So I'm playing with the Wikipedia dump file. It's an XML file that has been bzipped. I can write all the files to directories, but then when I want to do analysis, I have to reread all the files on the disk. This gives me random access, but it's slow. I have the ram to put the entire bzipped file into ram.
I can load the dump file just fine and read all the lines, but I cannot seek in it as it's gigantic. From what it seems, the bz2 library has to read and capture the offset before it can bring me there (and decompress it all, as the offset is in decompressed bytes).
Anyway, I'm trying to mmap the dump file (~9.5 gigs) and load it into bzip. I obviously want to test this on a bzip file before.
I want to map the mmap file to a BZ2File so I can seek through it (to get to a specific, uncompressed byte offset), but from what it seems, this is impossible without decompressing the entire mmap file (this would be well over 30 gigabytes).
Do I have any options?
Here's some code I wrote to test.
import bz2
import mmap
lines = '''This is my first line
This is the second
And the third
'''
with open("bz2TestFile", "wb") as f:
f.write(bz2.compress(lines))
with open("bz2TestFile", "rb") as f:
mapped = mmap.mmap(f.fileno(), 0, prot=mmap.PROT_READ)
print "Part of MMAPPED"
# This does not work until I hit a minimum length
# due to (I believe) the checksums in the bz2 algorithm
#
for x in range(len(mapped)+2):
line = mapped[0:x]
try:
print x
print bz2.decompress(line)
except:
pass
# I can decompress the entire mmapped file
print ":entire mmap file:"
print bz2.decompress(mapped)
# I can create a bz2File object from the file path
# Is there a way to map the mmap object to this function?
print ":BZ2 File readline:"
bzF = bz2.BZ2File("bz2TestFile")
# Seek to specific offset
bzF.seek(22)
# Read the data
print bzF.readline()
This all makes me wonder though, what is special about the bz2 file object that allows it to read a line after seeking? Does it have to read every line before it to get the checksums from the algorithm to work out correctly?
I found an answer! James Taylor wrote a couple scripts for seeking in BZ2 files, and his scripts are in the biopython module.
https://bitbucket.org/james_taylor/bx-python/overview
These work pretty well, although they do not allow for seeking to arbitrary byte offsets in the BZ2 file, his scripts read out blocks of BZ2 data and allow seeking based on blocks.
In particular, see bx-python / wiki / IO / SeekingInBzip2Files
Related
In Bash, when you gzip a file, the original is not retained, whereas in Python, you could use the gzip library like this (as shown here in the "Examples of Usage" section):
import gzip
import shutil
with open('/home/joe/file.txt', 'rb') as f_in:
with gzip.open('/home/joe/file.txt.gz', 'wb') as f_out:
shutil.copyfileobj(f_in, f_out)
By default, this retains the original file. I couldn't find a way to not retain it while gzipping. Do I have to wait till gzip is done to delete the file?
If you are on a unix-like system, you can unlink the file after opening so that it is no longer found in the file system. But it will still take disk space until you close the now-anonymous file.
import gzip
import shutil
import os
with open('deleteme', 'rb') as f_in:
with gzip.open('deleteme.gz', 'wb') as f_out:
os.unlink('deleteme') # *after* we knew the gzip open worked!
shutil.copyfileobj(f_in, f_out)
As far as I know, this doesn't work on Windows. You need to do the remove after the zip process completes. You could change its name to something like "thefile.temporary" or even move it to a different directory (fast if the directory is the same file system, but copied if its a different one).
Considering that when GZip runs (in Bash, or anywhere else for that matter):
GZip requires the original data to perform the zipping action
GZip is designed to handle data of basically arbitrary size
Therefore: GZip isn't likely to be creating a temp file in memory, rather it is almost certainly deleting the original after the gzip is done anyway.
With those points in mind, an identical strategy for your code is to do the gzip, then delete the file.
Certainly deleting the file isn't onerous — there are several ways to do it — and of course you could package the whole thing in a procedure so as to never have to concern yourself with it again.
The code below (partially based on tdelaney's answer), will do the following:
read the file, compressing on the fly, and storing all the compressed data in memory
delete the input file
then write the compressed data
This is for the use case where you have a full filesystem, which prevents you from writing the compressed data at the same time that the uncompressed file exists on disk. To get around this problem, it is therefore necessary to store all the data in memory (unless you have access to external storage), but to minimise this memory cost as far as possible, only the compressed data is fully stored in memory, while the uncompressed data is read in chunks.
There is of course a risk of data loss if the program is interrupted between deleting the input file and completing writing the compressed data to disk.
There is also the possibility of failure if there is insufficient memory, but the input file would not be deleted in that case because the MemoryError would be raised before the os.unlink is reached.
It is worth noting that this does not specifically answer what the question asks for, namely deleting the input file while still reading from it. This is possible under unix-like OSes, but there is no practical advantage in doing this over the regular command-line gzip behaviour, because freeing the disk space still does not happen until the file is closed, so it sacrifices recoverability in the event of failure, without gaining any additional space to juggle data in exchange for that sacrifice. (There would still need to be disk space for the uncompressed and compressed data to coexist.)
import gzip
import shutil
import os
from io import BytesIO
filename = 'deleteme'
buf = BytesIO()
# compress into memory - don't store all the uncompressed data in memory
# but do store all the compressed data in memory
with open(filename, 'rb') as fin:
with gzip.open(buf, 'wb') as zbuf:
shutil.copyfileobj(fin, zbuf)
# sanity check for already compressed data
length = buf.tell()
if length > os.path.getsize(filename):
raise RuntimeError("data *grew* in size - refusing to delete input")
# delete input file and then write out the compressed data
buf.seek(0)
os.unlink(filename)
with open(filename + '.gz', 'wb') as fout:
shutil.copyfileobj(buf, fout)
I've worked with decompressing and reading files on the fly in memory with the bz2 library. However, i've read through the documentation and can't seem to just simply decompress the file to create a brand new file on the file system with the decompressed data without memory storage. Sure, you could read line by line using BZ2Decompressor then write that to a file, but that would be insanely slow. (Decompressing massive files, 50GB+). Is there some method or library I have overlooked to achieve the same functionality as the terminal command bz2 -d myfile.ext.bz2 in python without using a hacky solution involving a subprocess to call that terminal command?
Example why bz2 is so slow:
Decompressing that file via bz2 -d: 104seconds
Analytics on a decompressed file(just involves reading line by line): 183seconds
with open(file_src) as x:
for l in x:
Decompressing on the file and using analytics: Over 600 seconds (This time should be max 104+183)
if file_src.endswith(".bz2"):
bz_file = bz2.BZ2File(file_src)
for l in bz_file:
You could use the bz2.BZ2File object which provides a transparent file-like handle.
(edit: you seem to use that already, but don't use readlines() on a binary file, or on a text file because in your case the block size isn't big enough which explains why it's slow)
Then use shutil.copyfileobj to copy to the write handle of your output file (you can adjust block size if you can afford the memory)
import bz2,shutil
with bz2.BZ2File("file.bz2") as fr, open("output.bin","wb") as fw:
shutil.copyfileobj(fr,fw)
Even if the file is big, it doesn't take more memory than the block size. Adjust the block size like this:
shutil.copyfileobj(fr,fw,length = 1000000) # read by 1MB chunks
For smaller files that you can store in memory before you save to a file, you can use bz2.open to decompress the file and save it as an uncompressed new file.
import bz2
#decompress data
with bz2.open('compressed_file.bz2', 'rb') as f:
uncompressed_content = f.read()
#store decompressed file
with open('new_uncompressed_file.dat', 'wb') as f:
f.write(uncompressed_content)
f.close()
I have ~1GB *.tbz files. Inside each of those files there is a single ~9GB file. I just need to read the header of this file, the first 1024 bytes.
I want this to do this as fast as possible as I have hundreds of this 1GB files I want to process. It takes about 1m30s to extract.
I tried using full extraction:
tar = tarfile.open(fn, mode='r|bz2')
for item in tar:
tar.extract(item)
and tarfile.getmembers() but with no speed imprevement:
tar = tarfile.open(fn, mode='r|bz2')
for member in tar.getmembers():
f = tar.extractfile(member)
headerbytes = f.read(1024)
headerdict = parseHeader(headerbytes)
The getmembers() method is what's taking all the time there.
Is there any way I can to this?
I think you should use the standard library bz2 interface. .tbz is the file extension for tar files that are compressed with the -j option to specify a bzip2 format.
As #bbayles pointed out in the comments, you can open your file as a bz2.BZ2File and use seek and read:
read([size])
Read at most size uncompressed bytes, returned as a
string. If the size argument is negative or omitted, read until EOF is
reached.
seek(offset[, whence])
Move to new file position. Argument offset is a
byte count.
f = bz2.BZ2File(path)
f.seek(512)
headerbytes = f.read(1024)
You can then parse that with your functions.
headerdict = parseHeader(headerbytes)
If you're sure that every tar archive will contain only a single bz2 file, you can simply skip the first 512 bytes when first reading the tar file (NOT the bz2 file contained in it, of course), because the tar file format has a padded (fixed size) header, after which your "real" content is stored.
A simple
f.seek(512)
instead of looping over getmembers() should do the trick.
I have a giant file, let's call it one-csv-file.xz. It is an XZ-compressed CSV file.
How can I open and parse through the file without first decompressing it to disk? What if the file is, for example, 100 GB? Python cannot read all of that into memory at once, of course. Will it page or run out of memory?
You can iterate through an LZMAFile object
import lzma # python 3, try lzmaffi in python 2
with open('one-csv-file.xz') as compressed:
with lzma.LZMAFile(compressed) as uncompressed:
for line in uncompressed:
do_stuff_with(line)
You can decompress incrementally. See Compression using the LZMA Algorithm. You create an LZMADecompressor object, and then use the decompress method with successive chunks of the compressed data to get successive chunks of the uncompressed data.
I use the following simple Python script to compress a large text file (say, 10GB) on an EC2 m3.large instance. However, I always got a MemoryError:
import gzip
with open('test_large.csv', 'rb') as f_in:
with gzip.open('test_out.csv.gz', 'wb') as f_out:
f_out.writelines(f_in)
# or the following:
# for line in f_in:
# f_out.write(line)
The traceback I got is:
Traceback (most recent call last):
File "test.py", line 8, in <module>
f_out.writelines(f_in)
MemoryError
I have read some discussion about this issue, but still not quite clear how to handle this. Can someone give me a more understandable answer about how to deal with this problem?
The problem here has nothing to do with gzip, and everything to do with reading line by line from a 10GB file with no newlines in it:
As an additional note, the file I used to test the Python gzip functionality is generated by fallocate -l 10G bigfile_file.
That gives you a 10GB sparse file made entirely of 0 bytes. Meaning there are no newline bytes. Meaning the first line is 10GB long. Meaning it will take 10GB to read the first line. (Or possibly even 20 or 40GB, if you're using pre-3.3 Python and trying to read it as Unicode.)
If you want to copy binary data, don't copy line by line. Whether it's a normal file, a GzipFile that's decompressing for you on the fly, a socket.makefile(), or anything else, you will have the same problem.
The solution is to copy chunk by chunk. Or just use copyfileobj, which does that for you automatically.
import gzip
import shutil
with open('test_large.csv', 'rb') as f_in:
with gzip.open('test_out.csv.gz', 'wb') as f_out:
shutil.copyfileobj(f_in, f_out)
By default, copyfileobj uses a chunk size optimized to be often very good and never very bad. In this case, you might actually want a smaller size, or a larger one; it's hard to predict which a priori.* So, test it by using timeit with different bufsize arguments (say, powers of 4 from 1KB to 8MB) to copyfileobj. But the default 16KB will probably be good enough unless you're doing a lot of this.
* If the buffer size is too big, you may end up alternating long chunks of I/O and long chunks of processing. If it's too small, you may end up needing multiple reads to fill a single gzip block.
That's odd. I would expect this error if you tried to compress a large binary file that didn't contain many newlines, since such a file could contain a "line" that was too big for your RAM, but it shouldn't happen on a line-structured .csv file.
But anyway, it's not very efficient to compress files line by line. Even though the OS buffers disk I/O it's generally much faster to read and write larger blocks of data, eg 64 kB.
I have 2GB of RAM on this machine, and I just successfully used the program below to compress a 2.8GB tar archive.
#! /usr/bin/env python
import gzip
import sys
blocksize = 1 << 16 #64kB
def gzipfile(iname, oname, level):
with open(iname, 'rb') as f_in:
f_out = gzip.open(oname, 'wb', level)
while True:
block = f_in.read(blocksize)
if block == '':
break
f_out.write(block)
f_out.close()
return
def main():
if len(sys.argv) < 3:
print "gzip compress in_file to out_file"
print "Usage:\n%s in_file out_file [compression_level]" % sys.argv[0]
exit(1)
iname = sys.argv[1]
oname = sys.argv[2]
level = int(sys.argv[3]) if len(sys.argv) > 3 else 6
gzipfile(iname, oname, level)
if __name__ == '__main__':
main()
I'm running Python 2.6.6 and gzip.open() doesn't support with.
As Andrew Bay notes in the comments, if block == '': won't work correctly in Python 3, since block contains bytes, not a string, and an empty bytes object doesn't compare as equal to an empty text string. We could check the block length, or compare to b'' (which will also work in Python 2.6+), but the simple way is if not block:.
It is weird to get a memory error even when reading a file line by line. I suppose it is because you have very little available memory and very large lines. You should then use binary reads :
import gzip
#adapt size value : small values will take more time, high value could cause memory errors
size = 8096
with open('test_large.csv', 'rb') as f_in:
with gzip.open('test_out.csv.gz', 'wb') as f_out:
while True:
data = f_in.read(size)
if data == '' : break
f_out.write(data)