Note that cwb-python is not actively maintained. The most recent PyPI-version is 0.2.2 from June 20, 2017. The code is identical to the 0.2.1-branch.

If you are starting a new project using Python and the Corpus Workbench, consider using cwb-ccc, which provides similar (and more) functionality.

CWB Python

This is a Python wrapper to the low-level API of CQP which allows you to access CWB corpora in the same way as Perl's CWB::CL.

Make sure installed CWB in the standard location (/usr/local tree). If you installed CWB in a non-standard location, point the setup in the right direction with e.g.

export CWB_DIR=/usr/local/cwb-3.4.10

To install the module, use setup.py:

sudo python setup.py install

Cython Compilation

The module ships with the generated .c files, so you do not need Cython for the installation.

If you want to re-generate the .c files, you need Cython:

pip install Cython

You can then compile the Cython code to C via

cython -2 cwb_python/CWB/CL.pyx

C Compilation

You can compile the C code via

pipenv run python setup.py build_ext --inplace

or just build (see above) or create a source distribution

pipenv run python setup.py sdist

Usage

To give you an idea how to use the library, see the following sample:

from cwb.cl import Corpus

# open the corpus
corpus=Corpus('TUEPP')
# get sentences and words
sentences=corpus.attribute('s','s')
words=corpus.attribute('word','p')
postags=corpus.attribute('pos','p')
# retrieve offsets of the 1235th sentence (0-based)
s_1234=sentences[1234]

for w,p in zip(words[s_1234[0]:s_1234[1]+1],postags[s_1234[0]:s_1234[1]+1]):
    print "%s/%s"%(w,p)

Tests

In order to test the CWB.CL module's correct installation independently of any CQP corpora, you can do a

python -m doctest tests/idlist.txt

which should terminate with no output when everything is well.