def syncdir(srcdir, destdir):
srcdir = normpath(srcdir)
destdir = normpath(destdir)
if not isdir(destdir):
sys.exit('destination directory "%s" does not exist' % destdir)
for root, dirs, files in os.walk(srcdir):
root = normpath(root)
prefix = os.path.commonprefix([root, srcdir])
root = root[len(prefix):]
if root.startswith('/'):
root = root[1:]
for rem in [ d for d in dirs if d.startswith('.') or d == 'SCCS']:
dirs.remove(rem)
for entry in dirs:
newdir = joinpath(destdir, root, entry)
if not isdir(newdir):
os.mkdir(newdir)
print 'mkdir', newdir
for i,d in enumerate(dirs):
if islink(joinpath(srcdir, root, d)):
dirs[i] = joinpath(d, '.')
for entry in files:
dest = normpath(joinpath(destdir, root, entry))
src = normpath(joinpath(srcdir, root, entry))
if not isfile(dest) or not filecmp(src, dest):
print 'copy %s %s' % (dest, src)
copy(src, dest)
def syncdir(srcdir, destdir):
srcdir = normpath(srcdir)
destdir = normpath(destdir)
if not isdir(destdir):
sys.exit('destination directory "%s" does not exist' % destdir)
for root, dirs, files in os.walk(srcdir):
root = normpath(root)
prefix = os.path.commonprefix([root, srcdir])
root = root[len(prefix):]
if root.startswith('/'):
root = root[1:]
for rem in [d for d in dirs if d.startswith('.') or d == 'SCCS']:
dirs.remove(rem)
for entry in dirs:
newdir = joinpath(destdir, root, entry)
if not isdir(newdir):
os.mkdir(newdir)
print 'mkdir', newdir
for i, d in enumerate(dirs):
if islink(joinpath(srcdir, root, d)):
dirs[i] = joinpath(d, '.')
for entry in files:
dest = normpath(joinpath(destdir, root, entry))
src = normpath(joinpath(srcdir, root, entry))
if not isfile(dest) or not filecmp(src, dest):
print 'copy %s %s' % (dest, src)
copy(src, dest)
def _binarycmp(filelist, onerror):
file0, file1 = filelist
try:
if filecmp(file0.path, file1.path, shallow=False):
dupdict = {True: filelist}
else:
dupdict = {}
errlist = []
except (IOError, OSError) as exc:
if onerror is not None:
onerror(exc, abspath(exc.filename))
dupdict = {}
errlist = filelist
return dupdict, errlist
def copyfiles(srcdir, dstdir):
from filecmp import cmp as filecmp
from shutil import copyfile
srcdir = normpath(srcdir)
dstdir = normpath(dstdir)
if not isdir(dstdir):
os.mkdir(dstdir)
for root, dirs, files in os.walk(srcdir):
root = normpath(root)
prefix = os.path.commonprefix([root, srcdir])
root = root[len(prefix):]
if root.startswith('/'):
root = root[1:]
for entry in dirs:
newdir = joinpath(dstdir, root, entry)
if not isdir(newdir):
os.mkdir(newdir)
for entry in files:
dest = normpath(joinpath(dstdir, root, entry))
src = normpath(joinpath(srcdir, root, entry))
if not isfile(dest) or not filecmp(src, dest):
copyfile(src, dest)
# some of the spec benchmarks expect to be run from one directory up.
# just create some symlinks that solve the problem
inlink = joinpath(dstdir, 'input')
outlink = joinpath(dstdir, 'output')
if not exists(inlink):
os.symlink('.', inlink)
if not exists(outlink):
os.symlink('.', outlink)
def copyfiles(srcdir, dstdir):
from filecmp import cmp as filecmp
from shutil import copyfile
srcdir = normpath(srcdir)
dstdir = normpath(dstdir)
if not isdir(dstdir):
os.mkdir(dstdir)
for root, dirs, files in os.walk(srcdir):
root = normpath(root)
prefix = os.path.commonprefix([root, srcdir])
root = root[len(prefix):]
if root.startswith('/'):
root = root[1:]
for entry in dirs:
newdir = joinpath(dstdir, root, entry)
if not isdir(newdir):
os.mkdir(newdir)
for entry in files:
dest = normpath(joinpath(dstdir, root, entry))
src = normpath(joinpath(srcdir, root, entry))
if not isfile(dest) or not filecmp(src, dest):
copyfile(src, dest)
# some of the spec benchmarks expect to be run from one directory up.
# just create some symlinks that solve the problem
inlink = joinpath(dstdir, 'input')
outlink = joinpath(dstdir, 'output')
if not exists(inlink):
os.symlink('.', inlink)
if not exists(outlink):
os.symlink('.', outlink)
def main():
opt_h = OptionHelper()
if opt_h.cascade():
input_type = opt_h.input_type()
if opt_h.output() is None:
opt_h.set_output('out')
opt_h.check_output()
cascade_dir = '%s/cascade00' % opt_h.output()
makedirs(cascade_dir)
results_fh = open('%s/results' % opt_h.output(), 'w')
cascade_train_out = '%s/train-out' % cascade_dir
cascade_test_out = '%s/test-out' % cascade_dir
basic_cmd = opt_h.basic_cmd()
output_file_type = ' -outputType UNDERSCORE4CCL'
log('running initial chunking')
run_cmd(basic_cmd \
+ opt_h.starter_train() \
+ opt_h.starter_train_out() \
+ output_file_type \
+ ' -output ' + cascade_train_out, \
verbose=opt_h.verbose())
run_cmd(basic_cmd \
+ opt_h.starter_train() \
+ opt_h.starter_test() \
+ opt_h.filter_flag() \
+ output_file_type \
+ ' -output ' + cascade_test_out, \
verbose=opt_h.verbose())
cascade_iter = 1
new_cascade_train_out_fname = get_output_fname(cascade_train_out)
cascade_expand_last = None
while True:
# convert test output to trees
cascade_test_out_fname = get_output_fname(cascade_test_out)
cascade_expand = []
log('building corpus record from ' + cascade_test_out)
for s_ind, sentence in enumerate(open(cascade_test_out_fname)):
i = 0
sentence_str = []
for chunk in sentence.split():
chunk = chunk.split('_')
chunk_str = []
for word in chunk:
if word.startswith('=') and len(word) > 1:
chunk_str.append(cascade_expand_last[s_ind][i])
else:
chunk_str.append(word)
i += 1
if len(chunk) == 1:
sentence_str.append(chunk_str[0])
else:
sentence_str.append('(' + (' '.join(chunk_str)) + ')')
cascade_expand.append(sentence_str)
cascade_test_eval_fname = cascade_dir + '/test-eval'
eval_fh = open(cascade_test_eval_fname, 'w')
for sent in cascade_expand:
print >>eval_fh, '(' + (' '.join(sent)).replace(' ;', '') + ')'
eval_fh.close()
# evaluate test output as trees
run_cmd(opt_h.eval_cmd() \
+ opt_h.starter_test() \
+ ' -cclpOutput ' + cascade_test_eval_fname \
+ opt_h.filter_flag(), fh=results_fh, \
verbose=opt_h.verbose())
cascade_expand_last = cascade_expand
log('running cascade level ' + str(cascade_iter))
# build term frequency map from last train output
cascade_train_out_fname = new_cascade_train_out_fname
phrasal_terms = PhrasalTerms(cascade_train_out_fname)
# create next-run train
next_run_train_fname = cascade_dir + '/next-train'
phrasal_terms.write_new_dataset(cascade_train_out_fname, \
next_run_train_fname)
# run chunker, output re-chunked train
new_cascade_dir = '%s/cascade%02d' % (opt_h.output(), cascade_iter)
makedirs(new_cascade_dir)
cascade_train_out = '%s/train-out' % new_cascade_dir
run_cmd(basic_cmd \
+ ' -train ' + next_run_train_fname \
+ ' -trainFileType SPL ' \
+ ' -test ' + next_run_train_fname \
+ ' -testFileType SPL ' \
+ output_file_type \
+ ' -output ' + cascade_train_out,
verbose=opt_h.verbose())
# if re-chunked train is the same as orig, break
new_cascade_train_out_fname = get_output_fname(cascade_train_out)
if filecmp(cascade_train_out_fname, new_cascade_train_out_fname):
break
# create next-run test
cascade_test_out = '%s/test-out' % new_cascade_dir
next_run_test_fname = cascade_dir + '/next-test'
phrasal_terms.write_new_dataset(cascade_test_out_fname, \
next_run_test_fname)
# run the chunker, output re-chunked test
run_cmd(basic_cmd \
+ ' -train ' + next_run_train_fname \
+ ' -trainFileType SPL ' \
+ ' -test ' + next_run_test_fname \
+ ' -testFileType SPL ' \
+ output_file_type \
+ ' -output ' + cascade_test_out,
verbose=opt_h.verbose())
cascade_dir = new_cascade_dir
cascade_iter += 1
results_fh.close()
else:
cmd = opt_h.basic_cmd()
output_flag = ''
if opt_h.stdout():
output_flag = ' -output -'
elif opt_h.output() is not None:
opt_h.check_output()
output_flag = ' -output ' + opt_h.output()
cmd += ' -outputType ' + opt_h.output_type()
cmd += output_flag
cmd += opt_h.starter_train()
cmd += opt_h.starter_test()
cmd += opt_h.filter_flag()
cmd += ' -E PRCL -e CLUMP,NPS,TREEBANKPREC'
run_cmd(cmd, verbose=opt_h.verbose())
def main():
opt_h = OptionHelper()
if opt_h.cascade():
input_type = opt_h.input_type()
if opt_h.output() is None:
opt_h.set_output('out')
opt_h.check_output()
cascade_dir = '%s/cascade00' % opt_h.output()
makedirs(cascade_dir)
results_fh = open('%s/results' % opt_h.output(), 'w')
cascade_train_out = '%s/train-out' % cascade_dir
cascade_test_out = '%s/test-out' % cascade_dir
basic_cmd = opt_h.basic_cmd()
output_file_type = ' -outputType UNDERSCORE4CCL'
log('running initial chunking')
run_cmd(basic_cmd \
+ opt_h.starter_train() \
+ opt_h.starter_train_out() \
+ output_file_type \
+ ' -output ' + cascade_train_out, \
verbose=opt_h.verbose())
run_cmd(basic_cmd \
+ opt_h.starter_train() \
+ opt_h.starter_test() \
+ opt_h.filter_flag() \
+ output_file_type \
+ ' -output ' + cascade_test_out, \
verbose=opt_h.verbose())
cascade_iter = 1
new_cascade_train_out_fname = get_output_fname(cascade_train_out)
cascade_expand_last = None
while True:
# convert test output to trees
cascade_test_out_fname = get_output_fname(cascade_test_out)
cascade_expand = []
log('building corpus record from ' + cascade_test_out)
for s_ind, sentence in enumerate(open(cascade_test_out_fname)):
i = 0
sentence_str = []
for chunk in sentence.split():
chunk = chunk.split('_')
chunk_str = []
for word in chunk:
if word.startswith('=') and len(word) > 1:
chunk_str.append(cascade_expand_last[s_ind][i])
else:
chunk_str.append(word)
i += 1
if len(chunk) == 1:
sentence_str.append(chunk_str[0])
else:
sentence_str.append('(' + (' '.join(chunk_str)) + ')')
cascade_expand.append(sentence_str)
cascade_test_eval_fname = cascade_dir + '/test-eval'
eval_fh = open(cascade_test_eval_fname, 'w')
for sent in cascade_expand:
print >> eval_fh, '(' + (' '.join(sent)).replace(' ;',
'') + ')'
eval_fh.close()
# evaluate test output as trees
run_cmd(opt_h.eval_cmd() \
+ opt_h.starter_test() \
+ ' -cclpOutput ' + cascade_test_eval_fname \
+ opt_h.filter_flag(), fh=results_fh, \
verbose=opt_h.verbose())
cascade_expand_last = cascade_expand
log('running cascade level ' + str(cascade_iter))
# build term frequency map from last train output
cascade_train_out_fname = new_cascade_train_out_fname
phrasal_terms = PhrasalTerms(cascade_train_out_fname)
# create next-run train
next_run_train_fname = cascade_dir + '/next-train'
phrasal_terms.write_new_dataset(cascade_train_out_fname, \
next_run_train_fname)
# run chunker, output re-chunked train
new_cascade_dir = '%s/cascade%02d' % (opt_h.output(), cascade_iter)
makedirs(new_cascade_dir)
cascade_train_out = '%s/train-out' % new_cascade_dir
run_cmd(basic_cmd \
+ ' -train ' + next_run_train_fname \
+ ' -trainFileType SPL ' \
+ ' -test ' + next_run_train_fname \
+ ' -testFileType SPL ' \
+ output_file_type \
+ ' -output ' + cascade_train_out,
verbose=opt_h.verbose())
# if re-chunked train is the same as orig, break
new_cascade_train_out_fname = get_output_fname(cascade_train_out)
if filecmp(cascade_train_out_fname, new_cascade_train_out_fname):
break
# create next-run test
cascade_test_out = '%s/test-out' % new_cascade_dir
next_run_test_fname = cascade_dir + '/next-test'
phrasal_terms.write_new_dataset(cascade_test_out_fname, \
next_run_test_fname)
# run the chunker, output re-chunked test
run_cmd(basic_cmd \
+ ' -train ' + next_run_train_fname \
+ ' -trainFileType SPL ' \
+ ' -test ' + next_run_test_fname \
+ ' -testFileType SPL ' \
+ output_file_type \
+ ' -output ' + cascade_test_out,
verbose=opt_h.verbose())
cascade_dir = new_cascade_dir
cascade_iter += 1
results_fh.close()
else:
cmd = opt_h.basic_cmd()
output_flag = ''
if opt_h.stdout():
output_flag = ' -output -'
elif opt_h.output() is not None:
opt_h.check_output()
output_flag = ' -output ' + opt_h.output()
cmd += ' -outputType ' + opt_h.output_type()
cmd += output_flag
cmd += opt_h.starter_train()
cmd += opt_h.starter_test()
cmd += opt_h.filter_flag()
cmd += ' -E PRCL -e CLUMP,NPS,TREEBANKPREC'
run_cmd(cmd, verbose=opt_h.verbose())
