Read a gzip file from inside a tar file

[rohisingh]

I have a tar file. The content of the file are as following.

rohits@sandman 12-08-04 $ tar tvf 20041208.tar
drwxr-xr-x root/root 0 2004-12-08 21:39:19 20041208/
-rw-r--r-- root/root 1576 2004-12-08 21:39:19 20041208/README
drwxr-xr-x root/root 0 2004-12-08 21:27:31
20041208/snapshot_01/
-rw-r--r-- was/was 103010606 2004-12-08 16:37:38
20041208/snapshot_01/tpv-2004 1208-1350.xml.gz


What is the best method to read the content of the
tpv-20041208-1350.xml.gz?

I want to do the following with minimum code
1) read above tar file
2) find the gzip file
3) read the content of this file
4) perform operations on content
5) continue

I tried various combination of following code but it does not work as
intended

fileName = sys.argv[1]
print "File Name is ", fileName
tar = tarfile.open(fileName, "r:")
for tarinfo in tar:
if tarinfo.isreg():
print tarinfo.name
if tarinfo.name.find("tpv") != -1:
#read the gzip file
print "\thttp plugin file"
fileLike = tar.extractfile(tarinfo)
fileText = fileLike.read()
stringio = StringIO.StringIO(fileText)
fileRead = gzip.GzipFile(stringio)
for aLine in fileRead:
print aLine

 

[Rohit]

if I change fileText = fileLike.read() to fileText =
fileLike.readLines().

It works for a while before it gets killed of out of memory.

These are huge files. My goal is to analyze the content of the gzip
file in the tar file without having to un gzip. If that is possible.

 

[Craig Ringer]

if I change fileText = fileLike.read() to fileText =
fileLike.readLines().

It works for a while before it gets killed of out of memory.

These are huge files. My goal is to analyze the content of the gzip
file in the tar file without having to un gzip. If that is possible.

As far as I know, gzip is a stream compression algorithm that can't be
decompressed in small blocks. That is, I don't think you can seek 500k
into a 1MB file and decompress the next 100k.

I'd say you'll have to progressively read the file from the beginning,
processing and discarding as you go. It looks like a no-brainer to me -
see zlib.decompressobj.

Note that you _do_ have to ungzip it, you just don't have to store the
whole decompressed thing in memory / on disk at once. If you need to do
anything to it that does require the entire thing to be loaded (or
anything that means you have to seek around the file), I'd say you're
SOL.

 

[Fredrik Lundh]

As far as I know, gzip is a stream compression algorithm that can't be
decompressed in small blocks. That is, I don't think you can seek 500k
into a 1MB file and decompress the next 100k.
correct.

I'd say you'll have to progressively read the file from the beginning,
processing and discarding as you go. It looks like a no-brainer to me -
see zlib.decompressobj.

it can be a bit tricky to set things up properly, though. here's a piece
of code that uses Python's good old consumer interface to decode things
incrementally:

http://effbot.org/zone/consumer-gzip.htm

you can either use this as is; just create a "target consumer", wrap it in the
gzip consumer, and feed data to the gzip consumer in suitable pieces.

alternatively, hack it until it does what you want.

</F>