What is the difference between incremental and one-shot compression?

Question

I am trying to use the bz2 and/or lzma packages in python. I am trying to compress a database dump in csv format and then put it to a zip file. I got it to work with one-shot compression with both the packages.

Code for which looks like this:

with ZipFile('something.zip', 'w') as zf:
    content = bz2.compress(bytes(csv_string, 'UTF-8'))  # also with lzma
    zf.writestr(
        'something.csv' + '.bz2',
        content,
        compress_type=ZIP_DEFLATED
    )

When I try to use incremental compression then it creates a .zip file which when I try to extract keeps giving some archive file recursively.

Code for which looks like this:

with ZipFile('something.zip', 'w') as zf:
    compressor = bz2.BZ2Compressor()
    content = compressor.compress(bytes(csv_string, 'UTF-8'))  # also with lzma
    zf.writestr(
        'something.csv' + '.bz2',
        content,
        compress_type=ZIP_DEFLATED
    )
    compressor.flush()

I went through the documentation and also look for information about the compression techniques, and there seems to be no comprehensive information about what one-shot and incremental compression are.

Answer 1

The difference between one-shot and incremental is that with one-shot mode you need to have the entire data in memory; if you are compressing a 100 gigabyte file, you ought to have loads of RAM.

With the incremental encoder your code can feed the compressor 1 megabyte or 1 kilobyte at a time and write whatever data results, into a file as soon as it is available. Another benefit is that an incremental compressor you can use to stream data - you can start writing compressed data before all uncompressed data is available!

---

Your second code is incorrect and it will cause you to lose your data. The flush may return more data that you need to save as well. Here I am compressing a string of 1000 'a' characters in Python 3; the result from compress is an empty string; the actual compressed data is returned from flush.

>>> c = bz2.BZ2Compressor()
>>> c.compress(b'a' * 1000)
b''
>>> c.flush()
b'BZh91AY&SYI\xdcOc\x00\x00\x01\x81\x01\xa0\x00\x00\x80\x00\x08 \x00 
\xaamA\x98\xba\x83\xc5\xdc\x91N\x14$\x12w\x13\xd8\xc0'

Thus your second code should be:

compressor = bz2.BZ2Compressor()
content = compressor.compress(bytes(csv_string, 'UTF-8'))  # also with lzma
content += compressor.flush()    

But actually you're still doing the one-shot compression, in a very complicated manner.