def test_cwb_scan_corpus_subcorpora(germaparl):
corpus = get_corpus(germaparl)
cqp = corpus.start_cqp()
# get a DumpFrame
cqp.nqr_from_query('[lemma="Helmut"]', name='Tmp')
# run cwb-scan-corpus on dump
with NamedTemporaryFile(mode="wt") as f:
cqp.Exec('dump Tmp > "%s"' % f.name)
df1, R1 = cwb_scan_corpus(germaparl['corpus_name'],
germaparl['registry_path'], f.name)
# activate a sub-corpus
cqp.nqr_from_query(query='[lemma="und"] expand to s', name='und')
cqp.nqr_activate(corpus.corpus_name, 'und')
# run same query on sub-corpus
cqp.nqr_from_query('[lemma="Helmut"]', name='Tmp')
# run cwb-scan-corpus on dump
with NamedTemporaryFile(mode="wt") as f:
cqp.Exec('dump Tmp > "%s"' % f.name)
df2, R2 = cwb_scan_corpus(germaparl['corpus_name'],
germaparl['registry_path'], f.name)
# check that results are different
assert (sum(df2['freq']) != sum(df1['freq']))
cqp.__kill__()
def test_cwb_scan_corpus_marginal(germaparl):
df1, R1 = cwb_scan_corpus(germaparl['corpus_name'],
germaparl['registry_path'])
df2, R2 = cwb_scan_corpus(germaparl['corpus_name'],
germaparl['registry_path'],
p_atts=['lemma', 'pos'])
assert (df1.index.name == 'item')
assert (df2.index.name == 'item')
assert (list(df1.columns) == ['freq', 'word'])
assert (list(df2.columns) == ['freq', 'lemma', 'pos'])
def test_cwb_scan_corpus_marginal(germaparl):
df1 = cwb_scan_corpus(None, germaparl['corpus_name'],
germaparl['registry_path'])
df2 = cwb_scan_corpus(None,
germaparl['corpus_name'],
germaparl['registry_path'],
p_atts=['lemma', 'pos'])
assert (isinstance(df1.index, pd.MultiIndex))
assert (isinstance(df2.index, pd.MultiIndex))
assert (df1.index.names == ['word'])
assert (df2.index.names == ['lemma', 'pos'])
assert (df1.columns == df2.columns)
def test_cwb_scan_corpus(germaparl):
from tempfile import NamedTemporaryFile
corpus = Corpus(germaparl['corpus_name'],
registry_path=germaparl['registry_path'])
cqp = corpus.start_cqp()
cqp.nqr_from_query('[lemma="Helmut"]', name='tmp')
with NamedTemporaryFile(mode="wt") as f:
cqp.Exec('dump tmp > "%s"' % f.name)
df1 = cwb_scan_corpus(f.name, germaparl['corpus_name'],
germaparl['registry_path'])
cqp.nqr_from_query(query='[lemma="und"] expand to s', name='und')
cqp.nqr_activate(corpus.corpus_name, 'und')
cqp.nqr_from_query('[lemma="Kohl"]', name='tmp')
with NamedTemporaryFile(mode="wt") as f:
cqp.Exec('dump tmp > "%s"' % f.name)
df2 = cwb_scan_corpus(f.name, germaparl['corpus_name'],
germaparl['registry_path'])
cqp.__kill__()
assert (sum(df2['freq']) != sum(df1['freq']))