:05:42
    ch = ('On 23/03/11 00:40:03\n'
          'JobID   Group.User          Ctime   Wtime   Status  QDate               CDate\n'
          '===================================================================================\n')
    f.write(ch)
    for i in xrange(repeat):
        line  = '430%s  group_atlas.atlas084    12  181 4   \t%s/%s/%s %s:%s:%s\t02/03/11 22:05:42\n' %\
                (choice(n),
                 choice(days),choice(months),choice(('10','11')),
                 choice(hours),choice(ms),choice(ms))
        f.write(line)


with open('yyB.txt','w') as f:
    # 430794  group_atlas.atlas084    12  181 4     02/03/11 22:02:37   02/03/11 22:05:42
    ch = ('On 25/03/11 13:45:24\n'
          'JobID   Group.User          Ctime   Wtime   Status  QDate               CDate\n'
          '===================================================================================\n')
    f.write(ch)
    for i in xrange(repeat):
        line  = '430%s  group_atlas.atlas084    12  181 4   \t%s/%s/%s %s:%s:%s\t02/03/11 22:05:42\n' %\
                (choice(n),
                 choice(days),choice(months),choice(('10','11')),
                 choice(hours),choice(ms),choice(ms))
        f.write(line)

with open('xxA.txt') as g:
    print 'readlines of xxA.txt :',len(g.readlines())
    g.seek(0,0)
    print 'set of xxA.txt :',len(set(g))

with open('yyB.txt') as g:
    print 'readlines of yyB.txt :',len(g.readlines())
    g.seek(0,0)
    print 'set of yyB.txt :',len(set(g))

Then I ran these 3 programs:

"merging regex.py"

#!/usr/bin/env python

from time import clock,mktime,strptime,strftime
from sets import Set
import re

infunc = []

def sorting_merge(o_file, c_file, m_file ):
    infunc.append(clock()) #infunc[0]
    pat = re.compile('([0123]\d/[01]\d/\d{2} [012]\d:[0-6]\d:[0-6]\d)')
    output = Set()

    def rmHead(filename, a_set):
        f_n = open(filename, 'r')
        f_n.readline()
        head = []
        for line in f_n:
            head.append(line) # line of the header
            if line.strip('= \r\n')=='':  break
        for line in f_n:
            a_set.add(line.rstrip())
        f_n.close()
        return head

    infunc.append(clock()) #infunc[1]
    head = rmHead(o_file, output)
    infunc.append(clock()) #infunc[2]
    head = rmHead(c_file, output)
    infunc.append(clock()) #infunc[3]
    if '' in output:  output.remove('')

    infunc.append(clock()) #infunc[4]
    output = [ (mktime(strptime(pat.search(line).group(),'%d/%m/%y %H:%M:%S')),line)
               for line in output ]
    infunc.append(clock()) #infunc[5]
    output.sort()
    infunc.append(clock()) #infunc[6]

    fm = open(m_file,'w')
    fm.write(strftime('On %d/%m/%y %H:%M:%S\n')+(''.join(head)))
    for t,line in output:
        fm.write(line + '\n')
    fm.close()
    infunc.append(clock()) #infunc[7]



c_f = "xxA.txt"
o_f = "yyB.txt"

t1 = clock()
sorting_merge(o_f, c_f, 'zz_mergedr.txt')
t2 = clock()
print 'merging regex'
print 'total time of execution :',t2-t1
print '              launching :',infunc[1] - t1
print '            preparation :',infunc[1] - infunc[0]
print '    reading of 1st file :',infunc[2] - infunc[1]
print '    reading of 2nd file :',infunc[3] - infunc[2]
print '      output.remove(\'\') :',infunc[4] - infunc[3]
print 'creation of list output :',infunc[5] - infunc[4]
print '      sorting of output :',infunc[6] - infunc[5]
print 'writing of merging file :',infunc[7] - infunc[6]
print 'closing of the function :',t2-infunc[7]

"merging split.py"

#!/usr/bin/env python

from time import clock,mktime,strptime,strftime
from sets import Set

infunc = []

def sorting_merge(o_file, c_file, m_file ):
    infunc.append(clock()) #infunc[0]
    output = Set()

    def rmHead(filename, a_set):
        f_n = open(filename, 'r')
        f_n.readline()
        head = []
        for line in f_n:
            head.append(line) # line of the header
            if line.strip('= \r\n')=='':  break
        for line in f_n:
            a_set.add(line.rstrip())
        f_n.close()
        return head

    infunc.append(clock()) #infunc[1]
    head = rmHead(o_file, output)
    infunc.append(clock()) #infunc[2]
    head = rmHead(c_file, output)
    infunc.append(clock()) #infunc[3]
    if '' in output:  output.remove('')

    infunc.append(clock()) #infunc[4]
    output = [ (mktime(strptime(line.split('\t')[-2],'%d/%m/%y %H:%M:%S')),line)
               for line in output ]
    infunc.append(clock()) #infunc[5]
    output.sort()
    infunc.append(clock()) #infunc[6]

    fm = open(m_file,'w')
    fm.write(strftime('On %d/%m/%y %H:%M:%S\n')+(''.join(head)))
    for t,line in output:
        fm.write(line + '\n')
    fm.close()
    infunc.append(clock()) #infunc[7]



c_f = "xxA.txt"
o_f = "yyB.txt"

t1 = clock()
sorting_merge(o_f, c_f, 'zz_mergeds.txt')
t2 = clock()
print 'merging split'
print 'total time of execution :',t2-t1
print '              launching :',infunc[1] - t1
print '            preparation :',infunc[1] - infunc[0]
print '    reading of 1st file :',infunc[2] - infunc[1]
print '    reading of 2nd file :',infunc[3] - infunc[2]
print '      output.remove(\'\') :',infunc[4] - infunc[3]
print 'creation of list output :',infunc[5] - infunc[4]
print '      sorting of output :',infunc[6] - infunc[5]
print 'writing of merging file :',infunc[7] - infunc[6]
print 'closing of the function :',t2-infunc[7]

"merging killer"

#!/usr/bin/env python

from time import clock,strftime
from sets import Set
import re

infunc = []

def sorting_merge(o_file, c_file, m_file ):
    infunc.append(clock()) #infunc[0]
    patk = re.compile('([0123]\d)/([01]\d)/(\d{2}) ([012]\d:[0-6]\d:[0-6]\d)')
    output = Set()

    def rmHead(filename, a_set):
        f_n = open(filename, 'r')
        f_n.readline()
        head = []
        for line in f_n:
            head.append(line) # line of the header
            if line.strip('= \r\n')=='':  break
        for line in f_n:
            a_set.add(line.rstrip())
        f_n.close()
        return head

    infunc.append(clock()) #infunc[1]
    head = rmHead(o_file, output)
    infunc.append(clock()) #infunc[2]
    head = rmHead(c_file, output)
    infunc.append(clock()) #infunc[3]
    if '' in output:  output.remove('')

    infunc.append(clock()) #infunc[4]
    output = [ (patk.search(line).group(3,2,1,4),line)for line in output ]
    infunc.append(clock()) #infunc[5]
    output.sort()
    infunc.append(clock()) #infunc[6]

    fm = open(m_file,'w')
    fm.write(strftime('On %d/%m/%y %H:%M:%S\n')+(''.join(head)))
    for t,line in output:
        fm.write(line + '\n')
    fm.close()
    infunc.append(clock()) #infunc[7]



c_f = "xxA.txt"
o_f = "yyB.txt"

t1 = clock()
sorting_merge(o_f, c_f, 'zz_mergedk.txt')
t2 = clock()
print 'merging killer'
print 'total time of execution :',t2-t1
print '              launching :',infunc[1] - t1
print '            preparation :',infunc[1] - infunc[0]
print '    reading of 1st file :',infunc[2] - infunc[1]
print '    reading of 2nd file :',infunc[3] - infunc[2]
print '      output.remove(\'\') :',infunc[4] - infunc[3]
print 'creation of list output :',infunc[5] - infunc[4]
print '      sorting of output :',infunc[6] - infunc[5]
print 'writing of merging file :',infunc[7] - infunc[6]
print 'closing of the function :',t2-infunc[7]

results

merging regex
total time of execution : 14.2816595405
              launching : 0.00169211450059
            preparation : 0.00168093989599
    reading of 1st file : 0.163582242995
    reading of 2nd file : 0.141301478261
      output.remove('') : 2.37460347614e-05
     creation of output : 13.4460212122
      sorting of output : 0.216363532237
writing of merging file : 0.232923737514
closing of the function : 0.0797514767938

merging split
total time of execution : 13.7824474898
              launching : 4.10666718815e-05
            preparation : 2.70984161395e-05
    reading of 1st file : 0.154349784679
    reading of 2nd file : 0.136050810927
      output.remove('') : 2.06730184981e-05
     creation of output : 12.9691854691
      sorting of output : 0.218704332534
writing of merging file : 0.225259076223
closing of the function : 0.0788362766776

merging killer
total time of execution : 2.14315311024
              launching : 0.00206199391263
            preparation : 0.00205026057781
    reading of 1st file : 0.158711791582
    reading of 2nd file : 0.138976601775
      output.remove('') : 2.37460347614e-05
     creation of output : 0.621466415424
      sorting of output : 0.823161602941
writing of merging file : 0.227701565422
closing of the function : 0.171049393149

During killer program, sorting output takes 4 times longer , but time of creation of output as a list is divided by 21 ! Then globaly, the execution's time is reduced at least by 85 %.