zc.zlibstorage 1.2.0

Usage

The primary storage is zc.zlibstorage.ZlibStorage. It is used as a wrapper around a lower-level storage. From Python, it is constructed by passing another storage, as in:

import ZODB.FileStorage, zc.zlibstorage

storage = zc.zlibstorage.ZlibStorage(
    ZODB.FileStorage.FileStorage('data.fs'))

When using a ZODB configuration file, the zlibstorage tag is used:

%import zc.zlibstorage

<zodb>
  <zlibstorage>
    <filestorage>
      path data.fs
    </filestorage>
  </zlibstorage>
</zodb>

Note the %import used to load the definition of the zlibstorage tag.

Use with ZEO

When used with a ZEO ClientStorage, you’ll need to use a server zlib storage on the storage server. This is necessary so that server operations that need to get at uncompressed record data can do so. This is accomplished using the serverzlibstorage tag in your ZEO server configuration file:

%import zc.zlibstorage

<zeo>
  address 8100
</zeo>

<serverzlibstorage>
  <filestorage>
    path data.fs
  </filestorage>
</serverzlibstorage>

Applying compression on the client this way is attractive because, in addition to reducing the size of stored database records on the server, you also reduce the size of records sent from the server to the client and the size of records stored in the client’s ZEO cache.

Decompressing only

By default, records are compressed when written to the storage and uncompressed when read from the storage. A compress option can be used to disable compression of records but still uncompress compressed records if they are encountered. Here’s an example from in Python:

import ZODB.FileStorage, zc.zlibstorage

storage = zc.zlibstorage.ZlibStorage(
    ZODB.FileStorage.FileStorage('data.fs'),
    compress=False)

and using the configurationb syntax:

%import zc.zlibstorage

<zodb>
  <zlibstorage>
    compress false
    <filestorage>
      path data.fs
    </filestorage>
  </zlibstorage>
</zodb>

This option is useful when deploying the storage when there are multiple clients. If you don’t want to update all of the clients at once, you can gradually update all of the clients with a zlib storage that doesn’t do compression, but recognizes compressed records. Then, in a second phase, you can update the clients to compress records, at which point, all of the clients will be able to read the compressed records produced.

Compressing entire databases

One way to compress all of the records in a database is to copy data from an uncompressed database to a compressed one, as in:

import ZODB.FileStorage, zc.zlibstorage

orig = ZODB.FileStorage.FileStorage('data.fs')
new = zc.zlibstorage.ZlibStorage(
    ZODB.FileStorage.FileStorage('data.fs-copy'))
new.copyTransactionsFrom(orig)

orig.close()
new.close()

Record prefix

Compressed records have a prefix of “.z”. This allows a database to have a mix of compressed and uncompressed records.

Stand-alone Compression and decompression functions

In anticipation of wanting to plug the compression and decompression logic into other tools without creating storages, the functions used to compress and uncompress data records are available as zc.zlibstorage module-level functions:

compress(data)

Compress the given data if:

• it is a string more than 20 characters in length,

• it doesn’t start with the compressed-record marker, b'.z', and

• the compressed size is less the original.

The compressed (or original) data are returned.

decompress(data)

Decompress the data if it is compressed.

The decompressed (or original) data are returned.

1.2.0 (2017-01-20)

• Add support for Python 3.6 and PyPy.

• Test with both ZODB/ZEO 4 and ZODB/ZEO 5. Note that ServerZlibStorage cannot be used in a ZODB 5 Connection (e.g., client-side, which wouldn’t make sense :-]). (https://github.com/zopefoundation/zc.zlibstorage/issues/5).

• Close the underlying iterator used by the iterator wrapper when it is closed. (https://github.com/zopefoundation/zc.zlibstorage/issues/4)

1.1.0 (2016-08-03)

• Fixed an incompatibility with ZODB5. The previously optional and ignored version argument to the database invalidate method is now disallowed.

• Drop Python 2.6, 3.2, and 3.3 support. Added Python 3.4 and 3.5 support.

1.0.0 (2015-11-11)

• Python 3 support contributed by Christian Tismer.

0.1.1 (2010-05-26)

• Fixed a packaging bug.

0.1.0 (2010-05-20)

Initial release