| 6 |
# Read the file COPYING coming with Thuban for details. |
# Read the file COPYING coming with Thuban for details. |
| 7 |
|
|
| 8 |
""" |
""" |
| 9 |
ClassGenerator |
Functions to generate Classifications |
| 10 |
""" |
""" |
| 11 |
|
|
| 12 |
__version__ = "$Revision$" |
__version__ = "$Revision$" |
| 20 |
from classification import Classification, ClassGroupSingleton, \ |
from classification import Classification, ClassGroupSingleton, \ |
| 21 |
ClassGroupRange, ClassGroupProperties |
ClassGroupRange, ClassGroupProperties |
| 22 |
|
|
| 23 |
class ClassGenerator: |
def GenSingletonsFromList(_list, numGroups, ramp): |
| 24 |
|
"""Generate a new classification consisting solely of singletons. |
| 25 |
|
|
| 26 |
def GenSingletonsFromList(self, _list, numGroups, ramp): |
The resulting classification will consist of at most 'numGroups' |
| 27 |
"""Generate a new classification consisting solely of singletons. |
groups whose group properties ramp between 'prop1' and 'prop2'. There |
| 28 |
|
could be fewer groups if '_list' contains fewer that 'numGroups' items. |
| 29 |
|
|
| 30 |
The resulting classification will consist of at most 'numGroups' |
_list -- any object that implements the iterator interface |
|
groups whose group properties ramp between 'prop1' and 'prop2'. There |
|
|
could be fewer groups if '_list' contains fewer that 'numGroups' items. |
|
| 31 |
|
|
| 32 |
_list -- any object that implements the iterator interface |
numGroups -- how many groups to generate. This can not be |
| 33 |
|
determined while the classification is being |
| 34 |
|
generated because the stepping values must |
| 35 |
|
be precalculated to ramp between prop1 and prop2. |
| 36 |
|
|
| 37 |
numGroups -- how many groups to generate. This can not be |
ramp -- an object which implements the CustomRamp interface |
| 38 |
determined while the classification is being |
""" |
|
generated because the stepping values must |
|
|
be precalculated to ramp between prop1 and prop2. |
|
| 39 |
|
|
| 40 |
ramp -- an object which implements the CustomRamp interface |
clazz = Classification() |
| 41 |
""" |
if numGroups == 0: return clazz |
| 42 |
|
|
| 43 |
clazz = Classification() |
ramp.SetNumGroups(numGroups) |
|
if numGroups == 0: return clazz |
|
| 44 |
|
|
| 45 |
ramp.SetNumGroups(numGroups) |
for value, prop in zip(_list, ramp): |
| 46 |
|
clazz.AppendGroup(ClassGroupSingleton(value, prop)) |
| 47 |
|
|
| 48 |
for value, prop in zip(_list, ramp): |
return clazz |
|
clazz.AppendGroup(ClassGroupSingleton(value, prop)) |
|
| 49 |
|
|
| 50 |
return clazz |
def GenSingletons(min, max, numGroups, ramp): |
| 51 |
|
|
| 52 |
def GenSingletons(self, min, max, numGroups, ramp): |
clazz = Classification() |
| 53 |
|
|
| 54 |
clazz = Classification() |
#step = int((max - min) / float(numGroups)) |
| 55 |
|
|
| 56 |
#step = int((max - min) / float(numGroups)) |
if numGroups > 0: |
| 57 |
|
|
| 58 |
if numGroups > 0: |
step = int((max - min + 1) / float(numGroups)) |
| 59 |
|
cur_value = min |
| 60 |
|
|
| 61 |
step = int((max - min + 1) / float(numGroups)) |
ramp.SetNumGroups(numGroups) |
|
cur_value = min |
|
| 62 |
|
|
| 63 |
ramp.SetNumGroups(numGroups) |
for prop in ramp: |
| 64 |
|
clazz.AppendGroup(ClassGroupSingleton(cur_value), prop) |
| 65 |
|
cur_value += step |
| 66 |
|
|
| 67 |
for prop in ramp: |
return clazz |
|
clazz.AppendGroup(ClassGroupSingleton(cur_value), prop) |
|
|
cur_value += step |
|
|
|
|
|
return clazz |
|
|
|
|
|
def GenUniformDistribution(self, min, max, numGroups, |
|
|
ramp, intStep = False): |
|
|
"""Generate a classification with numGroups range groups |
|
|
each with the same interval. |
|
|
|
|
|
intStep -- force the calculated stepping to an integer. |
|
|
Useful if the values are integers but the |
|
|
number of groups specified doesn't evenly |
|
|
divide (max - min). |
|
|
""" |
|
| 68 |
|
|
| 69 |
clazz = Classification() |
def GenUniformDistribution(min, max, numGroups, |
| 70 |
if numGroups == 0: return clazz |
ramp, intStep = False): |
| 71 |
|
"""Generate a classification with numGroups range groups |
| 72 |
|
each with the same interval. |
| 73 |
|
|
| 74 |
ramp.SetNumGroups(numGroups) |
intStep -- force the calculated stepping to an integer. |
| 75 |
|
Useful if the values are integers but the |
| 76 |
|
number of groups specified doesn't evenly |
| 77 |
|
divide (max - min). |
| 78 |
|
""" |
| 79 |
|
|
| 80 |
step = (max - min) / float(numGroups) |
clazz = Classification() |
| 81 |
|
if numGroups == 0: return clazz |
| 82 |
|
|
| 83 |
if intStep: |
ramp.SetNumGroups(numGroups) |
|
step = int(step) |
|
| 84 |
|
|
| 85 |
cur_min = min |
step = (max - min) / float(numGroups) |
|
cur_max = cur_min + step |
|
| 86 |
|
|
| 87 |
i = 0 |
if intStep: |
| 88 |
end = "[" |
step = int(step) |
|
for prop in ramp: |
|
| 89 |
|
|
| 90 |
if i == (numGroups - 1): |
cur_min = min |
| 91 |
cur_max = max |
cur_max = cur_min + step |
|
end = "]" |
|
| 92 |
|
|
| 93 |
|
i = 0 |
| 94 |
|
end = "[" |
| 95 |
|
for prop in ramp: |
| 96 |
|
|
| 97 |
# this check guards against rounding issues |
if i == (numGroups - 1): |
| 98 |
if cur_min != cur_max: |
cur_max = max |
| 99 |
range = Range(("[", cur_min, cur_max, end)) |
end = "]" |
|
clazz.AppendGroup(ClassGroupRange(range, None, prop)) |
|
| 100 |
|
|
|
cur_min = cur_max |
|
|
cur_max += step |
|
|
i += 1 |
|
| 101 |
|
|
| 102 |
return clazz |
# this check guards against rounding issues |
| 103 |
|
if cur_min != cur_max: |
| 104 |
|
range = Range(("[", cur_min, cur_max, end)) |
| 105 |
|
clazz.AppendGroup(ClassGroupRange(range, None, prop)) |
| 106 |
|
|
| 107 |
|
cur_min = cur_max |
| 108 |
|
cur_max += step |
| 109 |
|
i += 1 |
| 110 |
|
|
| 111 |
def GenQuantiles(self, _list, percents, ramp, _range): |
return clazz |
|
"""Generates a Classification which has groups of ranges that |
|
|
represent quantiles of _list at the percentages given in percents. |
|
|
Only the values that fall within _range are considered. |
|
| 112 |
|
|
|
Returns a tuple (adjusted, Classification) where adjusted is |
|
|
True if the Classification does not exactly represent the given |
|
|
range, or if the Classification is empty. |
|
| 113 |
|
|
| 114 |
_list -- a sort list of values |
def GenQuantiles(_list, percents, ramp, _range): |
| 115 |
|
"""Generates a Classification which has groups of ranges that |
| 116 |
|
represent quantiles of _list at the percentages given in percents. |
| 117 |
|
Only the values that fall within _range are considered. |
| 118 |
|
|
| 119 |
percents -- a sorted list of floats in the range 0.0-1.0 which |
Returns a tuple (adjusted, Classification) where adjusted is |
| 120 |
represent the upper bound of each quantile |
True if the Classification does not exactly represent the given |
| 121 |
|
range, or if the Classification is empty. |
| 122 |
|
|
| 123 |
ramp -- an object which implements the CustomRamp interface |
_list -- a sort list of values |
| 124 |
|
|
| 125 |
_range -- a Range object |
percents -- a sorted list of floats in the range 0.0-1.0 which |
| 126 |
""" |
represent the upper bound of each quantile |
| 127 |
|
|
| 128 |
clazz = Classification() |
ramp -- an object which implements the CustomRamp interface |
|
quantiles = self.CalculateQuantiles(_list, percents, _range) |
|
|
adjusted = True |
|
| 129 |
|
|
| 130 |
if quantiles is not None: |
_range -- a Range object |
| 131 |
|
""" |
| 132 |
|
|
| 133 |
numGroups = len(quantiles[3]) |
clazz = Classification() |
| 134 |
|
quantiles = CalculateQuantiles(_list, percents, _range) |
| 135 |
|
adjusted = True |
| 136 |
|
|
| 137 |
if numGroups != 0: |
if quantiles is not None: |
| 138 |
|
|
| 139 |
adjusted = quantiles[0] |
numGroups = len(quantiles[3]) |
| 140 |
|
|
| 141 |
ramp.SetNumGroups(numGroups) |
if numGroups != 0: |
| 142 |
|
|
| 143 |
start, min, endMax, right = _range.GetRange() |
adjusted = quantiles[0] |
| 144 |
|
|
| 145 |
oldp = 0 |
ramp.SetNumGroups(numGroups) |
|
i = 1 |
|
|
end = "]" |
|
| 146 |
|
|
| 147 |
for (q, p), prop in zip(quantiles[3], ramp): |
start, min, endMax, right = _range.GetRange() |
|
if i == numGroups: |
|
|
max = endMax |
|
|
end = right |
|
|
else: |
|
|
max = _list[q] |
|
| 148 |
|
|
| 149 |
group = ClassGroupRange(Range((start, min, max, end)), |
oldp = 0 |
| 150 |
None, prop) |
i = 1 |
| 151 |
|
end = "]" |
| 152 |
group.SetLabel("%s%% - %s%%" % (round(oldp*100, 2), |
|
| 153 |
round(p*100, 2))) |
for (q, p), prop in zip(quantiles[3], ramp): |
| 154 |
oldp = p |
if i == numGroups: |
| 155 |
start = "]" |
max = endMax |
| 156 |
min = max |
end = right |
| 157 |
clazz.AppendGroup(group) |
else: |
| 158 |
i += 1 |
max = _list[q] |
| 159 |
|
|
| 160 |
return (adjusted, clazz) |
group = ClassGroupRange(Range((start, min, max, end)), |
| 161 |
|
None, prop) |
| 162 |
def CalculateQuantiles(self, _list, percents, _range): |
|
| 163 |
"""Calculate quantiles for the given _list of percents from the |
group.SetLabel("%s%% - %s%%" % (round(oldp*100, 2), |
| 164 |
sorted list of values that are in range. |
round(p*100, 2))) |
| 165 |
|
oldp = p |
| 166 |
This may not actually generate len(percents) quantiles if |
start = "]" |
| 167 |
many of the values that fall on quantile borders are the same. |
min = max |
| 168 |
|
clazz.AppendGroup(group) |
| 169 |
Returns a tuple of the form: |
i += 1 |
| 170 |
(adjusted, minIndex, maxIndex, [quantile_list]) |
|
| 171 |
|
return (adjusted, clazz) |
| 172 |
where adjusted is True if the the quantile percentages differ from |
|
| 173 |
those supplied, minIndex is the index into _list where the |
def CalculateQuantiles(_list, percents, _range): |
| 174 |
minimum value used is located, maxIndex is the index into _list |
"""Calculate quantiles for the given _list of percents from the |
| 175 |
where the maximum value used is located, and quantile_list is a |
sorted list of values that are in range. |
| 176 |
list of tuples of the form: (list_index, quantile_percentage) |
|
| 177 |
|
This may not actually generate len(percents) quantiles if |
| 178 |
Returns None, if no quantiles could be generated based on the |
many of the values that fall on quantile borders are the same. |
| 179 |
given range or input list. |
|
| 180 |
|
Returns a tuple of the form: |
| 181 |
|
(adjusted, minIndex, maxIndex, [quantile_list]) |
| 182 |
|
|
| 183 |
|
where adjusted is True if the the quantile percentages differ from |
| 184 |
|
those supplied, minIndex is the index into _list where the |
| 185 |
|
minimum value used is located, maxIndex is the index into _list |
| 186 |
|
where the maximum value used is located, and quantile_list is a |
| 187 |
|
list of tuples of the form: (list_index, quantile_percentage) |
| 188 |
|
|
| 189 |
|
Returns None, if no quantiles could be generated based on the |
| 190 |
|
given range or input list. |
| 191 |
|
|
| 192 |
|
_list -- a sort list of values |
| 193 |
|
|
| 194 |
|
percents -- a sorted list of floats in the range 0.0-1.0 which |
| 195 |
|
represent the upper bound of each quantile |
| 196 |
|
|
| 197 |
_list -- a sort list of values |
_range -- a Range object |
| 198 |
|
""" |
| 199 |
|
|
| 200 |
percents -- a sorted list of floats in the range 0.0-1.0 which |
quantiles = [] |
| 201 |
represent the upper bound of each quantile |
adjusted = False |
| 202 |
|
|
| 203 |
_range -- a Range object |
if len(percents) != 0: |
| 204 |
""" |
|
| 205 |
|
# |
| 206 |
quantiles = [] |
# find what part of the _list range covers |
| 207 |
adjusted = False |
# |
| 208 |
|
minIndex = -1 |
| 209 |
|
maxIndex = -2 |
| 210 |
|
for i in xrange(0, len(_list), 1): |
| 211 |
|
if operator.contains(_range, _list[i]): |
| 212 |
|
minIndex = i |
| 213 |
|
break |
| 214 |
|
|
| 215 |
|
for i in xrange(len(_list)-1, -1, -1): |
| 216 |
|
if operator.contains(_range, _list[i]): |
| 217 |
|
maxIndex = i |
| 218 |
|
break |
| 219 |
|
|
| 220 |
|
numValues = maxIndex - minIndex + 1 |
| 221 |
|
|
| 222 |
|
if numValues > 0: |
| 223 |
|
|
|
if len(percents) != 0: |
|
|
|
|
| 224 |
# |
# |
| 225 |
# find what part of the _list range covers |
# build a list of unique indices into list of where each |
| 226 |
|
# quantile *should* be. set adjusted if the resulting |
| 227 |
|
# indices are different |
| 228 |
# |
# |
| 229 |
minIndex = -1 |
quantiles = {} |
| 230 |
maxIndex = -2 |
for p in percents: |
| 231 |
for i in xrange(0, len(_list), 1): |
index = min(minIndex + int(p*numValues)-1, maxIndex) |
| 232 |
if operator.contains(_range, _list[i]): |
|
| 233 |
minIndex = i |
adjusted = adjusted \ |
| 234 |
break |
or quantiles.has_key(index) \ |
| 235 |
|
or ((index - minIndex + 1) / float(numValues)) != p |
| 236 |
|
|
| 237 |
for i in xrange(len(_list)-1, -1, -1): |
quantiles[index] = 0 |
| 238 |
if operator.contains(_range, _list[i]): |
|
| 239 |
maxIndex = i |
quantiles = quantiles.keys() |
| 240 |
break |
quantiles.sort() |
| 241 |
|
|
| 242 |
|
# |
| 243 |
|
# the current quantile index must be strictly greater than |
| 244 |
|
# the lowerBound |
| 245 |
|
# |
| 246 |
|
lowerBound = minIndex - 1 |
| 247 |
|
|
| 248 |
numValues = maxIndex - minIndex + 1 |
for qindex in xrange(len(quantiles)): |
| 249 |
|
if lowerBound >= maxIndex: |
| 250 |
|
# discard higher quantiles |
| 251 |
|
quantiles = quantiles[:qindex] |
| 252 |
|
break |
| 253 |
|
|
| 254 |
if numValues > 0: |
# lowerBound + 1 is always a valid index |
| 255 |
|
|
| 256 |
# |
# |
| 257 |
# build a list of unique indices into list of where each |
# bump up the current quantile index to be a usable index |
| 258 |
# quantile *should* be. set adjusted if the resulting |
# if it currently falls below the lowerBound |
|
# indices are different |
|
| 259 |
# |
# |
| 260 |
quantiles = {} |
if quantiles[qindex] <= lowerBound: |
| 261 |
for p in percents: |
quantiles[qindex] = lowerBound + 1 |
| 262 |
index = min(minIndex + int(p*numValues)-1, maxIndex) |
|
| 263 |
|
listIndex = quantiles[qindex] |
| 264 |
adjusted = adjusted \ |
value = _list[listIndex] |
|
or quantiles.has_key(index) \ |
|
|
or ((index - minIndex + 1) / float(numValues)) != p |
|
|
|
|
|
quantiles[index] = 0 |
|
| 265 |
|
|
| 266 |
quantiles = quantiles.keys() |
# |
| 267 |
quantiles.sort() |
# look for similar values around the quantile index |
| 268 |
|
# |
| 269 |
|
lindex = listIndex - 1 |
| 270 |
|
while lindex > lowerBound and value == _list[lindex]: |
| 271 |
|
lindex -= 1 |
| 272 |
|
lcount = (listIndex - 1) - lindex |
| 273 |
|
|
| 274 |
|
rindex = listIndex + 1 |
| 275 |
|
while rindex < maxIndex + 1 and value == _list[rindex]: |
| 276 |
|
rindex += 1 |
| 277 |
|
rcount = (listIndex + 1) - rindex |
| 278 |
|
|
| 279 |
# |
# |
| 280 |
# the current quantile index must be strictly greater than |
# adjust the current quantile index based on how many |
| 281 |
# the lowerBound |
# numbers in the _list are the same as the current value |
| 282 |
# |
# |
| 283 |
lowerBound = minIndex - 1 |
newIndex = listIndex |
| 284 |
|
if lcount == rcount: |
| 285 |
for qindex in xrange(len(quantiles)): |
if lcount != 0: |
| 286 |
if lowerBound >= maxIndex: |
# |
| 287 |
# discard higher quantiles |
# there are an equal number of numbers to the left |
| 288 |
quantiles = quantiles[:qindex] |
# and right, try going to the left first unless |
|
break |
|
|
|
|
|
# lowerBound + 1 is always a valid index |
|
|
|
|
|
# |
|
|
# bump up the current quantile index to be a usable index |
|
|
# if it currently falls below the lowerBound |
|
|
# |
|
|
if quantiles[qindex] <= lowerBound: |
|
|
quantiles[qindex] = lowerBound + 1 |
|
|
|
|
|
listIndex = quantiles[qindex] |
|
|
value = _list[listIndex] |
|
|
|
|
|
# |
|
|
# look for similar values around the quantile index |
|
|
# |
|
|
lindex = listIndex - 1 |
|
|
while lindex > lowerBound and value == _list[lindex]: |
|
|
lindex -= 1 |
|
|
lcount = (listIndex - 1) - lindex |
|
|
|
|
|
rindex = listIndex + 1 |
|
|
while rindex < maxIndex + 1 and value == _list[rindex]: |
|
|
rindex += 1 |
|
|
rcount = (listIndex + 1) - rindex |
|
|
|
|
|
# |
|
|
# adjust the current quantile index based on how many |
|
|
# numbers in the _list are the same as the current value |
|
|
# |
|
|
newIndex = listIndex |
|
|
if lcount == rcount: |
|
|
if lcount != 0: |
|
|
# |
|
|
# there are an equal number of numbers to the left |
|
|
# and right, try going to the left first unless |
|
|
# doing so creates an empty quantile. |
|
|
# |
|
|
if lindex != lowerBound: |
|
|
newIndex = lindex |
|
|
else: |
|
|
newIndex = rindex - 1 |
|
|
|
|
|
elif lcount < rcount: |
|
|
# there are fewer items to the left, so |
|
|
# try going to the left first unless |
|
| 289 |
# doing so creates an empty quantile. |
# doing so creates an empty quantile. |
| 290 |
|
# |
| 291 |
if lindex != lowerBound: |
if lindex != lowerBound: |
| 292 |
newIndex = lindex |
newIndex = lindex |
| 293 |
else: |
else: |
| 294 |
newIndex = rindex - 1 |
newIndex = rindex - 1 |
| 295 |
|
|
| 296 |
elif rcount < lcount: |
elif lcount < rcount: |
| 297 |
# there are fewer items to the right, so go to the right |
# there are fewer items to the left, so |
| 298 |
|
# try going to the left first unless |
| 299 |
|
# doing so creates an empty quantile. |
| 300 |
|
if lindex != lowerBound: |
| 301 |
|
newIndex = lindex |
| 302 |
|
else: |
| 303 |
newIndex = rindex - 1 |
newIndex = rindex - 1 |
|
|
|
|
adjusted = adjusted or newIndex != listIndex |
|
| 304 |
|
|
| 305 |
quantiles[qindex] = newIndex |
elif rcount < lcount: |
| 306 |
lowerBound = quantiles[qindex] |
# there are fewer items to the right, so go to the right |
| 307 |
|
newIndex = rindex - 1 |
| 308 |
# |
|
| 309 |
# since quantiles is only set if the code is at least a little |
adjusted = adjusted or newIndex != listIndex |
| 310 |
# successful, an empty list will be generated in the case that |
|
| 311 |
# we fail to get to the real body of the algorithm |
quantiles[qindex] = newIndex |
| 312 |
# |
lowerBound = quantiles[qindex] |
| 313 |
if len(quantiles) == 0: |
|
| 314 |
return None |
if len(quantiles) == 0: |
| 315 |
else: |
return None |
| 316 |
return (adjusted, minIndex, maxIndex, |
else: |
| 317 |
[(q, (q - minIndex+1) / float(numValues)) \ |
return (adjusted, minIndex, maxIndex, |
| 318 |
for q in quantiles]) |
[(q, (q - minIndex+1) / float(numValues)) \ |
| 319 |
|
for q in quantiles]) |
| 320 |
|
|
| 321 |
CLR = 0 |
CLR = 0 |
| 322 |
STEP = 1 |
STEP = 1 |
Parent Directory
| 
Revision Log
| 
Patch