def newDb(self, path):
if os.path.exists(path): os.remove(path)
self.data = ParticleDataSQLite(path)
self.initInterface()
def openDb(self, path):
self.data = ParticleDataSQLite(path)
self.initInterface()
self.view.updateGeneralInformation()
self.view.updateStats()
self.view.updateCalendar()
class Controller():
def __init__(self, appPath):
self.data = None
self.view = None
self.operations = {}
self.appPath = appPath
self.nucleationSizeLimit = 0.000000025 # 25 nm
self.relevantWindowSize = 12
def setView(self, view):
"""
Set view instance for calling interface update methods
"""
self.view = view
self.initInterface()
def newDb(self, path):
if os.path.exists(path): os.remove(path)
self.data = ParticleDataSQLite(path)
self.initInterface()
def openDb(self, path):
self.data = ParticleDataSQLite(path)
self.initInterface()
self.view.updateGeneralInformation()
self.view.updateStats()
self.view.updateCalendar()
def initInterface(self):
"""
Llama a los procedimientos necesarios del interfaz para resetear el mismo.
"""
self.view.initInterface()
def _getDateTimeFromJulianDate(self, julianDate):
X = float(julianDate)
Z = math.floor(X)
F = X - Z
Y = math.floor((Z - 1867216.25) / 36524.25)
A = Z + 1 + Y - math.floor(Y / 4)
B = A + 1524
C = math.floor((B - 122.1)/365.25)
D = math.floor(365.25 * C)
G = math.floor((B - D) / 30.6001)
month = (G - 1) if (G < 13.5) else (G - 13)
year = (C - 4715) if (month < 2.5) else (C - 4716)
UT = B - D - math.floor(30.6001 * G) + F
day = int(math.floor(UT))
UT -= math.floor(UT)
UT *= 24
hour = int(math.floor(UT))
UT -= math.floor(UT)
UT *= 60
minute = int(math.floor(UT))
UT -= math.floor(UT)
UT *= 60
second = int(round(UT))
return datetime.datetime(int(year), int(month), day, hour, minute, second)
def _getTimeFromJulianDate(self, julianDate):
julianDate = float(julianDate)
UT = julianDate - math.floor(julianDate)
UT -= math.floor(UT)
UT *= 24
hour = int(math.floor(UT))
UT -= math.floor(UT)
UT *= 60
minute = int(math.floor(UT))
UT -= math.floor(UT)
UT *= 60
second = int(round(UT))
return datetime.time(hour, minute, second)
def importSUM(self, path, date):
"""
Imports a SUM file and update interface consequently.
"""
if self.data.isDayPresent(date):
if self.view.showDayPresentDialog():
self.data.deleteDayData(date, 'norm')
self.data.deleteDayData(date, 'mf')
self.data.deleteDayData(date, 'raw')
else:
return
result = self.data.importSumFile(path, date)
self.view.showImportResult(result)
if result:
self.view.updateCalendar()
def importTraining(self, path):
self.reader = csv.DictReader(open(path, "r"), delimiter=',')
classes = {}
# Delete previous training data
self.data.deleteTrainingData()
# Perform sum of all elements of the class
for record in self.reader:
date = datetime.datetime(int(record['YEAR']), 1, 1) + datetime.timedelta(int(record['DAY']) - 1)
date = date.date()
measures = self.data.getDayData(date, 'norm')
if len(measures):
class_id = int(record['CLASS'])
times = measures.keys()
times.sort()
sizes = measures[times[0]].keys()
sizes.sort()
if classes.get(class_id):
hits_count, class_measures = classes.get(class_id)
hits_count += 1
else:
hits_count = 1
class_measures = numpy.zeros((len(times), 14), dtype=numpy.float64)
offset = 14 - len(sizes)
for i in range(len(times)):
for j in range(len(sizes)):
class_measures[i][offset + j] += measures[times[i]][sizes[j]]
classes[class_id] = (hits_count, class_measures)
else:
print "No normalized data for %s" %(date)
# Compute average
class_ids = classes.keys()
for class_id in class_ids:
hits_count, class_measures = classes[class_id]
class_range = class_measures.shape
for i in range(class_range[0]):
for j in range(class_range[1]):
class_measures[i][j] /= hits_count
# Store training data
self.data.addTrainingData(class_id, class_measures)
def performKmeans(self):
# Construct centroid array
centroids = []
max_classes = 5
for i in range(max_classes): # TODO: Coger de la BD el nº de clases
centroid = self.data.getTrainingData(i)
shape = centroid.shape
centroids.append([item for sublist in centroid for item in sublist])
centroids = numpy.array(centroids)
dates = self.data.getAllActiveDays()
observations = []
for i in range(len(dates)):
measures = self.data.getDayData(dates[i], 'norm')
if len(measures):
times = measures.keys()
times.sort()
sizes = measures[times[0]].keys()
sizes.sort()
offset = shape[1] - len(sizes)
observation = numpy.zeros(shape[0] * shape[1], numpy.float64)
for i in range(len(times)):
for j in range(len(sizes)):
observation[i * shape[1] + offset + j] += measures[times[i]][sizes[j]]
observations.append(observation)
observations = numpy.array(observations)
centroid, labels = scipy.cluster.vq.kmeans2(observations, centroids, iter=10, thresh=1e-05, minit='matrix')
# Evaluation
classes = {}
rank_classes = {}
for i in range(max_classes):
classes[i] = []
rank_classes[i] = [0, 0]
for j in range(len(labels)):
if labels[j] == i:
classes[i].append(dates[j])
print "Clase %s - Total: %s" %(i, len(classes[i]))
#print classes[i]
self.reader = csv.DictReader(open(os.path.join(self.appPath, 'hyde_clean_06.csv'), "r"), delimiter=',')
for record in self.reader:
date = datetime.datetime(int(record['year']), 1, 1) + datetime.timedelta(int(record['day']) - 1)
date = date.strftime('%Y-%m-%d')
owner_class = None
if record['event_1'] == '1':
owner_class = 0
elif record['event_2'] == '1':
owner_class = 1
elif record['event_3'] == '1':
owner_class = 2
elif record['event_0'] == '1':
owner_class = 3
elif record['non_event'] == '1':
owner_class = 4
if not owner_class is None:
rank_classes[owner_class][0] += 1
for date_class in classes[i]:
if date_class == date:
rank_classes[owner_class][1] += 1
break
for i in range(max_classes):
print "Clase %s. Nº elementos manual: %s - Bien clasificados: %s" %(i, rank_classes[i][0], rank_classes[i][1])
def _constructReExpression(self, pattern):
regexPattern = pattern.replace('%y', '[0-9][0-9]')
regexPattern = regexPattern.replace('%yyy', '[0-9][0-9][0-9][0-9]')
regexPattern = regexPattern.replace('%m', '[0-9][0-9]')
regexPattern = regexPattern.replace('%d', '[0-9][0-9]')
regexPattern = '^' + regexPattern + '$'
return re.compile(regexPattern, re.IGNORECASE)
def importFolder(self, path, pattern, progressDialog, numberFiles):
"""
Preparan el patrón RegEx para pasarlo como patrón a un método recursivo que examina los posibles
archivos a importar.
@param path: Ruta de la carpeta raíz a importar
@param pattern: Patrón para obtener la fecha a través del nombre de archivo.
Códigos:
- %yyy - Año con cuatro cifras
- %y - Año con dos cifras
- %m - Mes con dos cifras
- %d - Día con dos cifras
"""
expression = self._constructReExpression(pattern)
result = True
count = 0
# Walk trough entire path
for dirname, dirnames, filenames in os.walk(path):
for filename in filenames:
# See if filename match the pattern
if expression.match(filename):
count += 1
# Extract date
index = pattern.find('%yyy')
if index > -1:
year = int(filename[index : index + 4])
else:
index = pattern.find('%y')
year = int(filename[index : index + 2])
# Ajuste a cuatro cifras
year += 2000 if year < 70 else 1900
index = pattern.find('%m')
month = int(filename[index : index + 2])
index = pattern.find('%d')
day = int(filename[index : index + 2])
date = datetime.date(year, month, day)
if self.data.isDayPresent(date):
self.data.deleteDayData(date, 'norm')
self.data.deleteDayData(date, 'mf')
self.data.deleteDayData(date, 'raw')
result = self.data.importSumFile(os.path.join(dirname, filename), date)
result = result and progressDialog.Update(count, 'Importing file %s/%s' %(count, numberFiles))[0]
while self.view.Pending(): self.view.Dispatch() # assure correct display
if not result: break
if not result: break
self.view.showImportResult(result)
return
def getNumberFiles(self, path, pattern):
expression = self._constructReExpression(pattern)
count = 0
for dirname, dirnames, filenames in os.walk(path):
for filename in filenames:
if expression.match(filename): count += 1
return count
def medianFilterOneDay(self, date, window):
"""
Makes a median filter of given date and stores in DB.
@param date: Date for which median filter is going to be calculated.
@param window: List of two elements specifying window size for the algorithm.
"""
result = self._medianFilter(self.data.getDayData(date, 'raw'), window)
if result:
times, sizes, values = result
# Delete previous data if any
self.data.deleteDayData(date, 'mf')
# Add new data
self.data.addDayData(date, times, sizes, values, 'mf')
else:
print "Incorrect data for date: %s" %(date)
return bool(result)
def medianFilterAll(self, window, progressDialog, numberDays):
"""
Makes a median filter of all dates and stores in DB.
@param window: List of two elements specifying window size for the algorithm.
"""
dates = self.data.getAllActiveDays()
count = 0
for date in dates:
count += 1
self.medianFilterOneDay(date, window)
if not progressDialog.Update(count, 'Calculating day %s/%s' %(count, numberDays))[0]: break
while self.view.Pending(): self.view.Dispatch() # assure correct display
def _medianFilter(self, events, window):
"""
Makes a median filter of given data.
@param events: Data for doing median filter.
@param window: List of two elements specifying window size for the algorithm.
@return: array with data filter by median
"""
if len(events):
# Build array from events dictionary
sizes = events.values()[0].keys()
sizes.sort()
times = events.keys()
times.sort()
eventsMatrix = []
for time in times:
rowTime = []
for size in sizes:
rowTime.append(events[time][size])
eventsMatrix.append(rowTime)
# Do median filter
try:
result = scipy.signal.medfilt2d(eventsMatrix, window)
except:
return None
return (times, sizes, result)
else:
return None
def relevantWindowDay(self, date, windowSize, method, source):
size_max, time_max = self._relevantWindow(date, windowSize, method=method, source=source)
if size_max:
self.data.setDayRelevantWindow(date, 0, size_max, time_max - windowSize + 1, time_max)
def relevantWindowAll(self, windowSize, method, source, progressDialog, numberDays):
dates = self.data.getAllActiveDays()
count = 0
for date in dates:
count += 1
self.relevantWindowDay(date, windowSize, method=method, source=source)
if not progressDialog.Update(count, 'Calculating day %s/%s' %(count, numberDays))[0]: break
while self.view.Pending(): self.view.Dispatch() # assure correct display
def _relevantWindow(self, date, windowSize, method='max', source='raw'):
"""
Calculates relevant data window of given date.
@param date: Date for which relevant window is going to be calculated.
@param windowSize: Vertical size of the window (time lapse). Horizontal window size is calculated
automatically from nucleationSizeLimit variable.
@param metod: 'max' for maximum intensity, 'dif' for differential intensity.
@return: tuple with left, right, top and bottom indexes of the window.
"""
captures = self.data.getDayData(date, source)
if len(captures):
times = captures.keys()
times.sort()
# Determine particle size interval relevant from first capture of the day
capture = captures[times[0]]
size_upper_limit = 0
sizes = capture.keys()
sizes.sort()
while sizes[size_upper_limit] < self.nucleationSizeLimit:
size_upper_limit += 1 # keep this value instead minus 1 to get index of next bin after nucleation mode size limit
windowSize -= 1 # Adjust window size for 0-base index issue
time_upper_limit = windowSize
max_sum = 0
for time_pos in range(0, len(times) - windowSize - 1):
local_sum = 0
for i in range(time_pos, time_pos + windowSize):
dataTime = captures[times[i]]
if dataTime:
for j in range(0, size_upper_limit):
if dataTime[sizes[j]]:
if method == 'max':
# Calculate sum of the values of the window
local_sum += dataTime[sizes[j]]
elif method == 'dif':
# Calculate sum of the increments of the window
if i > 0 and captures[times[i-1]][sizes[j]]: local_sum += dataTime[sizes[j]] - captures[times[i-1]][sizes[j]]
if local_sum > max_sum:
max_sum = local_sum
time_upper_limit = time_pos + windowSize
return size_upper_limit, time_upper_limit
else:
return (None, None)
def normalizationDay(self, date, source='raw'):
self.data.deleteDayData(date, 'norm')
# Get data
captures = self.data.getDayData(date, source)
relevantWindow = self.data.getDayRelevantWindow(date)
if relevantWindow and len(captures):
times = captures.keys()
times.sort()
sizes = captures[times[0]].keys()
sizes.sort()
# Reduce keys to relevant window frame
times = times[relevantWindow[2]:relevantWindow[3]+1]
sizes = sizes[relevantWindow[0]:relevantWindow[1]+1]
# Get maximum value of the window
maximum = 0
for time in times:
for size in sizes:
if captures[time][size] > maximum:
maximum = captures[time][size]
# Normalize
normalizedData = []
for time in times:
timeData = []
for size in sizes:
if captures[time][size]:
timeData.append(captures[time][size] / maximum)
else:
timeData.append(0.0)
normalizedData.append(timeData)
self.data.addDayData(date, times, sizes, normalizedData, 'norm')
else:
# TODO: Informar del error de otra forma
print "No data or relevant window for date %s" %(date)
def normalizationAll(self, source, progressDialog, numberDays):
dates = self.data.getAllActiveDays()
count = 0
for date in dates:
count += 1
self.normalizationDay(date, source=source)
if not progressDialog.Update(count, 'Calculating day %s/%s' %(count, numberDays))[0]: break
while self.view.Pending(): self.view.Dispatch() # assure correct display
