split large file by string/regex

[Martin Dieringer]

I am trying to split a file by a fixed string.
The file is too large to just read it into a string and split this.
I could probably use a lexer but there maybe anything more simple?
thanks
m.

 

[Steve Holden]

I am trying to split a file by a fixed string.
The file is too large to just read it into a string and split this.
I could probably use a lexer but there maybe anything more simple?
thanks
m.

Depends on your definition of "simple", I suppose. The problem with
*not* using a lexer is that you'd have to examine the file in a sequence
of overlapping chunks to make sure that a regex could pick up all
matches. For me that would be more complex than using a lexer, given the
excellent range of modules such as SPARK and PLY, to mention but two.

regards
Steve

 

[Jason Rennie]

I am trying to split a file by a fixed string.
The file is too large to just read it into a string and split this.
I could probably use a lexer but there maybe anything more simple?

If the pattern is contained within a single line, do something like this:

import re
myre = re.compile(r'foo')
fh = open(f)
fh1 = open(f1,'w')
s = fh.readline()
while not myre.search(s):
fh1.write(s)
s = fh.readline()
fh1.close()
fh2.open(f1,'w')
while fh
fh2.write(s)
s = fh.readline()
fh2.close()
fh.close()

I'm doing this off the top of my head, so this code almost certainly
has bugs. Hopefully its enough to get you started... Note that only
one line is held in memory at any point in time. Oh, if there's a
chance that the pattern does not appear in the file, you'll need to
check for eof in the first while loop.

Jason

 

[Diez B. Roggisch]

Depends on your definition of "simple", I suppose. The problem with

*not* using a lexer is that you'd have to examine the file in a sequence
of overlapping chunks to make sure that a regex could pick up all
matches. For me that would be more complex than using a lexer, given the
excellent range of modules such as SPARK and PLY, to mention but two.

At least spark operates on whole strings if used as lexer/tokenizer - you
can of course feed it a lazy sequence of tokens by using a generator - but
that's up to you.

 

[Martin Dieringer]

Depends on your definition of "simple", I suppose. The problem with
*not* using a lexer is that you'd have to examine the file in a
sequence of overlapping chunks to make sure that a regex could pick up
all matches. For me that would be more complex than using a lexer,
given the excellent range of modules such as SPARK and PLY, to mention
but two.

yes lexing would be the simplest, but PLY also can't read from streams
and it looks to me (from the examples) as if it's the same with SPARK.
I wonder why something like this is not in any lib.
Is there any known lexer that can do this?
I don't have to parse, just write the junks to separate files.
I really hate doing that sequence thing...

m.

 

[Martin Dieringer]

If the pattern is contained within a single line, do something like this:

Hmm it's binary data, I can't tell how long lines would be. OTOH a
line would certainly contain the pattern as it has no \n in it... and
the lines probably wouldn't be too large for memory...

m.

 

[Bengt Richter]

Hmm it's binary data, I can't tell how long lines would be. OTOH a
line would certainly contain the pattern as it has no \n in it... and
the lines probably wouldn't be too large for memory...

m.

Do you want to keep the splitting string? I.e., if you split with xxx
from '1231xxx45646xxx45646xxx78' do you want the long-file equivalent of
['1231', '45646', '45646', '78']

or (I chose this for below)
['1231', 'xxx', '45646', 'xxx', '45646', 'xxx', '78']

or maybe

['1231xxx', '45646xxx', '45646xxx', '78']

??

Anyway, I'd use a generator to iterate through the file and look for the delimiter.
This is case-sensitive, BTW (practically untested ;-):

--< splitfile.py >----------------------------------------------
def splitfile(path, splitstr, chunksize=1024*64): # try a megabyte?
splen = len(splitstr)
chunks = iter(lambda f=open(path,'rb'):f.read(chunksize), '')
buf = ''
for chunk in chunks:
buf += chunk
start = end = 0
while end>=0 and len(buf)>=splen:
start, end = end, buf.find(splitstr, end)
if end>=0:
yield buf[start:end] #not including splitstr
yield splitstr # == buf[end:end+splen] # splitstr
end += splen
else:
buf = buf[start:]
break

yield buf

def test(*args):
for chunk in splitfile(*args):
print repr(chunk)

if __name__ == '__main__':
import sys
args = sys.argv[1:]
try:
if len(args)==3: args[2]=int(args[2])
except Exception:
raise SystemExit, 'Usage: python splitfile.py path splitstr [chunksize=64k]'
test(*args)
----------------------------------------------------------------

Extent of testing follows
----------------------------------------
01234abc5678abc901234
567ab890abc
---------------------------------------- '01234'
'abc'
'5678'
'abc'
'901234\r\n567ab890'
'abc'
'\r\n' ''
'012'
'34abc5678abc9'
'012'
'34\r\n567ab890abc\r\n' Traceback (most recent call last):
File "<stdin>", line 1, in ?
StopIteration

(I put it in my ut package directory but you can put splitfile.py anywhere handy
and mod it to do what you need).

Regards,
Bengt Richter

 

[Denis S. Otkidach]

Depends on your definition of "simple", I suppose. The problem with
*not* using a lexer is that you'd have to examine the file in a sequence
of overlapping chunks to make sure that a regex could pick up all

re module works fine with mmap-ed file, so no need to read it into memory.

 

[Martin Dieringer]

re module works fine with mmap-ed file, so no need to read it into memory.

thank you, this is the solution!
Now I can mmap.find all locations and then read the chunks them via
file.seek and file.read

m.

 

[William Park]

Hmm it's binary data, I can't tell how long lines would be. OTOH a
line would certainly contain the pattern as it has no \n in it... and
the lines probably wouldn't be too large for memory...

man strings (-o option)

 

[Martin Dieringer]

man strings (-o option)

this doesn't make sense at all

m.

 

[Denis S. Otkidach]

thank you, this is the solution!
Now I can mmap.find all locations and then read the chunks them via
file.seek and file.read

mmap-ed files also support subscription and slicing. I guess
mmfile[start:stop] would more readable.

 

[Martin Dieringer]

thank you, this is the solution!
Now I can mmap.find all locations and then read the chunks them via
file.seek and file.read

mmap-ed files also support subscription and slicing. I guess
mmfile[start:stop] would more readable.

yes, even better

m.