text stringlengths 81 112k |
|---|
Takes an input path to the images folder of an experiment and generates
automatically the category - filenumber list needed to construct an
appropriate _categories object.
Parameters :
loader : Loader object which contains
impath : string
path to the input, i.e. image-, files of the experiment. All
subfolders in that path will be treated as categories. If no
subfolders are present, category 1 will be assigned and all
files in the folder are considered input images.
Images have to end in '.png'.
ftrpath : string
path to the feature folder. It is expected that the folder
structure corresponds to the structure in impath, i.e.
ftrpath/category/featurefolder/featuremap.mat
Furthermore, features are assumed to be the same for all
categories.
def DirectoryStimuliFactory(loader):
"""
Takes an input path to the images folder of an experiment and generates
automatically the category - filenumber list needed to construct an
appropriate _categories object.
Parameters :
loader : Loader object which contains
impath : string
path to the input, i.e. image-, files of the experiment. All
subfolders in that path will be treated as categories. If no
subfolders are present, category 1 will be assigned and all
files in the folder are considered input images.
Images have to end in '.png'.
ftrpath : string
path to the feature folder. It is expected that the folder
structure corresponds to the structure in impath, i.e.
ftrpath/category/featurefolder/featuremap.mat
Furthermore, features are assumed to be the same for all
categories.
"""
impath = loader.impath
ftrpath = loader.ftrpath
# checks whether user has reading permission for the path
assert os.access(impath, os.R_OK)
assert os.access(ftrpath, os.R_OK)
# EXTRACTING IMAGE NAMES
img_per_cat = {}
# extract only directories in the given folder
subfolders = [name for name in os.listdir(impath) if os.path.isdir(
os.path.join(impath, name))]
# if there are no subfolders, walk through files. Take 1 as key for the
# categories object
if not subfolders:
[_, _, files] = next(os.walk(os.path.join(impath)))
# this only takes entries that end with '.png'
entries = {1:
[int(cur_file[cur_file.find('_')+1:-4]) for cur_file
in files if cur_file.endswith('.png')]}
img_per_cat.update(entries)
subfolders = ['']
# if there are subfolders, walk through them
else:
for directory in subfolders:
[_, _, files] = next(os.walk(os.path.join(impath, directory)))
# this only takes entries that end with '.png'. Strips ending and
# considers everything after the first '_' as the imagenumber
imagenumbers = [int(cur_file[cur_file.find('_')+1:-4])
for cur_file in files
if (cur_file.endswith('.png') & (len(cur_file) > 4))]
entries = {int(directory): imagenumbers}
img_per_cat.update(entries)
del directory
del imagenumbers
# in case subfolders do not exist, '' is appended here.
_, features, files = next(os.walk(os.path.join(ftrpath,
subfolders[0])))
return Categories(loader, img_per_cat = img_per_cat, features = features) |
Filter the fixmat such that it only contains fixations on images
in categories that are also in the categories object
def fixations(self):
''' Filter the fixmat such that it only contains fixations on images
in categories that are also in the categories object'''
if not self._fixations:
raise RuntimeError('This Images object does not have'
+' an associated fixmat')
if len(list(self._categories.keys())) == 0:
return None
else:
idx = np.zeros(self._fixations.x.shape, dtype='bool')
for (cat, _) in list(self._categories.items()):
idx = idx | ((self._fixations.category == cat))
return self._fixations[idx] |
Saves a new image to disk
def data(self, value):
"""
Saves a new image to disk
"""
self.loader.save_image(self.category, self.image, value) |
Returns all fixations that are on this image.
A precondition for this to work is that a fixmat
is associated with this Image object.
def fixations(self):
"""
Returns all fixations that are on this image.
A precondition for this to work is that a fixmat
is associated with this Image object.
"""
if not self._fixations:
raise RuntimeError('This Images object does not have'
+' an associated fixmat')
return self._fixations[(self._fixations.category == self.category) &
(self._fixations.filenumber == self.image)] |
Generator for creating the cross-validation slices.
Returns
A tuple of that contains two fixmats (training and test)
and two Category objects (test and train).
def generate(self):
"""
Generator for creating the cross-validation slices.
Returns
A tuple of that contains two fixmats (training and test)
and two Category objects (test and train).
"""
for _ in range(0, self.num_slices):
#1. separate fixmat into test and training fixmat
subjects = np.unique(self.fm.SUBJECTINDEX)
test_subs = randsample(subjects,
self.subject_hold_out*len(subjects))
train_subs = [x for x in subjects if x not in test_subs]
test_fm = self.fm[ismember(self.fm.SUBJECTINDEX, test_subs)]
train_fm = self.fm[ismember(self.fm.SUBJECTINDEX, train_subs)]
#2. distribute images
test_imgs = {}
train_imgs = {}
id_test = (test_fm.x <1) & False
id_train = (train_fm.x <1) & False
for cat in self.categories:
imgs = cat.images()
test_imgs.update({cat.category:randsample(imgs,
self.image_hold_out*len(imgs)).tolist()})
train_imgs.update({cat.category:[x for x in imgs
if not ismember(x, test_imgs[cat.category])]})
id_test = id_test | ((ismember(test_fm.filenumber,
test_imgs[cat.category])) &
(test_fm.category == cat.category))
id_train = id_train | ((ismember(train_fm.filenumber,
train_imgs[cat.category])) &
(train_fm.category == cat.category))
#3. Create categories objects and yield result
test_stimuli = Categories(self.categories.loader, test_imgs,
features=self.categories._features,
fixations=test_fm)
train_stimuli = Categories(self.categories.loader, train_imgs,
features=self.categories._features,
fixations=train_fm)
yield (train_fm[id_train],
train_stimuli,
test_fm[id_test],
test_stimuli) |
Computes angle and length differences up to given order and deletes
suspiciously long fixations.
Input
fm: Fixmat
Fixmat for which to comput angle and length differences
max_back: Int
Computes delta angle and amplitude up to order max_back.
dur_cap: Int
Longest allowed fixation duration
Output
fm: Fixmat
Filtered fixmat that aligns to the other outputs.
durations: ndarray
Duration for each fixation in fm
forward_angle:
Angle between previous and next saccade.
def prepare_data(fm, max_back, dur_cap=700):
'''
Computes angle and length differences up to given order and deletes
suspiciously long fixations.
Input
fm: Fixmat
Fixmat for which to comput angle and length differences
max_back: Int
Computes delta angle and amplitude up to order max_back.
dur_cap: Int
Longest allowed fixation duration
Output
fm: Fixmat
Filtered fixmat that aligns to the other outputs.
durations: ndarray
Duration for each fixation in fm
forward_angle:
Angle between previous and next saccade.
'''
durations = np.roll(fm.end - fm.start, 1).astype(float)
angles, lengths, ads, lds = anglendiff(fm, roll=max_back, return_abs=True)
# durations and ads are aligned in a way that an entry in ads
# encodes the angle of the saccade away from a fixation in
# durations
forward_angle = abs(reshift(ads[0])).astype(float)
ads = [abs(reshift(a)) for a in ads]
# Now filter out weird fixation durations
id_in = durations > dur_cap
durations[id_in] = np.nan
forward_angle[id_in] = np.nan
return fm, durations, forward_angle, ads, lds |
Computes the mean fixation duration at forward angles.
def saccadic_momentum_effect(durations, forward_angle,
summary_stat=nanmean):
"""
Computes the mean fixation duration at forward angles.
"""
durations_per_da = np.nan * np.ones((len(e_angle) - 1,))
for i, (bo, b1) in enumerate(zip(e_angle[:-1], e_angle[1:])):
idx = (
bo <= forward_angle) & (
forward_angle < b1) & (
~np.isnan(durations))
durations_per_da[i] = summary_stat(durations[idx])
return durations_per_da |
Computes a measure of fixation durations at delta angle and delta
length combinations.
def ior_effect(durations, angle_diffs, length_diffs,
summary_stat=np.mean, parallel=True, min_samples=20):
"""
Computes a measure of fixation durations at delta angle and delta
length combinations.
"""
raster = np.empty((len(e_dist) - 1, len(e_angle) - 1), dtype=object)
for a, (a_low, a_upp) in enumerate(zip(e_angle[:-1], e_angle[1:])):
for d, (d_low, d_upp) in enumerate(zip(e_dist[:-1], e_dist[1:])):
idx = ((d_low <= length_diffs) & (length_diffs < d_upp) &
(a_low <= angle_diffs) & (angle_diffs < a_upp))
if sum(idx) < min_samples:
raster[d, a] = np.array([np.nan])
else:
raster[d, a] = durations[idx]
if parallel:
p = pool.Pool(3)
result = p.map(summary_stat, list(raster.flatten()))
p.terminate()
else:
result = list(map(summary_stat, list(raster.flatten())))
for idx, value in enumerate(result):
i, j = np.unravel_index(idx, raster.shape)
raster[i, j] = value
return raster |
Fits a non-linear piecewise regression to fixtaion durations for a fixmat.
Returns corrected fixation durations.
def predict_fixation_duration(
durations, angles, length_diffs, dataset=None, params=None):
"""
Fits a non-linear piecewise regression to fixtaion durations for a fixmat.
Returns corrected fixation durations.
"""
if dataset is None:
dataset = np.ones(durations.shape)
corrected_durations = np.nan * np.ones(durations.shape)
for i, ds in enumerate(np.unique(dataset)):
e = lambda v, x, y, z: (leastsq_dual_model(x, z, *v) - y)
v0 = [120, 220.0, -.1, 0.5, .1, .1]
id_ds = dataset == ds
idnan = (
~np.isnan(angles)) & (
~np.isnan(durations)) & (
~np.isnan(length_diffs))
v, s = leastsq(
e, v0, args=(
angles[
idnan & id_ds], durations[
idnan & id_ds], length_diffs[
idnan & id_ds]), maxfev=10000)
corrected_durations[id_ds] = (durations[id_ds] -
(leastsq_dual_model(angles[id_ds], length_diffs[id_ds], *v)))
if params is not None:
params['v' + str(i)] = v
params['s' + str(i)] = s
return corrected_durations |
Calculates the saccadic momentum effect for individual subjects.
Removes any effect of amplitude differences.
The parameters are fitted on unbinned data. The effects are
computed on binned data. See e_dist and e_angle for the binning
parameter.
def subject_predictions(fm, field='SUBJECTINDEX',
method=predict_fixation_duration, data=None):
'''
Calculates the saccadic momentum effect for individual subjects.
Removes any effect of amplitude differences.
The parameters are fitted on unbinned data. The effects are
computed on binned data. See e_dist and e_angle for the binning
parameter.
'''
if data is None:
fma, dura, faa, adsa, ldsa = prepare_data(fm, dur_cap=700, max_back=5)
adsa = adsa[0]
ldsa = ldsa[0]
else:
fma, dura, faa, adsa, ldsa = data
fma = fma.copy() # [ones(fm.x.shape)]
sub_effects = []
sub_predictions = []
parameters = []
for i, fmsub in enumerate(np.unique(fma.field(field))):
id = fma.field(field) == fmsub
#_, dur, fa, ads, lds = prepare_data(fmsub, dur_cap = 700, max_back=5)
dur, fa, ads, lds = dura[id], faa[id], adsa[id], ldsa[id]
params = {}
_ = method(dur, fa, lds, params=params)
ps = params['v0']
ld_corrected = leastsq_only_dist(lds, ps[4], ps[5])
prediction = leastsq_only_angle(fa, ps[0], ps[1], ps[2], ps[3])
sub_predictions += [saccadic_momentum_effect(prediction, fa)]
sub_effects += [saccadic_momentum_effect(dur - ld_corrected, fa)]
parameters += [ps]
return np.array(sub_effects), np.array(sub_predictions), parameters |
Calculates how well the fixations from a set of subjects on a set of
images can be predicted with the fixations from another set of subjects
on another set of images.
The prediction is carried out by computing a fixation density map from
fixations of predicting_subjects subjects on predicting_images images.
Prediction accuracy is assessed by measures.prediction_scores.
Parameters
fm : fixmat instance
category : int
Category from which the fixations are taken.
predicting_filenumbers : list
List of filenumbers used for prediction, i.e. images where fixations
for the prediction are taken from.
predicting_subjects : list
List of subjects whose fixations on images in predicting_filenumbers
are used for the prediction.
predicted_filenumnbers : list
List of images from which the to be predicted fixations are taken.
predicted_subjects : list
List of subjects used for evaluation, i.e subjects whose fixations
on images in predicted_filenumbers are taken for evaluation.
controls : bool, optional
If True (default), n_predict subjects are chosen from the fixmat.
If False, 1000 fixations are randomly generated and used for
testing.
scale_factor : int, optional
specifies the scaling of the fdm. Default is 1.
Returns
auc : area under the roc curve for sets of actuals and controls
true_pos_rate : ndarray
Rate of true positives for every given threshold value.
All values appearing in actuals are taken as thresholds. Uses lower
sum interpolation.
false_pos_rate : ndarray
See true_pos_rate but for false positives.
def intersubject_scores(fm, category, predicting_filenumbers,
predicting_subjects, predicted_filenumbers,
predicted_subjects, controls = True, scale_factor = 1):
"""
Calculates how well the fixations from a set of subjects on a set of
images can be predicted with the fixations from another set of subjects
on another set of images.
The prediction is carried out by computing a fixation density map from
fixations of predicting_subjects subjects on predicting_images images.
Prediction accuracy is assessed by measures.prediction_scores.
Parameters
fm : fixmat instance
category : int
Category from which the fixations are taken.
predicting_filenumbers : list
List of filenumbers used for prediction, i.e. images where fixations
for the prediction are taken from.
predicting_subjects : list
List of subjects whose fixations on images in predicting_filenumbers
are used for the prediction.
predicted_filenumnbers : list
List of images from which the to be predicted fixations are taken.
predicted_subjects : list
List of subjects used for evaluation, i.e subjects whose fixations
on images in predicted_filenumbers are taken for evaluation.
controls : bool, optional
If True (default), n_predict subjects are chosen from the fixmat.
If False, 1000 fixations are randomly generated and used for
testing.
scale_factor : int, optional
specifies the scaling of the fdm. Default is 1.
Returns
auc : area under the roc curve for sets of actuals and controls
true_pos_rate : ndarray
Rate of true positives for every given threshold value.
All values appearing in actuals are taken as thresholds. Uses lower
sum interpolation.
false_pos_rate : ndarray
See true_pos_rate but for false positives.
"""
predicting_fm = fm[
(ismember(fm.SUBJECTINDEX, predicting_subjects)) &
(ismember(fm.filenumber, predicting_filenumbers)) &
(fm.category == category)]
predicted_fm = fm[
(ismember(fm.SUBJECTINDEX,predicted_subjects)) &
(ismember(fm.filenumber,predicted_filenumbers))&
(fm.category == category)]
try:
predicting_fdm = compute_fdm(predicting_fm, scale_factor = scale_factor)
except RuntimeError:
predicting_fdm = None
if controls == True:
fm_controls = fm[
(ismember(fm.SUBJECTINDEX, predicted_subjects)) &
((ismember(fm.filenumber, predicted_filenumbers)) != True) &
(fm.category == category)]
return measures.prediction_scores(predicting_fdm, predicted_fm,
controls = (fm_controls.y, fm_controls.x))
return measures.prediction_scores(predicting_fdm, predicted_fm, controls = None) |
Calculates how well the fixations of n random subjects on one image can
be predicted with the fixations of m other random subjects.
Notes
Function that uses intersubject_auc for computing auc.
Parameters
fm : fixmat instance
category : int
Category from which the fixations are taken.
filnumber : int
Image from which fixations are taken.
n_train : int
The number of subjects which are used for prediction.
n_predict : int
The number of subjects to predict
controls : bool, optional
If True (default), n_predict subjects are chosen from the fixmat.
If False, 1000 fixations are randomly generated and used for
testing.
scale_factor : int, optional
specifies the scaling of the fdm. Default is 1.
Returns
tuple : prediction scores
def intersubject_scores_random_subjects(fm, category, filenumber, n_train,
n_predict, controls=True,
scale_factor = 1):
"""
Calculates how well the fixations of n random subjects on one image can
be predicted with the fixations of m other random subjects.
Notes
Function that uses intersubject_auc for computing auc.
Parameters
fm : fixmat instance
category : int
Category from which the fixations are taken.
filnumber : int
Image from which fixations are taken.
n_train : int
The number of subjects which are used for prediction.
n_predict : int
The number of subjects to predict
controls : bool, optional
If True (default), n_predict subjects are chosen from the fixmat.
If False, 1000 fixations are randomly generated and used for
testing.
scale_factor : int, optional
specifies the scaling of the fdm. Default is 1.
Returns
tuple : prediction scores
"""
subjects = np.unique(fm.SUBJECTINDEX)
if len(subjects) < n_train + n_predict:
raise ValueError("""Not enough subjects in fixmat""")
# draw a random sample of subjects for testing and evaluation, according
# to the specified set sizes (n_train, n_predict)
np.random.shuffle(subjects)
predicted_subjects = subjects[0 : n_predict]
predicting_subjects = subjects[n_predict : n_predict + n_train]
assert len(predicting_subjects) == n_train
assert len(predicted_subjects) == n_predict
assert [x not in predicting_subjects for x in predicted_subjects]
return intersubject_scores(fm, category, [filenumber], predicting_subjects,
[filenumber], predicted_subjects,
controls, scale_factor) |
compute the inter-subject consistency upper bound for a fixmat.
Input:
fm : a fixmat instance
nr_subs : the number of subjects used for the prediction. Defaults
to the total number of subjects in the fixmat minus 1
scale_factor : the scale factor of the FDMs. Default is 1.
Returns:
A list of scores; the list contains one dictionary for each measure.
Each dictionary contains one key for each category and corresponding
values is an array with scores for each subject.
def upper_bound(fm, nr_subs = None, scale_factor = 1):
"""
compute the inter-subject consistency upper bound for a fixmat.
Input:
fm : a fixmat instance
nr_subs : the number of subjects used for the prediction. Defaults
to the total number of subjects in the fixmat minus 1
scale_factor : the scale factor of the FDMs. Default is 1.
Returns:
A list of scores; the list contains one dictionary for each measure.
Each dictionary contains one key for each category and corresponding
values is an array with scores for each subject.
"""
nr_subs_total = len(np.unique(fm.SUBJECTINDEX))
if not nr_subs:
nr_subs = nr_subs_total - 1
assert (nr_subs < nr_subs_total)
# initialize output structure; every measure gets one dict with
# category numbers as keys and numpy-arrays as values
intersub_scores = []
for measure in range(len(measures.scores)):
res_dict = {}
result_vectors = [np.empty(nr_subs_total) + np.nan
for _ in np.unique(fm.category)]
res_dict.update(list(zip(np.unique(fm.category), result_vectors)))
intersub_scores.append(res_dict)
#compute inter-subject scores for every stimulus, with leave-one-out
#over subjects
for fm_cat in fm.by_field('category'):
cat = fm_cat.category[0]
for (sub_counter, sub) in enumerate(np.unique(fm_cat.SUBJECTINDEX)):
image_scores = []
for fm_single in fm_cat.by_field('filenumber'):
predicting_subs = (np.setdiff1d(np.unique(
fm_single.SUBJECTINDEX),[sub]))
np.random.shuffle(predicting_subs)
predicting_subs = predicting_subs[0:nr_subs]
predicting_fm = fm_single[
(ismember(fm_single.SUBJECTINDEX, predicting_subs))]
predicted_fm = fm_single[fm_single.SUBJECTINDEX == sub]
try:
predicting_fdm = compute_fdm(predicting_fm,
scale_factor = scale_factor)
except RuntimeError:
predicting_fdm = None
image_scores.append(measures.prediction_scores(
predicting_fdm, predicted_fm))
for (measure, score) in enumerate(nanmean(image_scores, 0)):
intersub_scores[measure][cat][sub_counter] = score
return intersub_scores |
Compute the spatial bias lower bound for a fixmat.
Input:
fm : a fixmat instance
nr_subs : the number of subjects used for the prediction. Defaults
to the total number of subjects in the fixmat minus 1
nr_imgs : the number of images used for prediction. If given, the
same number will be used for every category. If not given,
leave-one-out will be used in all categories.
scale_factor : the scale factor of the FDMs. Default is 1.
Returns:
A list of spatial bias scores; the list contains one dictionary for each
measure. Each dictionary contains one key for each category and
corresponding values is an array with scores for each subject.
def lower_bound(fm, nr_subs = None, nr_imgs = None, scale_factor = 1):
"""
Compute the spatial bias lower bound for a fixmat.
Input:
fm : a fixmat instance
nr_subs : the number of subjects used for the prediction. Defaults
to the total number of subjects in the fixmat minus 1
nr_imgs : the number of images used for prediction. If given, the
same number will be used for every category. If not given,
leave-one-out will be used in all categories.
scale_factor : the scale factor of the FDMs. Default is 1.
Returns:
A list of spatial bias scores; the list contains one dictionary for each
measure. Each dictionary contains one key for each category and
corresponding values is an array with scores for each subject.
"""
nr_subs_total = len(np.unique(fm.SUBJECTINDEX))
if nr_subs is None:
nr_subs = nr_subs_total - 1
assert (nr_subs < nr_subs_total)
# initialize output structure; every measure gets one dict with
# category numbers as keys and numpy-arrays as values
sb_scores = []
for measure in range(len(measures.scores)):
res_dict = {}
result_vectors = [np.empty(nr_subs_total) + np.nan
for _ in np.unique(fm.category)]
res_dict.update(list(zip(np.unique(fm.category),result_vectors)))
sb_scores.append(res_dict)
# compute mean spatial bias predictive power for all subjects in all
# categories
for fm_cat in fm.by_field('category'):
cat = fm_cat.category[0]
nr_imgs_cat = len(np.unique(fm_cat.filenumber))
if not nr_imgs:
nr_imgs_current = nr_imgs_cat - 1
else:
nr_imgs_current = nr_imgs
assert(nr_imgs_current < nr_imgs_cat)
for (sub_counter, sub) in enumerate(np.unique(fm.SUBJECTINDEX)):
image_scores = []
for fm_single in fm_cat.by_field('filenumber'):
# Iterating by field filenumber makes filenumbers
# in fm_single unique: Just take the first one to get the
# filenumber for this fixmat
fn = fm_single.filenumber[0]
predicting_subs = (np.setdiff1d(np.unique(
fm_cat.SUBJECTINDEX), [sub]))
np.random.shuffle(predicting_subs)
predicting_subs = predicting_subs[0:nr_subs]
predicting_fns = (np.setdiff1d(np.unique(
fm_cat.filenumber), [fn]))
np.random.shuffle(predicting_fns)
predicting_fns = predicting_fns[0:nr_imgs_current]
predicting_fm = fm_cat[
(ismember(fm_cat.SUBJECTINDEX, predicting_subs)) &
(ismember(fm_cat.filenumber, predicting_fns))]
predicted_fm = fm_single[fm_single.SUBJECTINDEX == sub]
try:
predicting_fdm = compute_fdm(predicting_fm,
scale_factor = scale_factor)
except RuntimeError:
predicting_fdm = None
image_scores.append(measures.prediction_scores(predicting_fdm,
predicted_fm))
for (measure, score) in enumerate(nanmean(image_scores, 0)):
sb_scores[measure][cat][sub_counter] = score
return sb_scores |
Calculates subscripts for indices into regularly spaced matrixes.
def ind2sub(ind, dimensions):
"""
Calculates subscripts for indices into regularly spaced matrixes.
"""
# check that the index is within range
if ind >= np.prod(dimensions):
raise RuntimeError("ind2sub: index exceeds array size")
cum_dims = list(dimensions)
cum_dims.reverse()
m = 1
mult = []
for d in cum_dims:
m = m*d
mult.append(m)
mult.pop()
mult.reverse()
mult.append(1)
indices = []
for d in mult:
indices.append((ind/d)+1)
ind = ind - (ind/d)*d
return indices |
An exemplary sub2ind implementation to create randomization
scripts.
This function calculates indices from subscripts into regularly spaced
matrixes.
def sub2ind(indices, dimensions):
"""
An exemplary sub2ind implementation to create randomization
scripts.
This function calculates indices from subscripts into regularly spaced
matrixes.
"""
# check that none of the indices exceeds the size of the array
if any([i > j for i, j in zip(indices, dimensions)]):
raise RuntimeError("sub2ind:an index exceeds its dimension's size")
dims = list(dimensions)
dims.append(1)
dims.remove(dims[0])
dims.reverse()
ind = list(indices)
ind.reverse()
idx = 0
mult = 1
for (cnt, dim) in zip(ind, dims):
mult = dim*mult
idx = idx + (cnt-1)*mult
return idx |
Restores a task store from file.
def RestoreTaskStoreFactory(store_class, chunk_size, restore_file, save_file):
"""
Restores a task store from file.
"""
intm_results = np.load(restore_file)
intm = intm_results[intm_results.files[0]]
idx = np.isnan(intm).flatten().nonzero()[0]
partitions = math.ceil(len(idx) / float(chunk_size))
task_store = store_class(partitions, idx.tolist(), save_file)
task_store.num_tasks = len(idx)
# Also set up matrices for saving results
for f in intm_results.files:
task_store.__dict__[f] = intm_results[f]
return task_store |
Reschedule all running tasks.
def xmlrpc_reschedule(self):
"""
Reschedule all running tasks.
"""
if not len(self.scheduled_tasks) == 0:
self.reschedule = list(self.scheduled_tasks.items())
self.scheduled_tasks = {}
return True |
Return a new task description: ID and necessary parameters,
all are given in a dictionary
def xmlrpc_get_task(self):
"""
Return a new task description: ID and necessary parameters,
all are given in a dictionary
"""
try:
if len(self.reschedule) == 0:
(task_id, cur_task) = next(self.task_iterator)
else:
(task_id, cur_task) = self.reschedule.pop()
self.scheduled_tasks.update({task_id: cur_task})
return (task_id, cur_task.to_dict())
except StopIteration:
print('StopIteration: No more tasks')
return False
except Exception as err:
print('Some other error')
print(err)
return False |
Take the results of a computation and put it into the results list.
def xmlrpc_task_done(self, result):
"""
Take the results of a computation and put it into the results list.
"""
(task_id, task_results) = result
del self.scheduled_tasks[task_id]
self.task_store.update_results(task_id, task_results)
self.results += 1
return True |
Return a status message
def xmlrpc_status(self):
"""
Return a status message
"""
return ("""
%i Jobs are still wating for execution
%i Jobs are being processed
%i Jobs are done
""" %(self.task_store.partitions -
self.results -
len(self.scheduled_tasks),
len(self.scheduled_tasks),
self.results)) |
Save results and own state into file.
def xmlrpc_save2file(self, filename):
"""
Save results and own state into file.
"""
savefile = open(filename,'wb')
try:
pickle.dump({'scheduled':self.scheduled_tasks,
'reschedule':self.reschedule},savefile)
except pickle.PicklingError:
return -1
savefile.close()
return 1 |
This function needs to be called to start the computation.
def run(self):
"""This function needs to be called to start the computation."""
(task_id, tasks) = self.server.get_task()
self.task_store.from_dict(tasks)
for (index, task) in self.task_store:
result = self.compute(index, task)
self.results.append(result)
self.server.task_done((task_id, self.results)) |
Configures the task store to be the task_store described
in description
def from_dict(self, description):
"""Configures the task store to be the task_store described
in description"""
assert(self.ident == description['ident'])
self.partitions = description['partitions']
self.indices = description['indices'] |
Partitions all tasks into groups of tasks. A group is
represented by a task_store object that indexes a sub-
set of tasks.
def partition(self):
"""Partitions all tasks into groups of tasks. A group is
represented by a task_store object that indexes a sub-
set of tasks."""
step = int(math.ceil(self.num_tasks / float(self.partitions)))
if self.indices == None:
slice_ind = list(range(0, self.num_tasks, step))
for start in slice_ind:
yield self.__class__(self.partitions,
list(range(start, start + step)))
else:
slice_ind = list(range(0, len(self.indices), step))
for start in slice_ind:
if start + step <= len(self.indices):
yield self.__class__(self.partitions,
self.indices[start: start + step])
else:
yield self.__class__(self.partitions, self.indices[start:]) |
Fits a 3D distribution with splines.
Input:
samples: Array
Array of samples from a probability distribution
e_x: Array
Edges that define the events in the probability
distribution along the x direction. For example,
e_x[0] < samples[0] <= e_x[1] picks out all
samples that are associated with the first event.
e_y: Array
See e_x, but for the y direction.
remove_zeros: Bool
If True, events that are not observed will not
be part of the fitting process. If False, those
events will be modelled as finfo('float').eps
**kw: Arguments that are passed on to spline_bse1d.
Returns:
distribution: Array
An array that gives an estimate of probability for
events defined by e.
knots: Tuple of arrays
Sequence of knots that were used for the spline basis (x,y)
def fit3d(samples, e_x, e_y, e_z, remove_zeros = False, **kw):
"""Fits a 3D distribution with splines.
Input:
samples: Array
Array of samples from a probability distribution
e_x: Array
Edges that define the events in the probability
distribution along the x direction. For example,
e_x[0] < samples[0] <= e_x[1] picks out all
samples that are associated with the first event.
e_y: Array
See e_x, but for the y direction.
remove_zeros: Bool
If True, events that are not observed will not
be part of the fitting process. If False, those
events will be modelled as finfo('float').eps
**kw: Arguments that are passed on to spline_bse1d.
Returns:
distribution: Array
An array that gives an estimate of probability for
events defined by e.
knots: Tuple of arrays
Sequence of knots that were used for the spline basis (x,y)
"""
height, width, depth = len(e_y)-1, len(e_x)-1, len(e_z)-1
(p_est, _) = np.histogramdd(samples, (e_x, e_y, e_z))
p_est = p_est/sum(p_est.flat)
p_est = p_est.flatten()
if remove_zeros:
non_zero = ~(p_est == 0)
else:
non_zero = (p_est >= 0)
basis = spline_base3d(width,height, depth, **kw)
model = linear_model.BayesianRidge()
model.fit(basis[:, non_zero].T, p_est[:,np.newaxis][non_zero,:])
return (model.predict(basis.T).reshape((width, height, depth)),
p_est.reshape((width, height, depth))) |
Fits a 2D distribution with splines.
Input:
samples: Matrix or list of arrays
If matrix, it must be of size Nx2, where N is the number of
observations. If list, it must contain two arrays of length
N.
e_x: Array
Edges that define the events in the probability
distribution along the x direction. For example,
e_x[0] < samples[0] <= e_x[1] picks out all
samples that are associated with the first event.
e_y: Array
See e_x, but for the y direction.
remove_zeros: Bool
If True, events that are not observed will not
be part of the fitting process. If False, those
events will be modelled as finfo('float').eps
**kw: Arguments that are passed on to spline_bse1d.
Returns:
distribution: Array
An array that gives an estimate of probability for
events defined by e.
knots: Tuple of arrays
Sequence of knots that were used for the spline basis (x,y)
def fit2d(samples,e_x, e_y, remove_zeros = False, p_est = None, **kw):
"""Fits a 2D distribution with splines.
Input:
samples: Matrix or list of arrays
If matrix, it must be of size Nx2, where N is the number of
observations. If list, it must contain two arrays of length
N.
e_x: Array
Edges that define the events in the probability
distribution along the x direction. For example,
e_x[0] < samples[0] <= e_x[1] picks out all
samples that are associated with the first event.
e_y: Array
See e_x, but for the y direction.
remove_zeros: Bool
If True, events that are not observed will not
be part of the fitting process. If False, those
events will be modelled as finfo('float').eps
**kw: Arguments that are passed on to spline_bse1d.
Returns:
distribution: Array
An array that gives an estimate of probability for
events defined by e.
knots: Tuple of arrays
Sequence of knots that were used for the spline basis (x,y)
"""
if p_est is None:
height = len(e_y)-1
width = len(e_x)-1
(p_est, _) = np.histogramdd(samples, (e_x, e_y))
else:
p_est = p_est.T
width, height = p_est.shape
# p_est contains x in dim 1 and y in dim 0
shape = p_est.shape
p_est = (p_est/sum(p_est.flat)).reshape(shape)
mx = p_est.sum(1)
my = p_est.sum(0)
# Transpose hist to have x in dim 0
p_est = p_est.T.flatten()
basis, knots = spline_base2d(width, height, marginal_x = mx, marginal_y = my, **kw)
model = linear_model.BayesianRidge()
if remove_zeros:
non_zero = ~(p_est == 0)
model.fit(basis[:, non_zero].T, p_est[non_zero])
else:
non_zero = (p_est >= 0)
p_est[~non_zero,:] = np.finfo(float).eps
model.fit(basis.T, p_est)
return (model.predict(basis.T).reshape((height, width)),
p_est.reshape((height, width)), knots) |
Fits a 1D distribution with splines.
Input:
samples: Array
Array of samples from a probability distribution
e: Array
Edges that define the events in the probability
distribution. For example, e[0] < x <= e[1] is
the range of values that are associated with the
first event.
**kw: Arguments that are passed on to spline_bse1d.
Returns:
distribution: Array
An array that gives an estimate of probability for
events defined by e.
knots: Array
Sequence of knots that were used for the spline basis
def fit1d(samples, e, remove_zeros = False, **kw):
"""Fits a 1D distribution with splines.
Input:
samples: Array
Array of samples from a probability distribution
e: Array
Edges that define the events in the probability
distribution. For example, e[0] < x <= e[1] is
the range of values that are associated with the
first event.
**kw: Arguments that are passed on to spline_bse1d.
Returns:
distribution: Array
An array that gives an estimate of probability for
events defined by e.
knots: Array
Sequence of knots that were used for the spline basis
"""
samples = samples[~np.isnan(samples)]
length = len(e)-1
hist,_ = np.histogramdd(samples, (e,))
hist = hist/sum(hist)
basis, knots = spline_base1d(length, marginal = hist, **kw)
non_zero = hist>0
model = linear_model.BayesianRidge()
if remove_zeros:
model.fit(basis[non_zero, :], hist[:,np.newaxis][non_zero,:])
else:
hist[~non_zero] = np.finfo(float).eps
model.fit(basis, hist[:,np.newaxis])
return model.predict(basis), hist, knots |
Determines knot placement based on a marginal distribution.
It places knots such that each knot covers the same amount
of probability mass. Two of the knots are reserved for the
borders which are treated seperatly. For example, a uniform
distribution with 5 knots will cause the knots to be equally
spaced with 25% of the probability mass between each two
knots.
Input:
marginal: Array
Estimate of the marginal distribution used to estimate
knot placement.
nr_knots: int
Number of knots to be placed.
spline_order: int
Order of the splines
Returns:
knots: Array
Sequence of knot positions
def knots_from_marginal(marginal, nr_knots, spline_order):
"""
Determines knot placement based on a marginal distribution.
It places knots such that each knot covers the same amount
of probability mass. Two of the knots are reserved for the
borders which are treated seperatly. For example, a uniform
distribution with 5 knots will cause the knots to be equally
spaced with 25% of the probability mass between each two
knots.
Input:
marginal: Array
Estimate of the marginal distribution used to estimate
knot placement.
nr_knots: int
Number of knots to be placed.
spline_order: int
Order of the splines
Returns:
knots: Array
Sequence of knot positions
"""
cumsum = np.cumsum(marginal)
cumsum = cumsum/cumsum.max()
borders = np.linspace(0,1,nr_knots)
knot_placement = [0] + np.unique([np.where(cumsum>=b)[0][0] for b in borders[1:-1]]).tolist() +[len(marginal)-1]
knots = augknt(knot_placement, spline_order)
return knots |
Computes a 1D spline basis
Input:
length: int
length of each basis
nr_knots: int
Number of knots, i.e. number of basis functions.
spline_order: int
Order of the splines.
marginal: array, optional
Estimate of the marginal distribution of the input to be fitted.
If given, it is used to determine the positioning of knots, each
knot will cover the same amount of probability mass. If not given,
knots are equally spaced.
def spline_base1d(length, nr_knots = 20, spline_order = 5, marginal = None):
"""Computes a 1D spline basis
Input:
length: int
length of each basis
nr_knots: int
Number of knots, i.e. number of basis functions.
spline_order: int
Order of the splines.
marginal: array, optional
Estimate of the marginal distribution of the input to be fitted.
If given, it is used to determine the positioning of knots, each
knot will cover the same amount of probability mass. If not given,
knots are equally spaced.
"""
if marginal is None:
knots = augknt(np.linspace(0,length+1, nr_knots), spline_order)
else:
knots = knots_from_marginal(marginal, nr_knots, spline_order)
x_eval = np.arange(1,length+1).astype(float)
Bsplines = spcol(x_eval,knots,spline_order)
return Bsplines, knots |
Computes a set of 2D spline basis functions.
The basis functions cover the entire space in height*width and can
for example be used to create fixation density maps.
Input:
width: int
width of each basis
height: int
height of each basis
nr_knots_x: int
of knots in x (width) direction.
nr_knots_y: int
of knots in y (height) direction.
spline_order: int
Order of the spline.
marginal_x: array, optional
Estimate of marginal distribution of the input to be fitted
along the x-direction (width). If given, it is used to determine
the positioning of knots, each knot will cover the same amount
of probability mass. If not given, knots are equally spaced.
marginal_y: array, optional
Marginal distribution along the y-direction (height). If
given, it is used to determine the positioning of knots.
Each knot will cover the same amount of probability mass.
Output:
basis: Matrix
Matrix of size n*(width*height) that contains in each row
one vectorized basis.
knots: Tuple
(x,y) are knot arrays that show the placement of knots.
def spline_base2d(width, height, nr_knots_x = 20.0, nr_knots_y = 20.0,
spline_order = 5, marginal_x = None, marginal_y = None):
"""Computes a set of 2D spline basis functions.
The basis functions cover the entire space in height*width and can
for example be used to create fixation density maps.
Input:
width: int
width of each basis
height: int
height of each basis
nr_knots_x: int
of knots in x (width) direction.
nr_knots_y: int
of knots in y (height) direction.
spline_order: int
Order of the spline.
marginal_x: array, optional
Estimate of marginal distribution of the input to be fitted
along the x-direction (width). If given, it is used to determine
the positioning of knots, each knot will cover the same amount
of probability mass. If not given, knots are equally spaced.
marginal_y: array, optional
Marginal distribution along the y-direction (height). If
given, it is used to determine the positioning of knots.
Each knot will cover the same amount of probability mass.
Output:
basis: Matrix
Matrix of size n*(width*height) that contains in each row
one vectorized basis.
knots: Tuple
(x,y) are knot arrays that show the placement of knots.
"""
if not (nr_knots_x<width and nr_knots_y<height):
raise RuntimeError("Too many knots for size of the base")
if marginal_x is None:
knots_x = augknt(np.linspace(0,width+1,nr_knots_x), spline_order)
else:
knots_x = knots_from_marginal(marginal_x, nr_knots_x, spline_order)
if marginal_y is None:
knots_y = augknt(np.linspace(0,height+1, nr_knots_y), spline_order)
else:
knots_y = knots_from_marginal(marginal_y, nr_knots_y, spline_order)
x_eval = np.arange(1,width+1).astype(float)
y_eval = np.arange(1,height+1).astype(float)
spline_setx = spcol(x_eval, knots_x, spline_order)
spline_sety = spcol(y_eval, knots_y, spline_order)
nr_coeff = [spline_sety.shape[1], spline_setx.shape[1]]
dim_bspline = [nr_coeff[0]*nr_coeff[1], len(x_eval)*len(y_eval)]
# construct 2D B-splines
nr_basis = 0
bspline = np.zeros(dim_bspline)
for IDX1 in range(0,nr_coeff[0]):
for IDX2 in range(0, nr_coeff[1]):
rand_coeff = np.zeros((nr_coeff[0] , nr_coeff[1]))
rand_coeff[IDX1,IDX2] = 1
tmp = np.dot(spline_sety,rand_coeff)
bspline[nr_basis,:] = np.dot(tmp,spline_setx.T).reshape((1,-1))
nr_basis = nr_basis+1
return bspline, (knots_x, knots_y) |
Computes a set of 3D spline basis functions.
For a description of the parameters see spline_base2d.
def spline_base3d( width, height, depth, nr_knots_x = 10.0, nr_knots_y = 10.0,
nr_knots_z=10, spline_order = 3, marginal_x = None, marginal_y = None,
marginal_z = None):
"""Computes a set of 3D spline basis functions.
For a description of the parameters see spline_base2d.
"""
if not nr_knots_z < depth:
raise RuntimeError("Too many knots for size of the base")
basis2d, (knots_x, knots_y) = spline_base2d(height, width, nr_knots_x,
nr_knots_y, spline_order, marginal_x, marginal_y)
if marginal_z is not None:
knots_z = knots_from_marginal(marginal_z, nr_knots_z, spline_order)
else:
knots_z = augknt(np.linspace(0,depth+1, nr_knots_z), spline_order)
z_eval = np.arange(1,depth+1).astype(float)
spline_setz = spcol(z_eval, knots_z, spline_order)
bspline = np.zeros((basis2d.shape[0]*len(z_eval), height*width*depth))
basis_nr = 0
for spline_a in spline_setz.T:
for spline_b in basis2d:
spline_b = spline_b.reshape((height, width))
bspline[basis_nr, :] = (spline_b[:,:,np.newaxis] * spline_a[:]).flat
basis_nr +=1
return bspline, (knots_x, knots_y, knots_z) |
Evaluates the ith spline basis given by knots on points in x
def spline(x,knots,p,i=0.0):
"""Evaluates the ith spline basis given by knots on points in x"""
assert(p+1<len(knots))
return np.array([N(float(u),float(i),float(p),knots) for u in x]) |
Computes the spline colocation matrix for knots in x.
The spline collocation matrix contains all m-p-1 bases
defined by knots. Specifically it contains the ith basis
in the ith column.
Input:
x: vector to evaluate the bases on
knots: vector of knots
spline_order: order of the spline
Output:
colmat: m x m-p matrix
The colocation matrix has size m x m-p where m
denotes the number of points the basis is evaluated
on and p is the spline order. The colums contain
the ith basis of knots evaluated on x.
def spcol(x,knots,spline_order):
"""Computes the spline colocation matrix for knots in x.
The spline collocation matrix contains all m-p-1 bases
defined by knots. Specifically it contains the ith basis
in the ith column.
Input:
x: vector to evaluate the bases on
knots: vector of knots
spline_order: order of the spline
Output:
colmat: m x m-p matrix
The colocation matrix has size m x m-p where m
denotes the number of points the basis is evaluated
on and p is the spline order. The colums contain
the ith basis of knots evaluated on x.
"""
colmat = np.nan*np.ones((len(x),len(knots) - spline_order-1))
for i in range(0,len(knots) - spline_order -1):
colmat[:,i] = spline(x,knots,spline_order,i)
return colmat |
Augment knot sequence such that some boundary conditions
are met.
def augknt(knots,order):
"""Augment knot sequence such that some boundary conditions
are met."""
a = []
[a.append(knots[0]) for t in range(0,order)]
[a.append(k) for k in knots]
[a.append(knots[-1]) for t in range(0,order)]
return np.array(a) |
Compute Spline Basis
Evaluates the spline basis of order p defined by knots
at knot i and point u.
def N(u,i,p,knots):
"""Compute Spline Basis
Evaluates the spline basis of order p defined by knots
at knot i and point u.
"""
if p == 0:
if knots[i] < u and u <=knots[i+1]:
return 1.0
else:
return 0.0
else:
try:
k = (( float((u-knots[i]))/float((knots[i+p] - knots[i]) ))
* N(u,i,p-1,knots))
except ZeroDivisionError:
k = 0.0
try:
q = (( float((knots[i+p+1] - u))/float((knots[i+p+1] - knots[i+1])))
* N(u,i+1,p-1,knots))
except ZeroDivisionError:
q = 0.0
return float(k + q) |
Evaluates a prediction against fixations in a fixmat with different measures.
The default measures which are used are AUC, NSS and KL-divergence. This
can be changed by setting the list of measures with set_scores.
As different measures need potentially different parameters, the kw
dictionary can be used to pass arguments to measures. Every named
argument (except fm and prediction) of a measure that is included in
kw.keys() will be filled with the value stored in kw.
Example:
>>> prediction_scores(P, FM, ctr_loc = (y,x))
In this case the AUC will be computed with control points (y,x), because
the measure 'roc_model' has 'ctr_loc' as named argument.
Input:
prediction : 2D numpy array
The prediction that should be evaluated
fm : Fixmat
The eyetracking data to evaluate against
Output:
Tuple of prediction scores. The order of the scores is determined
by order of measures.scores.
def prediction_scores(prediction, fm, **kw):
"""
Evaluates a prediction against fixations in a fixmat with different measures.
The default measures which are used are AUC, NSS and KL-divergence. This
can be changed by setting the list of measures with set_scores.
As different measures need potentially different parameters, the kw
dictionary can be used to pass arguments to measures. Every named
argument (except fm and prediction) of a measure that is included in
kw.keys() will be filled with the value stored in kw.
Example:
>>> prediction_scores(P, FM, ctr_loc = (y,x))
In this case the AUC will be computed with control points (y,x), because
the measure 'roc_model' has 'ctr_loc' as named argument.
Input:
prediction : 2D numpy array
The prediction that should be evaluated
fm : Fixmat
The eyetracking data to evaluate against
Output:
Tuple of prediction scores. The order of the scores is determined
by order of measures.scores.
"""
if prediction == None:
return [np.NaN for measure in scores]
results = []
for measure in scores:
(args, _, _, _) = inspect.getargspec(measure)
if len(args)>2:
# Filter dictionary, such that only the keys that are
# expected by the measure are in it
mdict = {}
[mdict.update({key:value}) for (key, value) in list(kw.items())
if key in args]
score = measure(prediction, fm, **mdict)
else:
score = measure(prediction, fm)
results.append(score)
return results |
wraps kldiv functionality for model evaluation
input:
prediction: 2D matrix
the model salience map
fm : fixmat
Should be filtered for the image corresponding to the prediction
def kldiv_model(prediction, fm):
"""
wraps kldiv functionality for model evaluation
input:
prediction: 2D matrix
the model salience map
fm : fixmat
Should be filtered for the image corresponding to the prediction
"""
(_, r_x) = calc_resize_factor(prediction, fm.image_size)
q = np.array(prediction, copy=True)
q -= np.min(q.flatten())
q /= np.sum(q.flatten())
return kldiv(None, q, distp = fm, scale_factor = r_x) |
Computes the Kullback-Leibler divergence between two distributions.
Parameters
p : Matrix
The first probability distribution
q : Matrix
The second probability distribution
distp : fixmat
If p is None, distp is used to compute a FDM which
is then taken as 1st probability distribution.
distq : fixmat
If q is None, distq is used to compute a FDM which is
then taken as 2dn probability distribution.
scale_factor : double
Determines the size of FDM computed from distq or distp.
def kldiv(p, q, distp = None, distq = None, scale_factor = 1):
"""
Computes the Kullback-Leibler divergence between two distributions.
Parameters
p : Matrix
The first probability distribution
q : Matrix
The second probability distribution
distp : fixmat
If p is None, distp is used to compute a FDM which
is then taken as 1st probability distribution.
distq : fixmat
If q is None, distq is used to compute a FDM which is
then taken as 2dn probability distribution.
scale_factor : double
Determines the size of FDM computed from distq or distp.
"""
assert q != None or distq != None, "Either q or distq have to be given"
assert p != None or distp != None, "Either p or distp have to be given"
try:
if p == None:
p = compute_fdm(distp, scale_factor = scale_factor)
if q == None:
q = compute_fdm(distq, scale_factor = scale_factor)
except RuntimeError:
return np.NaN
q += np.finfo(q.dtype).eps
p += np.finfo(p.dtype).eps
kl = np.sum( p * (np.log2(p / q)))
return kl |
Computes Chao-Shen corrected KL-divergence between prediction
and fdm made from fixations in fm.
Parameters :
prediction : np.ndarray
a fixation density map
fm : FixMat object
def kldiv_cs_model(prediction, fm):
"""
Computes Chao-Shen corrected KL-divergence between prediction
and fdm made from fixations in fm.
Parameters :
prediction : np.ndarray
a fixation density map
fm : FixMat object
"""
# compute histogram of fixations needed for ChaoShen corrected kl-div
# image category must exist (>-1) and image_size must be non-empty
assert(len(fm.image_size) == 2 and (fm.image_size[0] > 0) and
(fm.image_size[1] > 0))
assert(-1 not in fm.category)
# check whether fixmat contains fixations
if len(fm.x) == 0:
return np.NaN
(scale_factor, _) = calc_resize_factor(prediction, fm.image_size)
# this specifies left edges of the histogram bins, i.e. fixations between
# ]0 binedge[0]] are included. --> fixations are ceiled
e_y = np.arange(0, np.round(scale_factor*fm.image_size[0]+1))
e_x = np.arange(0, np.round(scale_factor*fm.image_size[1]+1))
samples = np.array(list(zip((scale_factor*fm.y), (scale_factor*fm.x))))
(fdm, _) = np.histogramdd(samples, (e_y, e_x))
# compute ChaoShen corrected kl-div
q = np.array(prediction, copy = True)
q[q == 0] = np.finfo(q.dtype).eps
q /= np.sum(q)
(H, pa, la) = chao_shen(fdm)
q = q[fdm > 0]
cross_entropy = -np.sum((pa * np.log2(q)) / la)
return (cross_entropy - H) |
Computes some terms needed for the Chao-Shen KL correction.
def chao_shen(q):
"""
Computes some terms needed for the Chao-Shen KL correction.
"""
yx = q[q > 0] # remove bins with zero counts
n = np.sum(yx)
p = yx.astype(float)/n
f1 = np.sum(yx == 1) # number of singletons in the sample
if f1 == n: # avoid C == 0
f1 -= 1
C = 1 - (f1/n) # estimated coverage of the sample
pa = C * p # coverage adjusted empirical frequencies
la = (1 - (1 - pa) ** n) # probability to see a bin (species) in the sample
H = -np.sum((pa * np.log2(pa)) / la)
return (H, pa, la) |
wraps numpy.corrcoef functionality for model evaluation
input:
prediction: 2D Matrix
the model salience map
fm: fixmat
Used to compute a FDM to which the prediction is compared.
def correlation_model(prediction, fm):
"""
wraps numpy.corrcoef functionality for model evaluation
input:
prediction: 2D Matrix
the model salience map
fm: fixmat
Used to compute a FDM to which the prediction is compared.
"""
(_, r_x) = calc_resize_factor(prediction, fm.image_size)
fdm = compute_fdm(fm, scale_factor = r_x)
return np.corrcoef(fdm.flatten(), prediction.flatten())[0,1] |
wraps nss functionality for model evaluation
input:
prediction: 2D matrix
the model salience map
fm : fixmat
Fixations that define the actuals
def nss_model(prediction, fm):
"""
wraps nss functionality for model evaluation
input:
prediction: 2D matrix
the model salience map
fm : fixmat
Fixations that define the actuals
"""
(r_y, r_x) = calc_resize_factor(prediction, fm.image_size)
fix = ((np.array(fm.y-1)*r_y).astype(int),
(np.array(fm.x-1)*r_x).astype(int))
return nss(prediction, fix) |
Compute the normalized scanpath salience
input:
fix : list, l[0] contains y, l[1] contains x
def nss(prediction, fix):
"""
Compute the normalized scanpath salience
input:
fix : list, l[0] contains y, l[1] contains x
"""
prediction = prediction - np.mean(prediction)
prediction = prediction / np.std(prediction)
return np.mean(prediction[fix[0], fix[1]]) |
wraps roc functionality for model evaluation
Parameters:
prediction: 2D array
the model salience map
fm : fixmat
Fixations that define locations of the actuals
ctr_loc : tuple of (y.x) coordinates, optional
Allows to specify control points for spatial
bias correction
ctr_size : two element tuple, optional
Specifies the assumed image size of the control locations,
defaults to fm.image_size
def roc_model(prediction, fm, ctr_loc = None, ctr_size = None):
"""
wraps roc functionality for model evaluation
Parameters:
prediction: 2D array
the model salience map
fm : fixmat
Fixations that define locations of the actuals
ctr_loc : tuple of (y.x) coordinates, optional
Allows to specify control points for spatial
bias correction
ctr_size : two element tuple, optional
Specifies the assumed image size of the control locations,
defaults to fm.image_size
"""
# check if prediction is a valid numpy array
assert type(prediction) == np.ndarray
# check whether scaling preserved aspect ratio
(r_y, r_x) = calc_resize_factor(prediction, fm.image_size)
# read out values in the fdm at actual fixation locations
# .astype(int) floors numbers in np.array
y_index = (r_y * np.array(fm.y-1)).astype(int)
x_index = (r_x * np.array(fm.x-1)).astype(int)
actuals = prediction[y_index, x_index]
if not ctr_loc:
xc = np.random.randint(0, prediction.shape[1], 1000)
yc = np.random.randint(0, prediction.shape[0], 1000)
ctr_loc = (yc.astype(int), xc.astype(int))
else:
if not ctr_size:
ctr_size = fm.image_size
else:
(r_y, r_x) = calc_resize_factor(prediction, ctr_size)
ctr_loc = ((r_y * np.array(ctr_loc[0])).astype(int),
(r_x * np.array(ctr_loc[1])).astype(int))
controls = prediction[ctr_loc[0], ctr_loc[1]]
return fast_roc(actuals, controls)[0] |
approximates the area under the roc curve for sets of actuals and controls.
Uses all values appearing in actuals as thresholds and lower sum
interpolation. Also returns arrays of the true positive rate and the false
positive rate that can be used for plotting the roc curve.
Parameters:
actuals : list
A list of numeric values for positive observations.
controls : list
A list of numeric values for negative observations.
def fast_roc(actuals, controls):
"""
approximates the area under the roc curve for sets of actuals and controls.
Uses all values appearing in actuals as thresholds and lower sum
interpolation. Also returns arrays of the true positive rate and the false
positive rate that can be used for plotting the roc curve.
Parameters:
actuals : list
A list of numeric values for positive observations.
controls : list
A list of numeric values for negative observations.
"""
assert(type(actuals) is np.ndarray)
assert(type(controls) is np.ndarray)
actuals = np.ravel(actuals)
controls = np.ravel(controls)
if np.isnan(actuals).any():
raise RuntimeError('NaN found in actuals')
if np.isnan(controls).any():
raise RuntimeError('NaN found in controls')
thresholds = np.hstack([-np.inf, np.unique(actuals), np.inf])[::-1]
true_pos_rate = np.empty(thresholds.size)
false_pos_rate = np.empty(thresholds.size)
num_act = float(len(actuals))
num_ctr = float(len(controls))
for i, value in enumerate(thresholds):
true_pos_rate[i] = (actuals >= value).sum() / num_act
false_pos_rate[i] = (controls >= value).sum() / num_ctr
auc = np.dot(np.diff(false_pos_rate), true_pos_rate[0:-1])
# treat cases where TPR of one is not reached before FPR of one
# by using trapezoidal integration for the last segment
# (add the missing triangle)
if false_pos_rate[-2] == 1:
auc += ((1-true_pos_rate[-3])*.5*(1-false_pos_rate[-3]))
return (auc, true_pos_rate, false_pos_rate) |
Histogram based implementation of AUC unde ROC curve.
Parameters:
actuals : list
A list of numeric values for positive observations.
controls : list
A list of numeric values for negative observations.
def faster_roc(actuals, controls):
"""
Histogram based implementation of AUC unde ROC curve.
Parameters:
actuals : list
A list of numeric values for positive observations.
controls : list
A list of numeric values for negative observations.
"""
assert(type(actuals) is np.ndarray)
assert(type(controls) is np.ndarray)
if len(actuals)<500:
raise RuntimeError('This method might be incorrect when '+
'not enough actuals are present. Needs to be checked before '+
'proceeding. Stopping here for you to do so.')
actuals = np.ravel(actuals)
controls = np.ravel(controls)
if np.isnan(actuals).any():
raise RuntimeError('NaN found in actuals')
if np.isnan(controls).any():
raise RuntimeError('NaN found in controls')
thresholds = np.hstack([-np.inf, np.unique(actuals), np.inf])+np.finfo(float).eps
true_pos_rate = np.nan*np.empty(thresholds.size-1)
false_pos_rate = np.nan*np.empty(thresholds.size-1)
num_act = float(len(actuals))
num_ctr = float(len(controls))
actuals = 1-(np.cumsum(np.histogram(actuals, thresholds)[0])/num_act)
controls = 1-(np.cumsum(np.histogram(controls, thresholds)[0])/num_ctr)
true_pos_rate = actuals
false_pos_rate = controls
#true_pos_rate = np.concatenate(([0], true_pos_rate, [1]))
false_pos_rate = false_pos_rate
auc = -1*np.dot(np.diff(false_pos_rate), true_pos_rate[0:-1])
# treat cases where TPR of one is not reached before FPR of one
# by using trapezoidal integration for the last segment
# (add the missing triangle)
if false_pos_rate[-2] == 1:
auc += ((1-true_pos_rate[-3])*.5*(1-false_pos_rate[-3]))
return (auc, true_pos_rate, false_pos_rate) |
wraps emd functionality for model evaluation
requires:
OpenCV python bindings
input:
prediction: the model salience map
fm : fixmat filtered for the image corresponding to the prediction
def emd_model(prediction, fm):
"""
wraps emd functionality for model evaluation
requires:
OpenCV python bindings
input:
prediction: the model salience map
fm : fixmat filtered for the image corresponding to the prediction
"""
(_, r_x) = calc_resize_factor(prediction, fm.image_size)
gt = fixmat.compute_fdm(fm, scale_factor = r_x)
return emd(prediction, gt) |
Compute the Eart Movers Distance between prediction and model.
This implementation uses opencv for doing the actual work.
Unfortunately, at the time of implementation only the SWIG
bindings werer available and the numpy arrays have to
converted by hand. This changes with opencv 2.1.
def emd(prediction, ground_truth):
"""
Compute the Eart Movers Distance between prediction and model.
This implementation uses opencv for doing the actual work.
Unfortunately, at the time of implementation only the SWIG
bindings werer available and the numpy arrays have to
converted by hand. This changes with opencv 2.1. """
import opencv
if not (prediction.shape == ground_truth.shape):
raise RuntimeError('Shapes of prediction and ground truth have' +
' to be equal. They are: %s, %s'
%(str(prediction.shape), str(ground_truth.shape)))
(x, y) = np.meshgrid(list(range(0, prediction.shape[1])),
list(range(0, prediction.shape[0])))
s1 = np.array([x.flatten(), y.flatten(), prediction.flatten()]).T
s2 = np.array([x.flatten(), y.flatten(), ground_truth.flatten()]).T
s1m = opencv.cvCreateMat(s1.shape[0], s2.shape[1], opencv.CV_32FC1)
s2m = opencv.cvCreateMat(s1.shape[0], s2.shape[1], opencv.CV_32FC1)
for r in range(0, s1.shape[0]):
for c in range(0, s1.shape[1]):
s1m[r, c] = float(s1[r, c])
s2m[r, c] = float(s2[r, c])
d = opencv.cvCalcEMD2(s1m, s2m, opencv.CV_DIST_L2)
return d |
Parse RFC 822 dates and times
http://tools.ietf.org/html/rfc822#section-5
There are some formatting differences that are accounted for:
1. Years may be two or four digits.
2. The month and day can be swapped.
3. Additional timezone names are supported.
4. A default time and timezone are assumed if only a date is present.
5.
def _rfc822(date):
"""Parse RFC 822 dates and times
http://tools.ietf.org/html/rfc822#section-5
There are some formatting differences that are accounted for:
1. Years may be two or four digits.
2. The month and day can be swapped.
3. Additional timezone names are supported.
4. A default time and timezone are assumed if only a date is present.
5.
"""
daynames = set(['mon', 'tue', 'wed', 'thu', 'fri', 'sat', 'sun'])
months = {
'jan': 1, 'feb': 2, 'mar': 3, 'apr': 4, 'may': 5, 'jun': 6,
'jul': 7, 'aug': 8, 'sep': 9, 'oct': 10, 'nov': 11, 'dec': 12,
}
timezonenames = {
'ut': 0, 'gmt': 0, 'z': 0,
'adt': -3, 'ast': -4, 'at': -4,
'edt': -4, 'est': -5, 'et': -5,
'cdt': -5, 'cst': -6, 'ct': -6,
'mdt': -6, 'mst': -7, 'mt': -7,
'pdt': -7, 'pst': -8, 'pt': -8,
'a': -1, 'n': 1,
'm': -12, 'y': 12,
}
parts = date.lower().split()
if len(parts) < 5:
# Assume that the time and timezone are missing
parts.extend(('00:00:00', '0000'))
# Remove the day name
if parts[0][:3] in daynames:
parts = parts[1:]
if len(parts) < 5:
# If there are still fewer than five parts, there's not enough
# information to interpret this
return None
try:
day = int(parts[0])
except ValueError:
# Check if the day and month are swapped
if months.get(parts[0][:3]):
try:
day = int(parts[1])
except ValueError:
return None
else:
parts[1] = parts[0]
else:
return None
month = months.get(parts[1][:3])
if not month:
return None
try:
year = int(parts[2])
except ValueError:
return None
# Normalize two-digit years:
# Anything in the 90's is interpreted as 1990 and on
# Anything 89 or less is interpreted as 2089 or before
if len(parts[2]) <= 2:
year += (1900, 2000)[year < 90]
timeparts = parts[3].split(':')
timeparts = timeparts + ([0] * (3 - len(timeparts)))
try:
(hour, minute, second) = map(int, timeparts)
except ValueError:
return None
tzhour = 0
tzmin = 0
# Strip 'Etc/' from the timezone
if parts[4].startswith('etc/'):
parts[4] = parts[4][4:]
# Normalize timezones that start with 'gmt':
# GMT-05:00 => -0500
# GMT => GMT
if parts[4].startswith('gmt'):
parts[4] = ''.join(parts[4][3:].split(':')) or 'gmt'
# Handle timezones like '-0500', '+0500', and 'EST'
if parts[4] and parts[4][0] in ('-', '+'):
try:
tzhour = int(parts[4][1:3])
tzmin = int(parts[4][3:])
except ValueError:
return None
if parts[4].startswith('-'):
tzhour = tzhour * -1
tzmin = tzmin * -1
else:
tzhour = timezonenames.get(parts[4], 0)
# Create the datetime object and timezone delta objects
try:
stamp = datetime.datetime(year, month, day, hour, minute, second)
except ValueError:
return None
delta = datetime.timedelta(0, 0, 0, 0, tzmin, tzhour)
# Return the date and timestamp in a UTC 9-tuple
try:
return stamp - delta
except OverflowError:
return None |
_to_rfc822(datetime.datetime) -> str
The datetime `strftime` method is subject to locale-specific
day and month names, so this function hardcodes the conversion.
def _to_rfc822(date):
"""_to_rfc822(datetime.datetime) -> str
The datetime `strftime` method is subject to locale-specific
day and month names, so this function hardcodes the conversion."""
months = ['Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun',
'Jul', 'Aug', 'Sep', 'Oct', 'Nov', 'Dec']
days = ['Mon', 'Tue', 'Wed', 'Thu', 'Fri', 'Sat', 'Sun']
fmt = '{day}, {d:02} {month} {y:04} {h:02}:{m:02}:{s:02} GMT'
return fmt.format(
day=days[date.weekday()],
d=date.day,
month=months[date.month - 1],
y=date.year,
h=date.hour,
m=date.minute,
s=date.second,
) |
Formats the SQL query to use ordinal parameters instead of named
parameters.
*sql* (|string|) is the SQL query.
*params* (|dict|) maps each named parameter (|string|) to value
(|object|). If |self.named| is "numeric", then *params* can be
simply a |sequence| of values mapped by index.
Returns a 2-|tuple| containing: the formatted SQL query (|string|),
and the ordinal parameters (|list|).
def format(self, sql, params):
"""
Formats the SQL query to use ordinal parameters instead of named
parameters.
*sql* (|string|) is the SQL query.
*params* (|dict|) maps each named parameter (|string|) to value
(|object|). If |self.named| is "numeric", then *params* can be
simply a |sequence| of values mapped by index.
Returns a 2-|tuple| containing: the formatted SQL query (|string|),
and the ordinal parameters (|list|).
"""
if isinstance(sql, unicode):
string_type = unicode
elif isinstance(sql, bytes):
string_type = bytes
sql = sql.decode(_BYTES_ENCODING)
else:
raise TypeError("sql:{!r} is not a unicode or byte string.".format(sql))
if self.named == 'numeric':
if isinstance(params, collections.Mapping):
params = {string_type(idx): val for idx, val in iteritems(params)}
elif isinstance(params, collections.Sequence) and not isinstance(params, (unicode, bytes)):
params = {string_type(idx): val for idx, val in enumerate(params, 1)}
if not isinstance(params, collections.Mapping):
raise TypeError("params:{!r} is not a dict.".format(params))
# Find named parameters.
names = self.match.findall(sql)
# Map named parameters to ordinals.
ord_params = []
name_to_ords = {}
for name in names:
value = params[name]
if isinstance(value, tuple):
ord_params.extend(value)
if name not in name_to_ords:
name_to_ords[name] = '(' + ','.join((self.replace,) * len(value)) + ')'
else:
ord_params.append(value)
if name not in name_to_ords:
name_to_ords[name] = self.replace
# Replace named parameters with ordinals.
sql = self.match.sub(lambda m: name_to_ords[m.group(1)], sql)
# Make sure the query is returned as the proper string type.
if string_type is bytes:
sql = sql.encode(_BYTES_ENCODING)
# Return formatted SQL and new ordinal parameters.
return sql, ord_params |
Formats the SQL query to use ordinal parameters instead of named
parameters.
*sql* (|string|) is the SQL query.
*many_params* (|iterable|) contains each *params* to format.
- *params* (|dict|) maps each named parameter (|string|) to value
(|object|). If |self.named| is "numeric", then *params* can be
simply a |sequence| of values mapped by index.
Returns a 2-|tuple| containing: the formatted SQL query (|string|),
and a |list| containing each ordinal parameters (|list|).
def formatmany(self, sql, many_params):
"""
Formats the SQL query to use ordinal parameters instead of named
parameters.
*sql* (|string|) is the SQL query.
*many_params* (|iterable|) contains each *params* to format.
- *params* (|dict|) maps each named parameter (|string|) to value
(|object|). If |self.named| is "numeric", then *params* can be
simply a |sequence| of values mapped by index.
Returns a 2-|tuple| containing: the formatted SQL query (|string|),
and a |list| containing each ordinal parameters (|list|).
"""
if isinstance(sql, unicode):
string_type = unicode
elif isinstance(sql, bytes):
string_type = bytes
sql = sql.decode(_BYTES_ENCODING)
else:
raise TypeError("sql:{!r} is not a unicode or byte string.".format(sql))
if not isinstance(many_params, collections.Iterable) or isinstance(many_params, (unicode, bytes)):
raise TypeError("many_params:{!r} is not iterable.".format(many_params))
# Find named parameters.
names = self.match.findall(sql)
name_set = set(names)
# Map named parameters to ordinals.
many_ord_params = []
name_to_ords = {}
name_to_len = {}
repl_str = self.replace
repl_tuple = (repl_str,)
for i, params in enumerate(many_params):
if self.named == 'numeric':
if isinstance(params, collections.Mapping):
params = {string_type(idx): val for idx, val in iteritems(params)}
elif isinstance(params, collections.Sequence) and not isinstance(params, (unicode, bytes)):
params = {string_type(idx): val for idx, val in enumerate(params, 1)}
if not isinstance(params, collections.Mapping):
raise TypeError("many_params[{}]:{!r} is not a dict.".format(i, params))
if not i: # first
# Map names to ordinals, and determine what names are tuples and
# what their lengths are.
for name in name_set:
value = params[name]
if isinstance(value, tuple):
tuple_len = len(value)
name_to_ords[name] = '(' + ','.join(repl_tuple * tuple_len) + ')'
name_to_len[name] = tuple_len
else:
name_to_ords[name] = repl_str
name_to_len[name] = None
# Make sure tuples match up and collapse tuples into ordinals.
ord_params = []
for name in names:
value = params[name]
tuple_len = name_to_len[name]
if tuple_len is not None:
if not isinstance(value, tuple):
raise TypeError("many_params[{}][{!r}]:{!r} was expected to be a tuple.".format(i, name, value))
elif len(value) != tuple_len:
raise ValueError("many_params[{}][{!r}]:{!r} length was expected to be {}.".format(i, name, value, tuple_len))
ord_params.extend(value)
else:
ord_params.append(value)
many_ord_params.append(ord_params)
# Replace named parameters with ordinals.
sql = self.match.sub(lambda m: name_to_ords[m.group(1)], sql)
# Make sure the query is returned as the proper string type.
if string_type is bytes:
sql = sql.encode(_BYTES_ENCODING)
# Return formatted SQL and new ordinal parameters.
return sql, many_ord_params |
Gets the parser for the command f, if it not exists it creates a new one
def _get_parser(f):
"""
Gets the parser for the command f, if it not exists it creates a new one
"""
_COMMAND_GROUPS[f.__module__].load()
if f.__name__ not in _COMMAND_GROUPS[f.__module__].parsers:
parser = _COMMAND_GROUPS[f.__module__].parser_generator.add_parser(f.__name__, help=f.__doc__,
description=f.__doc__)
parser.set_defaults(func=f)
_COMMAND_GROUPS[f.__module__].parsers[f.__name__] = parser
return _COMMAND_GROUPS[f.__module__].parsers[f.__name__] |
Find all <a /> elements in the given html for a post. Only scan html element matching all criteria in look_in.
optionally the content to be scanned can be given as an argument.
If any have an href attribute that is not from the
one of the items in exclude_domains, append it to our lists.
:param sourceURL: the URL for the post we are scanning
:param exclude_domains: a list of domains to exclude from the search
:type exclude_domains: list
:param content: the content to be scanned for mentions
:param look_in: dictionary with name, id and class_. only element matching all of these will be scanned
:param test_urls: optional flag to test URLs for validation
:param headers: optional headers to send with any web requests
:type headers: dict
:param timeout: optional timeout for web requests
:type timeout float
:rtype: dictionary of Mentions
def findMentions(sourceURL, targetURL=None, exclude_domains=[], content=None, test_urls=True, headers={}, timeout=None):
"""Find all <a /> elements in the given html for a post. Only scan html element matching all criteria in look_in.
optionally the content to be scanned can be given as an argument.
If any have an href attribute that is not from the
one of the items in exclude_domains, append it to our lists.
:param sourceURL: the URL for the post we are scanning
:param exclude_domains: a list of domains to exclude from the search
:type exclude_domains: list
:param content: the content to be scanned for mentions
:param look_in: dictionary with name, id and class_. only element matching all of these will be scanned
:param test_urls: optional flag to test URLs for validation
:param headers: optional headers to send with any web requests
:type headers: dict
:param timeout: optional timeout for web requests
:type timeout float
:rtype: dictionary of Mentions
"""
__doc__ = None
if test_urls:
URLValidator(message='invalid source URL')(sourceURL)
if content:
result = {'status': requests.codes.ok,
'headers': None,
}
else:
r = requests.get(sourceURL, verify=True, headers=headers, timeout=timeout)
result = {'status': r.status_code,
'headers': r.headers
}
# Check for character encodings and use 'correct' data
if 'charset' in r.headers.get('content-type', ''):
content = r.text
else:
content = r.content
result.update({'refs': set(), 'post-url': sourceURL})
if result['status'] == requests.codes.ok:
# Allow passing BS doc as content
if isinstance(content, BeautifulSoup):
__doc__ = content
# result.update({'content': unicode(__doc__)})
result.update({'content': str(__doc__)})
else:
__doc__ = BeautifulSoup(content, _html_parser)
result.update({'content': content})
# try to find first h-entry else use full document
entry = __doc__.find(class_="h-entry") or __doc__
# Allow finding particular URL
if targetURL:
# find only targetURL
all_links = entry.find_all('a', href=targetURL)
else:
# find all links with a href
all_links = entry.find_all('a', href=True)
for link in all_links:
href = link.get('href', None)
if href:
url = urlparse(href)
if url.scheme in ('http', 'https'):
if url.hostname and url.hostname not in exclude_domains:
result['refs'].add(href)
return result |
Search the given html content for all <link /> elements
and return any discovered WebMention URL.
:param html: html content
:rtype: WebMention URL
def findEndpoint(html):
"""Search the given html content for all <link /> elements
and return any discovered WebMention URL.
:param html: html content
:rtype: WebMention URL
"""
poss_rels = ['webmention', 'http://webmention.org', 'http://webmention.org/', 'https://webmention.org', 'https://webmention.org/']
# find elements with correct rels and a href value
all_links = BeautifulSoup(html, _html_parser).find_all(rel=poss_rels, href=True)
for link in all_links:
s = link.get('href', None)
if s is not None:
return s
return None |
Discover any WebMention endpoint for a given URL.
:param link: URL to discover WebMention endpoint
:param test_urls: optional flag to test URLs for validation
:param headers: optional headers to send with any web requests
:type headers dict
:param timeout: optional timeout for web requests
:type timeout float
:param request: optional Requests request object to avoid another GET
:rtype: tuple (status_code, URL, [debug])
def discoverEndpoint(url, test_urls=True, headers={}, timeout=None, request=None, debug=False):
"""Discover any WebMention endpoint for a given URL.
:param link: URL to discover WebMention endpoint
:param test_urls: optional flag to test URLs for validation
:param headers: optional headers to send with any web requests
:type headers dict
:param timeout: optional timeout for web requests
:type timeout float
:param request: optional Requests request object to avoid another GET
:rtype: tuple (status_code, URL, [debug])
"""
if test_urls:
URLValidator(message='invalid URL')(url)
# status, webmention
endpointURL = None
debugOutput = []
try:
if request is not None:
targetRequest = request
else:
targetRequest = requests.get(url, verify=False, headers=headers, timeout=timeout)
returnCode = targetRequest.status_code
debugOutput.append('%s %s' % (returnCode, url))
if returnCode == requests.codes.ok:
try:
linkHeader = parse_link_header(targetRequest.headers['link'])
endpointURL = linkHeader.get('webmention', '') or \
linkHeader.get('http://webmention.org', '') or \
linkHeader.get('http://webmention.org/', '') or \
linkHeader.get('https://webmention.org', '') or \
linkHeader.get('https://webmention.org/', '')
# force searching in the HTML if not found
if not endpointURL:
raise AttributeError
debugOutput.append('found in link headers')
except (KeyError, AttributeError):
endpointURL = findEndpoint(targetRequest.text)
debugOutput.append('found in body')
if endpointURL is not None:
endpointURL = urljoin(url, endpointURL)
except (requests.exceptions.RequestException, requests.exceptions.ConnectionError,
requests.exceptions.HTTPError, requests.exceptions.URLRequired,
requests.exceptions.TooManyRedirects, requests.exceptions.Timeout):
debugOutput.append('exception during GET request')
returnCode = 500
debugOutput.append('endpointURL: %s %s' % (returnCode, endpointURL))
if debug:
return (returnCode, endpointURL, debugOutput)
else:
return (returnCode, endpointURL) |
Send to the :targetURL: a WebMention for the :sourceURL:
The WebMention will be discovered if not given in the :webmention:
parameter.
:param sourceURL: URL that is referencing :targetURL:
:param targetURL: URL of mentioned post
:param webmention: optional WebMention endpoint
:param test_urls: optional flag to test URLs for validation
:param headers: optional headers to send with any web requests
:type headers dict
:param timeout: optional timeout for web requests
:type timeout float
:rtype: HTTPrequest object if WebMention endpoint was valid
def sendWebmention(sourceURL, targetURL, webmention=None, test_urls=True, vouchDomain=None,
headers={}, timeout=None, debug=False):
"""Send to the :targetURL: a WebMention for the :sourceURL:
The WebMention will be discovered if not given in the :webmention:
parameter.
:param sourceURL: URL that is referencing :targetURL:
:param targetURL: URL of mentioned post
:param webmention: optional WebMention endpoint
:param test_urls: optional flag to test URLs for validation
:param headers: optional headers to send with any web requests
:type headers dict
:param timeout: optional timeout for web requests
:type timeout float
:rtype: HTTPrequest object if WebMention endpoint was valid
"""
if test_urls:
v = URLValidator()
v(sourceURL)
v(targetURL)
debugOutput = []
originalURL = targetURL
try:
targetRequest = requests.get(targetURL)
if targetRequest.status_code == requests.codes.ok:
if len(targetRequest.history) > 0:
redirect = targetRequest.history[-1]
if (redirect.status_code == 301 or redirect.status_code == 302) and 'Location' in redirect.headers:
targetURL = urljoin(targetURL, redirect.headers['Location'])
debugOutput.append('targetURL redirected: %s' % targetURL)
if webmention is None:
wStatus, wUrl = discoverEndpoint(targetURL, headers=headers, timeout=timeout, request=targetRequest)
else:
wStatus = 200
wUrl = webmention
debugOutput.append('endpointURL: %s %s' % (wStatus, wUrl))
if wStatus == requests.codes.ok and wUrl is not None:
if test_urls:
v(wUrl)
payload = {'source': sourceURL,
'target': originalURL}
if vouchDomain is not None:
payload['vouch'] = vouchDomain
try:
result = requests.post(wUrl, data=payload, headers=headers, timeout=timeout)
debugOutput.append('POST %s -- %s' % (wUrl, result.status_code))
if result.status_code == 405 and len(result.history) > 0:
redirect = result.history[-1]
if redirect.status_code == 301 and 'Location' in redirect.headers:
result = requests.post(redirect.headers['Location'], data=payload, headers=headers, timeout=timeout)
debugOutput.append('redirected POST %s -- %s' % (redirect.headers['Location'], result.status_code))
except Exception as e:
result = None
except (requests.exceptions.RequestException, requests.exceptions.ConnectionError,
requests.exceptions.HTTPError, requests.exceptions.URLRequired,
requests.exceptions.TooManyRedirects, requests.exceptions.Timeout):
debugOutput.append('exception during GET request')
result = None
return result |
takes the link header as a string and returns a dictionary with rel values as keys and urls as values
:param link: link header as a string
:rtype: dictionary {rel_name: rel_value}
def parse_link_header(link):
"""takes the link header as a string and returns a dictionary with rel values as keys and urls as values
:param link: link header as a string
:rtype: dictionary {rel_name: rel_value}
"""
rel_dict = {}
for rels in link.split(','):
rel_break = quoted_split(rels, ';')
try:
rel_url = re.search('<(.+?)>', rel_break[0]).group(1)
rel_names = quoted_split(rel_break[1], '=')[-1]
if rel_names.startswith('"') and rel_names.endswith('"'):
rel_names = rel_names[1:-1]
for name in rel_names.split():
rel_dict[name] = rel_url
except (AttributeError, IndexError):
pass
return rel_dict |
Find all <a /> elements in the given html for a post.
If any have an href attribute that is rel="me" then include
it in the result.
:param sourceURL: the URL for the post we are scanning
:rtype: dictionary of RelMe references
def findRelMe(sourceURL):
"""Find all <a /> elements in the given html for a post.
If any have an href attribute that is rel="me" then include
it in the result.
:param sourceURL: the URL for the post we are scanning
:rtype: dictionary of RelMe references
"""
r = requests.get(sourceURL)
result = {'status': r.status_code,
'headers': r.headers,
'history': r.history,
'content': r.text,
'relme': [],
'url': sourceURL
}
if r.status_code == requests.codes.ok:
dom = BeautifulSoup(r.text, _html_parser)
for link in dom.find_all('a', rel='me'):
rel = link.get('rel')
href = link.get('href')
if rel is not None and href is not None:
url = urlparse(href)
if url is not None and url.scheme in ('http', 'https'):
result['relme'].append(cleanURL(href))
return result |
Determine if a given :resourceURL: is authoritative for the :profileURL:
TODO add https/http filtering for those who wish to limit/restrict urls to match fully
TODO add code to ensure that each item in the redirect chain is authoritative
:param profileURL: URL of the user
:param resourceURL: URL of the resource to validate
:param profileRelMes: optional list of rel="me" links within the profile URL
:param resourceRelMes: optional list of rel="me" links found within resource URL
:rtype: True if confirmed
def confirmRelMe(profileURL, resourceURL, profileRelMes=None, resourceRelMes=None):
"""Determine if a given :resourceURL: is authoritative for the :profileURL:
TODO add https/http filtering for those who wish to limit/restrict urls to match fully
TODO add code to ensure that each item in the redirect chain is authoritative
:param profileURL: URL of the user
:param resourceURL: URL of the resource to validate
:param profileRelMes: optional list of rel="me" links within the profile URL
:param resourceRelMes: optional list of rel="me" links found within resource URL
:rtype: True if confirmed
"""
result = False
profile = normalizeURL(profileURL)
if profileRelMes is None:
profileRelMe = findRelMe(profileURL)
profileRelMes = profileRelMe['relme']
if resourceRelMes is None:
resourceRelMe = findRelMe(resourceURL)
resourceRelMes = resourceRelMe['relme']
for url in resourceRelMes:
if profile in (url, normalizeURL(url)):
result = True
break
return result |
Indent every line of text in a newline-delimited string
def indent_text(string, indent_level=2):
"""Indent every line of text in a newline-delimited string"""
indented_lines = []
indent_spaces = ' ' * indent_level
for line in string.split('\n'):
indented_lines.append(indent_spaces + line)
return '\n'.join(indented_lines) |
Download a file using requests.
This is like urllib.request.urlretrieve, but:
- requests validates SSL certificates by default
- you can pass tracker objects to e.g. display a progress bar or calculate
a file hash.
def download(url, target, headers=None, trackers=()):
"""Download a file using requests.
This is like urllib.request.urlretrieve, but:
- requests validates SSL certificates by default
- you can pass tracker objects to e.g. display a progress bar or calculate
a file hash.
"""
if headers is None:
headers = {}
headers.setdefault('user-agent', 'requests_download/'+__version__)
r = requests.get(url, headers=headers, stream=True)
r.raise_for_status()
for t in trackers:
t.on_start(r)
with open(target, 'wb') as f:
for chunk in r.iter_content(chunk_size=8192):
if chunk:
f.write(chunk)
for t in trackers:
t.on_chunk(chunk)
for t in trackers:
t.on_finish() |
Writes an object created by `parse` to either a file or a bytearray.
If the object doesn't end on a byte boundary, zeroes are appended to it
until it does.
def write(parsed_obj, spec=None, filename=None):
"""Writes an object created by `parse` to either a file or a bytearray.
If the object doesn't end on a byte boundary, zeroes are appended to it
until it does.
"""
if not isinstance(parsed_obj, BreadStruct):
raise ValueError(
'Object to write must be a structure created '
'by bread.parse')
if filename is not None:
with open(filename, 'wb') as fp:
parsed_obj._data_bits[:parsed_obj._length].tofile(fp)
else:
return bytearray(parsed_obj._data_bits[:parsed_obj._length].tobytes()) |
Uploads a file to an S3 bucket, as a public file.
def deploy_file(file_path, bucket):
""" Uploads a file to an S3 bucket, as a public file. """
# Paths look like:
# index.html
# css/bootstrap.min.css
logger.info("Deploying {0}".format(file_path))
# Upload the actual file to file_path
k = Key(bucket)
k.key = file_path
try:
k.set_contents_from_filename(file_path)
k.set_acl('public-read')
except socket.error:
logger.warning("Caught socket.error while trying to upload {0}".format(
file_path))
msg = "Please file an issue with alotofeffort if you see this,"
logger.warning(msg)
logger.warning("providing as much info as you can.") |
Deploy to the configured S3 bucket.
def deploy(www_dir, bucket_name):
""" Deploy to the configured S3 bucket. """
# Set up the connection to an S3 bucket.
conn = boto.connect_s3()
bucket = conn.get_bucket(bucket_name)
# Deploy each changed file in www_dir
os.chdir(www_dir)
for root, dirs, files in os.walk('.'):
for f in files:
# Use full relative path. Normalize to remove dot.
file_path = os.path.normpath(os.path.join(root, f))
if has_changed_since_last_deploy(file_path, bucket):
deploy_file(file_path, bucket)
else:
logger.info("Skipping {0}".format(file_path))
# Make the whole bucket public
bucket.set_acl('public-read')
# Configure it to be a website
bucket.configure_website('index.html', 'error.html')
# Print the endpoint, so you know the URL
msg = "Your website is now live at {0}".format(
bucket.get_website_endpoint())
logger.info(msg)
logger.info("If you haven't done so yet, point your domain name there!") |
Checks if a file has changed since the last time it was deployed.
:param file_path: Path to file which should be checked. Should be relative
from root of bucket.
:param bucket_name: Name of S3 bucket to check against.
:returns: True if the file has changed, else False.
def has_changed_since_last_deploy(file_path, bucket):
"""
Checks if a file has changed since the last time it was deployed.
:param file_path: Path to file which should be checked. Should be relative
from root of bucket.
:param bucket_name: Name of S3 bucket to check against.
:returns: True if the file has changed, else False.
"""
msg = "Checking if {0} has changed since last deploy.".format(file_path)
logger.debug(msg)
with open(file_path) as f:
data = f.read()
file_md5 = hashlib.md5(data.encode('utf-8')).hexdigest()
logger.debug("file_md5 is {0}".format(file_md5))
key = bucket.get_key(file_path)
# HACK: Boto's md5 property does not work when the file hasn't been
# downloaded. The etag works but will break for multi-part uploaded files.
# http://stackoverflow.com/questions/16872679/how-to-programmatically-
# get-the-md5-checksum-of-amazon-s3-file-using-boto/17607096#17607096
# Also the double quotes around it must be stripped. Sketchy...boto's fault
if key:
key_md5 = key.etag.replace('"', '').strip()
logger.debug("key_md5 is {0}".format(key_md5))
else:
logger.debug("File does not exist in bucket")
return True
if file_md5 == key_md5:
logger.debug("File has not changed.")
return False
logger.debug("File has changed.")
return True |
Entry point for the package, as defined in setup.py.
def main():
""" Entry point for the package, as defined in setup.py. """
# Log info and above to console
logging.basicConfig(
format='%(levelname)s: %(message)s', level=logging.INFO)
# Get command line input/output arguments
msg = 'Instantly deploy static HTML sites to S3 at the command line.'
parser = argparse.ArgumentParser(description=msg)
parser.add_argument(
'www_dir',
help='Directory containing the HTML files for your website.'
)
parser.add_argument(
'bucket_name',
help='Name of S3 bucket to deploy to, e.g. mybucket.'
)
args = parser.parse_args()
# Deploy the site to S3!
deploy(args.www_dir, args.bucket_name) |
This keyword is used to start sikuli java process.
If library is inited with mode "OLD", sikuli java process is started automatically.
If library is inited with mode "NEW", this keyword should be used.
:param port: port of sikuli java process, if value is None or 0, a random free port will be used
:return: None
def start_sikuli_process(self, port=None):
"""
This keyword is used to start sikuli java process.
If library is inited with mode "OLD", sikuli java process is started automatically.
If library is inited with mode "NEW", this keyword should be used.
:param port: port of sikuli java process, if value is None or 0, a random free port will be used
:return: None
"""
if port is None or int(port) == 0:
port = self._get_free_tcp_port()
self.port = port
start_retries = 0
started = False
while start_retries < 5:
try:
self._start_sikuli_java_process()
except RuntimeError as err:
print('error........%s' % err)
if self.process:
self.process.terminate_process()
self.port = self._get_free_tcp_port()
start_retries += 1
continue
started = True
break
if not started:
raise RuntimeError('Start sikuli java process failed!')
self.remote = self._connect_remote_library() |
Respond to POSTed username/password with token.
def post(self, request):
"""Respond to POSTed username/password with token."""
serializer = AuthTokenSerializer(data=request.data)
if serializer.is_valid():
token, _ = ExpiringToken.objects.get_or_create(
user=serializer.validated_data['user']
)
if token.expired():
# If the token is expired, generate a new one.
token.delete()
token = ExpiringToken.objects.create(
user=serializer.validated_data['user']
)
data = {'token': token.key}
return Response(data)
return Response(serializer.errors, status=HTTP_400_BAD_REQUEST) |
Return the allowed lifespan of a token as a TimeDelta object.
Defaults to 30 days.
def EXPIRING_TOKEN_LIFESPAN(self):
"""
Return the allowed lifespan of a token as a TimeDelta object.
Defaults to 30 days.
"""
try:
val = settings.EXPIRING_TOKEN_LIFESPAN
except AttributeError:
val = timedelta(days=30)
return val |
Return boolean indicating token expiration.
def expired(self):
"""Return boolean indicating token expiration."""
now = timezone.now()
if self.created < now - token_settings.EXPIRING_TOKEN_LIFESPAN:
return True
return False |
Test if a token is made entirely of Unicode characters of the following
classes:
- P: punctuation
- S: symbols
- Z: separators
- M: combining marks
- C: control characters
>>> unicode_is_punctuation('word')
False
>>> unicode_is_punctuation('。')
True
>>> unicode_is_punctuation('-')
True
>>> unicode_is_punctuation('-3')
False
>>> unicode_is_punctuation('あ')
False
def unicode_is_punctuation(text):
"""
Test if a token is made entirely of Unicode characters of the following
classes:
- P: punctuation
- S: symbols
- Z: separators
- M: combining marks
- C: control characters
>>> unicode_is_punctuation('word')
False
>>> unicode_is_punctuation('。')
True
>>> unicode_is_punctuation('-')
True
>>> unicode_is_punctuation('-3')
False
>>> unicode_is_punctuation('あ')
False
"""
for char in str_func(text):
category = unicodedata.category(char)[0]
if category not in 'PSZMC':
return False
return True |
Store the actual process in _process. If it doesn't exist yet, create
it.
def process(self):
"""
Store the actual process in _process. If it doesn't exist yet, create
it.
"""
if hasattr(self, '_process'):
return self._process
else:
self._process = self._get_process()
return self._process |
Create the process by running the specified command.
def _get_process(self):
"""
Create the process by running the specified command.
"""
command = self._get_command()
return subprocess.Popen(command, bufsize=-1, close_fds=True,
stdout=subprocess.PIPE,
stdin=subprocess.PIPE) |
Split a text into separate words.
def tokenize_list(self, text):
"""
Split a text into separate words.
"""
return [self.get_record_token(record) for record in self.analyze(text)] |
Determine whether a single word is a stopword, or whether a short
phrase is made entirely of stopwords, disregarding context.
Use of this function should be avoided; it's better to give the text
in context and let the process determine which words are the stopwords.
def is_stopword(self, text):
"""
Determine whether a single word is a stopword, or whether a short
phrase is made entirely of stopwords, disregarding context.
Use of this function should be avoided; it's better to give the text
in context and let the process determine which words are the stopwords.
"""
found_content_word = False
for record in self.analyze(text):
if not self.is_stopword_record(record):
found_content_word = True
break
return not found_content_word |
Get a canonical list representation of text, with words
separated and reduced to their base forms.
TODO: use the cache.
def normalize_list(self, text, cache=None):
"""
Get a canonical list representation of text, with words
separated and reduced to their base forms.
TODO: use the cache.
"""
words = []
analysis = self.analyze(text)
for record in analysis:
if not self.is_stopword_record(record):
words.append(self.get_record_root(record))
if not words:
# Don't discard stopwords if that's all you've got
words = [self.get_record_token(record) for record in analysis]
return words |
Given some text, return a sequence of (stem, pos, text) triples as
appropriate for the reader. `pos` can be as general or specific as
necessary (for example, it might label all parts of speech, or it might
only distinguish function words from others).
Twitter-style hashtags and at-mentions have the stem and pos they would
have without the leading # or @. For instance, if the reader's triple
for "thing" is ('thing', 'NN', 'things'), then "#things" would come out
as ('thing', 'NN', '#things').
def tag_and_stem(self, text, cache=None):
"""
Given some text, return a sequence of (stem, pos, text) triples as
appropriate for the reader. `pos` can be as general or specific as
necessary (for example, it might label all parts of speech, or it might
only distinguish function words from others).
Twitter-style hashtags and at-mentions have the stem and pos they would
have without the leading # or @. For instance, if the reader's triple
for "thing" is ('thing', 'NN', 'things'), then "#things" would come out
as ('thing', 'NN', '#things').
"""
analysis = self.analyze(text)
triples = []
for record in analysis:
root = self.get_record_root(record)
token = self.get_record_token(record)
if token:
if unicode_is_punctuation(token):
triples.append((token, '.', token))
else:
pos = self.get_record_pos(record)
triples.append((root, pos, token))
return triples |
Given some text, extract phrases of up to 2 content words,
and map their normalized form to the complete phrase.
def extract_phrases(self, text):
"""
Given some text, extract phrases of up to 2 content words,
and map their normalized form to the complete phrase.
"""
analysis = self.analyze(text)
for pos1 in range(len(analysis)):
rec1 = analysis[pos1]
if not self.is_stopword_record(rec1):
yield self.get_record_root(rec1), rec1[0]
for pos2 in range(pos1 + 1, len(analysis)):
rec2 = analysis[pos2]
if not self.is_stopword_record(rec2):
roots = [self.get_record_root(rec1),
self.get_record_root(rec2)]
pieces = [analysis[i][0] for i in range(pos1, pos2+1)]
term = ' '.join(roots)
phrase = ''.join(pieces)
yield term, phrase
break |
Use MeCab to turn any text into its phonetic spelling, as katakana
separated by spaces.
def to_kana(text):
"""
Use MeCab to turn any text into its phonetic spelling, as katakana
separated by spaces.
"""
records = MECAB.analyze(text)
kana = []
for record in records:
if record.pronunciation:
kana.append(record.pronunciation)
elif record.reading:
kana.append(record.reading)
else:
kana.append(record.surface)
return ' '.join(k for k in kana if k) |
Return two things about each character:
- Its transliterated value (in Roman characters, if it's a kana)
- A class of characters indicating how it affects the romanization
def get_kana_info(char):
"""
Return two things about each character:
- Its transliterated value (in Roman characters, if it's a kana)
- A class of characters indicating how it affects the romanization
"""
try:
name = unicodedata.name(char)
except ValueError:
return char, NOT_KANA
# The names we're dealing with will probably look like
# "KATAKANA CHARACTER ZI".
if (name.startswith('HIRAGANA LETTER') or
name.startswith('KATAKANA LETTER') or
name.startswith('KATAKANA-HIRAGANA')):
names = name.split()
syllable = str_func(names[-1].lower())
if name.endswith('SMALL TU'):
# The small tsu (っ) doubles the following consonant.
# It'll show up as 't' on its own.
return 't', SMALL_TSU
elif names[-1] == 'N':
return 'n', NN
elif names[1] == 'PROLONGED':
# The prolongation marker doubles the previous vowel.
# It'll show up as '_' on its own.
return '_', PROLONG
elif names[-2] == 'SMALL':
# Small characters tend to modify the sound of the previous
# kana. If they can't modify anything, they're appended to
# the letter 'x' instead.
if syllable.startswith('y'):
return 'x' + syllable, SMALL_Y
else:
return 'x' + syllable, SMALL
return syllable, KANA
else:
if char in ROMAN_PUNCTUATION_TABLE:
char = ROMAN_PUNCTUATION_TABLE[char]
return char, NOT_KANA |
Runs a line of text through MeCab, and returns the results as a
list of lists ("records") that contain the MeCab analysis of each
word.
def analyze(self, text):
"""
Runs a line of text through MeCab, and returns the results as a
list of lists ("records") that contain the MeCab analysis of each
word.
"""
try:
self.process # make sure things are loaded
text = render_safe(text).replace('\n', ' ').lower()
results = []
for chunk in string_pieces(text):
self.send_input((chunk + '\n').encode('utf-8'))
while True:
out_line = self.receive_output_line().decode('utf-8')
if out_line == 'EOS\n':
break
word, info = out_line.strip('\n').split('\t')
record_parts = [word] + info.split(',')
# Pad the record out to have 10 parts if it doesn't
record_parts += [None] * (10 - len(record_parts))
record = MeCabRecord(*record_parts)
# special case for detecting nai -> n
if (record.surface == 'ん' and
record.conjugation == '不変化型'):
# rebuild the record so that record.root is 'nai'
record_parts[MeCabRecord._fields.index('root')] = 'ない'
record = MeCabRecord(*record_parts)
results.append(record)
return results
except ProcessError:
self.restart_process()
return self.analyze(text) |
Determine whether a single MeCab record represents a stopword.
This mostly determines words to strip based on their parts of speech.
If common_words is set to True (default), it will also strip common
verbs and nouns such as くる and よう. If more_stopwords is True, it
will look at the sub-part of speech to remove more categories.
def is_stopword_record(self, record):
"""
Determine whether a single MeCab record represents a stopword.
This mostly determines words to strip based on their parts of speech.
If common_words is set to True (default), it will also strip common
verbs and nouns such as くる and よう. If more_stopwords is True, it
will look at the sub-part of speech to remove more categories.
"""
# preserve negations
if record.root == 'ない':
return False
return (
record.pos in STOPWORD_CATEGORIES or
record.subclass1 in STOPWORD_CATEGORIES or
record.root in STOPWORD_ROOTS
) |
Given a record, get the word's part of speech.
Here we're going to return MeCab's part of speech (written in
Japanese), though if it's a stopword we prefix the part of speech
with '~'.
def get_record_pos(self, record):
"""
Given a record, get the word's part of speech.
Here we're going to return MeCab's part of speech (written in
Japanese), though if it's a stopword we prefix the part of speech
with '~'.
"""
if self.is_stopword_record(record):
return '~' + record.pos
else:
return record.pos |
Run text through the external process, and get a list of lists
("records") that contain the analysis of each word.
def analyze(self, text):
"""
Run text through the external process, and get a list of lists
("records") that contain the analysis of each word.
"""
try:
text = render_safe(text).strip()
if not text:
return []
chunks = text.split('\n')
results = []
for chunk_text in chunks:
if chunk_text.strip():
textbytes = (chunk_text + '\n').encode('utf-8')
self.send_input(textbytes)
out_line = ''
while True:
out_line = self.receive_output_line()
out_line = out_line.decode('utf-8')
if out_line == '\n':
break
record = out_line.strip('\n').split(' ')
results.append(record)
return results
except ProcessError:
self.restart_process()
return self.analyze(text) |
Untokenizing a text undoes the tokenizing operation, restoring
punctuation and spaces to the places that people expect them to be.
Ideally, `untokenize(tokenize(text))` should be identical to `text`,
except for line breaks.
def untokenize(words):
"""
Untokenizing a text undoes the tokenizing operation, restoring
punctuation and spaces to the places that people expect them to be.
Ideally, `untokenize(tokenize(text))` should be identical to `text`,
except for line breaks.
"""
text = ' '.join(words)
step1 = text.replace("`` ", '"').replace(" ''", '"').replace('. . .', '...')
step2 = step1.replace(" ( ", " (").replace(" ) ", ") ")
step3 = re.sub(r' ([.,:;?!%]+)([ \'"`])', r"\1\2", step2)
step4 = re.sub(r' ([.,:;?!%]+)$', r"\1", step3)
step5 = step4.replace(" '", "'").replace(" n't", "n't").replace(
"can not", "cannot")
step6 = step5.replace(" ` ", " '")
return step6.strip() |
r"""
Splits apart words that are written in CamelCase.
Bugs:
- Non-ASCII characters are treated as lowercase letters, even if they are
actually capital letters.
Examples:
>>> un_camel_case('1984ZXSpectrumGames')
'1984 ZX Spectrum Games'
>>> un_camel_case('aaAa aaAaA 0aA AAAa!AAA')
'aa Aa aa Aa A 0a A AA Aa! AAA'
>>> un_camel_case('MotörHead')
'Mot\xf6r Head'
>>> un_camel_case('MSWindows3.11ForWorkgroups')
'MS Windows 3.11 For Workgroups'
This should not significantly affect text that is not camel-cased:
>>> un_camel_case('ACM_Computing_Classification_System')
'ACM Computing Classification System'
>>> un_camel_case('Anne_Blunt,_15th_Baroness_Wentworth')
'Anne Blunt, 15th Baroness Wentworth'
>>> un_camel_case('Hindi-Urdu')
'Hindi-Urdu'
def un_camel_case(text):
r"""
Splits apart words that are written in CamelCase.
Bugs:
- Non-ASCII characters are treated as lowercase letters, even if they are
actually capital letters.
Examples:
>>> un_camel_case('1984ZXSpectrumGames')
'1984 ZX Spectrum Games'
>>> un_camel_case('aaAa aaAaA 0aA AAAa!AAA')
'aa Aa aa Aa A 0a A AA Aa! AAA'
>>> un_camel_case('MotörHead')
'Mot\xf6r Head'
>>> un_camel_case('MSWindows3.11ForWorkgroups')
'MS Windows 3.11 For Workgroups'
This should not significantly affect text that is not camel-cased:
>>> un_camel_case('ACM_Computing_Classification_System')
'ACM Computing Classification System'
>>> un_camel_case('Anne_Blunt,_15th_Baroness_Wentworth')
'Anne Blunt, 15th Baroness Wentworth'
>>> un_camel_case('Hindi-Urdu')
'Hindi-Urdu'
"""
revtext = text[::-1]
pieces = []
while revtext:
match = CAMEL_RE.match(revtext)
if match:
pieces.append(match.group(1))
revtext = revtext[match.end():]
else:
pieces.append(revtext)
revtext = ''
revstr = ' '.join(piece.strip(' _') for piece in pieces
if piece.strip(' _'))
return revstr[::-1].replace('- ', '-') |
Takes a (unicode) string and yields pieces of it that are at most `maxlen`
characters, trying to break it at punctuation/whitespace. This is an
important step before using a tokenizer with a maximum buffer size.
def string_pieces(s, maxlen=1024):
"""
Takes a (unicode) string and yields pieces of it that are at most `maxlen`
characters, trying to break it at punctuation/whitespace. This is an
important step before using a tokenizer with a maximum buffer size.
"""
if not s:
return
i = 0
while True:
j = i + maxlen
if j >= len(s):
yield s[i:]
return
# Using "j - 1" keeps boundary characters with the left chunk
while unicodedata.category(s[j - 1]) not in BOUNDARY_CATEGORIES:
j -= 1
if j == i:
# No boundary available; oh well.
j = i + maxlen
break
yield s[i:j]
i = j |
Assign a heuristic to possible outputs from Morphy. Minimizing this
heuristic avoids incorrect stems.
def _word_badness(word):
"""
Assign a heuristic to possible outputs from Morphy. Minimizing this
heuristic avoids incorrect stems.
"""
if word.endswith('e'):
return len(word) - 2
elif word.endswith('ess'):
return len(word) - 10
elif word.endswith('ss'):
return len(word) - 4
else:
return len(word) |
Get the most likely stem for a word using Morphy, once the input has been
pre-processed by morphy_stem().
def _morphy_best(word, pos=None):
"""
Get the most likely stem for a word using Morphy, once the input has been
pre-processed by morphy_stem().
"""
results = []
if pos is None:
pos = 'nvar'
for pos_item in pos:
results.extend(morphy(word, pos_item))
if not results:
return None
results.sort(key=lambda x: _word_badness(x))
return results[0] |
Get the most likely stem for a word. If a part of speech is supplied,
the stem will be more accurate.
Valid parts of speech are:
- 'n' or 'NN' for nouns
- 'v' or 'VB' for verbs
- 'a' or 'JJ' for adjectives
- 'r' or 'RB' for adverbs
Any other part of speech will be treated as unknown.
def morphy_stem(word, pos=None):
"""
Get the most likely stem for a word. If a part of speech is supplied,
the stem will be more accurate.
Valid parts of speech are:
- 'n' or 'NN' for nouns
- 'v' or 'VB' for verbs
- 'a' or 'JJ' for adjectives
- 'r' or 'RB' for adverbs
Any other part of speech will be treated as unknown.
"""
word = word.lower()
if pos is not None:
if pos.startswith('NN'):
pos = 'n'
elif pos.startswith('VB'):
pos = 'v'
elif pos.startswith('JJ'):
pos = 'a'
elif pos.startswith('RB'):
pos = 'r'
if pos is None and word.endswith('ing') or word.endswith('ed'):
pos = 'v'
if pos is not None and pos not in 'nvar':
pos = None
if word in EXCEPTIONS:
return EXCEPTIONS[word]
if pos is None:
if word in AMBIGUOUS_EXCEPTIONS:
return AMBIGUOUS_EXCEPTIONS[word]
return _morphy_best(word, pos) or word |
Returns a list of (stem, tag, token) triples:
- stem: the word's uninflected form
- tag: the word's part of speech
- token: the original word, so we can reconstruct it later
def tag_and_stem(text):
"""
Returns a list of (stem, tag, token) triples:
- stem: the word's uninflected form
- tag: the word's part of speech
- token: the original word, so we can reconstruct it later
"""
tokens = tokenize(text)
tagged = nltk.pos_tag(tokens)
out = []
for token, tag in tagged:
stem = morphy_stem(token, tag)
out.append((stem, tag, token))
return out |
Get a list of word stems that appear in the text. Stopwords and an initial
'to' will be stripped, unless this leaves nothing in the stem.
>>> normalize_list('the dog')
['dog']
>>> normalize_list('big dogs')
['big', 'dog']
>>> normalize_list('the')
['the']
def normalize_list(text):
"""
Get a list of word stems that appear in the text. Stopwords and an initial
'to' will be stripped, unless this leaves nothing in the stem.
>>> normalize_list('the dog')
['dog']
>>> normalize_list('big dogs')
['big', 'dog']
>>> normalize_list('the')
['the']
"""
pieces = [morphy_stem(word) for word in tokenize(text)]
pieces = [piece for piece in pieces if good_lemma(piece)]
if not pieces:
return [text]
if pieces[0] == 'to':
pieces = pieces[1:]
return pieces |
Get a canonical representation of a Wikipedia topic, which may include
a disambiguation string in parentheses.
Returns (name, disambig), where "name" is the normalized topic name,
and "disambig" is a string corresponding to the disambiguation text or
None.
def normalize_topic(topic):
"""
Get a canonical representation of a Wikipedia topic, which may include
a disambiguation string in parentheses.
Returns (name, disambig), where "name" is the normalized topic name,
and "disambig" is a string corresponding to the disambiguation text or
None.
"""
# find titles of the form Foo (bar)
topic = topic.replace('_', ' ')
match = re.match(r'([^(]+) \(([^)]+)\)', topic)
if not match:
return normalize(topic), None
else:
return normalize(match.group(1)), 'n/' + match.group(2).strip(' _') |
split key to elements
def key2elements(key):
"""split key to elements"""
# words = key.split('.')
# if len(words) == 4:
# return words
# # there is a dot in object name
# fieldword = words.pop(-1)
# nameword = '.'.join(words[-2:])
# if nameword[-1] in ('"', "'"):
# # The object name is in quotes
# nameword = nameword[1:-1]
# elements = words[:-2] + [nameword, fieldword, ]
# return elements
words = key.split('.')
first2words = words[:2]
lastword = words[-1]
namewords = words[2:-1]
namephrase = '.'.join(namewords)
if namephrase.startswith("'") and namephrase.endswith("'"):
namephrase = namephrase[1:-1]
return first2words + [namephrase] + [lastword] |
update idf using dct
def updateidf(idf, dct):
"""update idf using dct"""
for key in list(dct.keys()):
if key.startswith('idf.'):
idftag, objkey, objname, field = key2elements(key)
if objname == '':
try:
idfobj = idf.idfobjects[objkey.upper()][0]
except IndexError as e:
idfobj = idf.newidfobject(objkey.upper())
else:
idfobj = idf.getobject(objkey.upper(), objname)
if idfobj == None:
idfobj = idf.newidfobject(objkey.upper(), Name=objname)
idfobj[field] = dct[key] |
return the fan power in bhp given fan efficiency, Pressure rise (Pa) and flow (m3/s)
def fan_bhp(fan_tot_eff, pascal, m3s):
"""return the fan power in bhp given fan efficiency, Pressure rise (Pa) and flow (m3/s)"""
# from discussion in
# http://energy-models.com/forum/baseline-fan-power-calculation
inh2o = pascal2inh2o(pascal)
cfm = m3s2cfm(m3s)
return (cfm * inh2o * 1.0) / (6356.0 * fan_tot_eff) |
return inputs for E+ in pascal and m3/s
def bhp2pascal(bhp, cfm, fan_tot_eff):
"""return inputs for E+ in pascal and m3/s"""
inh2o = bhp * 6356.0 * fan_tot_eff / cfm
pascal = inh2o2pascal(inh2o)
m3s = cfm2m3s(cfm)
return pascal, m3s |
return the fan power in watts given fan efficiency, Pressure rise (Pa) and flow (m3/s)
def fan_watts(fan_tot_eff, pascal, m3s):
"""return the fan power in watts given fan efficiency, Pressure rise (Pa) and flow (m3/s)"""
# got this from a google search
bhp = fan_bhp(fan_tot_eff, pascal, m3s)
return bhp2watts(bhp) |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.