wormtable 0.1.0a1

Write-once read-many data sets using Berkeley DB.

Wormtable is a write-once read-many table for large scale datasets. It provides Python programmers with a simple and efficient method of storing, processing and searching datasets of essentially unlimited size. A wormtable consists of a set of rows, each of which contains values belonging to a fixed number of columns. Rows are encoded in a custom binary format, designed to be flexible, compact and portable. These rows are then stored on disc using Berkeley DB, a highly respected embedded database toolkit. Wormtable also supports efficient searching and retrieval of rows with particular values through the use of indexes.

The Variant Call Format (VCF) is supported directly by wormtable through a command line conversion program, vcf2wt. There is also a command line utility wtadmin to manage wormtables, including the ability to dump values and add, remove and view indexes.

Documentation

Here’s a quick example for the impatient:

import wormtable as wt
with wt.open_table("example.wt") as t:
    c = t.cursor("POS", "ALT", "REF")
    for pos, alt, ref in c:
        print(pos, ":", alt, ":", ref)

Full documentation for wormtable is available at http://jeromekelleher.github.com/wormtable.

Installation

Wormtable requires Berkeley DB, which is available for all major platforms. Any recent version of Berkeley DB should work, but the various versions have not been tested extensively. Development and testing have been carried out primarily on the DB 4.x series.

Once DB has been installed (see below) we can build the wormtable module using the standard Python methods. For example, using pip we have

$ sudo pip install wormtable

Or, we can manually download the package, unpack it and then run:

$ python setup.py build
$ sudo python setup.py install

Most of the time this will compile and install the module without difficulty.

It is also possible to download the latest development version of wormtable from github.

Installing Berkeley DB

Linux

Wormtable is primarily developed on Linux and should work very well on any modern Linux distribution. Installing Berkeley DB is very easy on Linux distributions.

On Debian/Ubuntu use:

$ sudo apt-get install libdb-dev

and on Red Hat/Fedora use:

# yum install db4-devel

Other distributions and package managers should provide a similarly easy option to install the DB development files.

Mac OS X

TODO: document installation on a mac. MacPorts/Homebrew/installer package?

Unix

Most Unix systems provide Berkeley DB packages. For example, on FreeBSD we have:

# pkg_add -r db48

If necessary, Berkeley DB can be built from source and installed manually quite easily.

Potential problems

On platforms that Berkeley DB is not available as part of the native packaging system (or DB was installed locally because of non-root access) there can be issues with finding the correct headers and libraries when compiling wormtable. For example, on FreeBSD we get something like this:

$ python setup.py build
... [Messages cut for brevity] ...
_wormtablemodule.c:3727: error: 'DB_NEXT_NODUP' undeclared (first use in this function)
_wormtablemodule.c:3733: error: 'DB_NOTFOUND' undeclared (first use in this function)
_wormtablemodule.c:3739: error: 'DistinctValueIterator' has no member named 'cursor'
_wormtablemodule.c:3739: error: 'DistinctValueIterator' has no member named 'cursor'
_wormtablemodule.c:3740: error: 'DistinctValueIterator' has no member named 'cursor'
error: command 'cc' failed with exit status 1

To remedy this we must set the LDFLAGS and CFLAGS environment variables to their correct values. Unfortunately there is no simple method to do this and some knowledge of where your system keeps headers and libraries is needed. On FreeBSD (after installing the db48 package as above) we might use:

$ CFLAGS=-I/usr/local/include/db48 LDFLAGS=-L/usr/local/lib/db48 python setup.py build

Test suite

Wormtable has an extensive suite of tests to ensure that data is stored correctly. It is a good idea to run these immediately after installation:

$ python tests.py

Tested platforms

Wormtable is higly portable, and has been successfully built and tested on the following platforms:

Operating system	Platform	Python	Compiler
Ubuntu 13.04	x86-64	2.7.4	gcc 4.7.3
Ubuntu 13.04	x86-64	3.3.1	gcc 4.7.3
Ubuntu 13.04	x86-64	2.7.4	clang 3.2.1
Debian squeeze	x86-64	2.6.6	gcc 4.4.5
Debian squeeze	x86-64	3.1.3	gcc 4.4.5
Debian squeeze	x86-64	3.1.3	clang 1.1
Debian squeeze	ppc64	2.6.6	gcc 4.4.5
Debian squeeze	ppc64	3.1.3	gcc 4.4.5
Debian squeeze	ppc64	3.1.3	clang 1.1
Debian wheezy	armv6l	2.7.3	gcc 4.6.3
Fedora 17	i386	2.7.3	gcc 4.7.2
Fedora 17	i386	3.2.3	gcc 4.7.2
FreeBSD 9.0	i386	3.2.2	gcc 4.2.2
FreeBSD 9.0	i386	2.7.2	gcc 4.2.2
FreeBSD 9.0	i386	3.1.4	clang 3.0