1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
|
#===========================================================================#
# caTools - R library #
# Copyright (C) 2005 Jarek Tuszynski #
# Distributed under GNU General Public License version 3 #
#===========================================================================#
#source('C:/programs/R/R-2.9.2/src/library/caTools/R/runfunc.R')
runmean = function(x, k, alg=c("C", "R", "fast", "exact"),
endrule=c("mean", "NA", "trim", "keep", "constant", "func"),
align = c("center", "left", "right"))
{
alg = match.arg(alg)
endrule = match.arg(endrule)
align = match.arg(align)
dimx = dim(x) # Capture dimension of input array - to be used for formating y
x = as.vector(x)
n = length(x)
if (k<=1) return (x)
if (k >n) k = n
k2 = k%/%2
y=double(n)
if (alg=="exact") {
.C("runmean_exact", x, y , as.integer(n), as.integer(k),
NAOK=TRUE, DUP=FALSE, PACKAGE="caTools")
} else if (alg=="C") {
.C("runmean", as.double(x), y , as.integer(n), as.integer(k),
NAOK=TRUE, DUP=FALSE, PACKAGE="caTools")
} else if (alg=="fast") {
.C("runmean_lite", as.double(x), y , as.integer(n), as.integer(k),
NAOK=TRUE, DUP=FALSE, PACKAGE="caTools")
} else { # the similar algorithm implemented in R language
k1 = k-k2-1
y = c( sum(x[1:k]), diff(x,k) ); # find the first sum and the differences from it
y = cumsum(y)/k # apply precomputed differences
y = c(rep(0,k1), y, rep(0,k2)) # make y the same length as x
if (endrule=="mean") endrule="func"
}
y = EndRule(x, y, k, dimx, endrule, align, mean, na.rm=TRUE)
return(y)
}
#==============================================================================
runmin = function(x, k, alg=c("C", "R"),
endrule=c("min", "NA", "trim", "keep", "constant", "func"),
align = c("center", "left", "right"))
{
alg = match.arg(alg)
align = match.arg(align)
endrule = match.arg(endrule)
dimx = dim(x) # Capture dimension of input array - to be used for formating y
x = as.vector(x)
n = length(x)
if (k<=1) return (x)
if (k >n) k = n
y = double(n)
if (alg=="C") {
.C("runmin", as.double(x) ,y , as.integer(n), as.integer(k),
NAOK=TRUE, DUP=FALSE, PACKAGE="caTools")
} else { # the similar algorithm implemented in R language
k2 = k%/%2
k1 = k-k2-1
a <- y[k1+1] <- min(x[1:k], na.rm=TRUE)
if (k!=n) for (i in (2+k1):(n-k2)) {
if (a==y[i-1]) # point leaving the window was the min, so ...
y[i] = min(x[(i-k1):(i+k2)], na.rm=TRUE) # recalculate min of the window
else # min=y[i-1] is still inside the window
y[i] = min(y[i-1], x[i+k2 ], na.rm=TRUE) # compare it with the new point
a = x[i-k1] # point that will be removed from the window next
if (!is.finite(a)) a=y[i-1]+1 # this will force the 'else' option
}
if (endrule=="min") endrule="func"
}
y = EndRule(x, y, k, dimx, endrule, align, min, na.rm=TRUE)
return(y)
}
#==============================================================================
runmax = function(x, k, alg=c("C", "R"),
endrule=c("max", "NA", "trim", "keep", "constant", "func"),
align = c("center", "left", "right"))
{
alg = match.arg(alg)
endrule = match.arg(endrule)
align = match.arg(align)
dimx = dim(x) # Capture dimension of input array - to be used for formating y
x = as.vector(x)
n = length(x)
k = as.integer(k)
if (k<=1) return (x)
if (k >n) k = n
y = double(n)
if (alg=="C") {
.C("runmax", as.double(x) ,y , as.integer(n), as.integer(k),
NAOK=TRUE, DUP=FALSE, PACKAGE="caTools")
} else { # the same algorithm implemented in R language
k2 = k%/%2
k1 = k-k2-1
a <- y[k1+1] <- max(x[1:k], na.rm=TRUE)
if (k!=n) for (i in (2+k1):(n-k2)) {
if (a==y[i-1]) # point leaving the window was the max, so ...
y[i] = max(x[(i-k1):(i+k2)], na.rm=TRUE) # recalculate max of the window
else # max=y[i-1] is still inside the window
y[i] = max(y[i-1], x[i+k2 ], na.rm=TRUE) # compare it with the new point
a = x[i-k1] # point that will be removed from the window next
if (!is.finite(a)) a=y[i-1]+1 # this will force the 'else' option
}
if (endrule=="max") endrule="func"
}
y = EndRule(x, y, k, dimx, endrule, align, max, na.rm=TRUE)
return(y)
}
#==============================================================================
runquantile = function(x, k, probs, type=7,
endrule=c("quantile", "NA", "trim", "keep", "constant", "func"),
align = c("center", "left", "right"))
{ ## see http://mathworld.wolfram.com/Quantile.html for very clear definition
## of different quantile types
endrule = match.arg(endrule)
align = match.arg(align)
dimx = dim(x) # Capture dimension of input array - to be used for formating y
yIsVec = is.null(dimx) # original x was a vector
x = as.vector(x)
n = length(x)
np = length(probs)
k = as.integer(k)
type = as.integer(type)
if (k<=1) return (rep(x,n,np))
if (k >n) k = n
if (is.na(type) || (type < 1 | type > 9))
warning("'type' outside allowed range [1,9]; changing 'type' to ", type<-7)
y = double(n*np)
.C("runquantile", as.double(x) ,y , as.integer(n), as.integer(k),
as.double(probs), as.integer(np),as.integer(type),
NAOK=TRUE, DUP=FALSE, PACKAGE="caTools")
dim(y) = c(n,np)
for (i in 1:np) { # for each percentile
yTmp = EndRule(x, y[,i], k, dimx, endrule, align, quantile, probs=probs[i], type=type, na.rm=TRUE)
if (i==1) {
if (is.null(dimx)) dimy = length(yTmp) else dimy = dim(yTmp)
yy = matrix(0,length(yTmp),np) # initialize output array
}
yy[,i] = as.vector(yTmp)
}
if (np>1) dim(yy) = c(dimy,np) else dim(yy) = dimy
return(yy)
}
#==============================================================================
runmad = function(x, k, center = runmed(x,k), constant = 1.4826,
endrule=c("mad", "NA", "trim", "keep", "constant", "func"),
align = c("center", "left", "right"))
{
endrule = match.arg(endrule)
align = match.arg(align)
dimx = dim(x) # Capture dimension of input array - to be used for formating y
x = as.vector(x)
n = length(x)
if (k<3) stop("'k' must be larger than 2")
if (k>n) k = n
y = double(n)
.C("runmad", as.double(x), as.double(center), y, as.integer(n),
as.integer(k), NAOK=TRUE, DUP=FALSE, PACKAGE="caTools")
y = EndRule(x, y, k, dimx, endrule, align, mad, constant=1, na.rm=TRUE)
return(constant*y)
}
#==============================================================================
runsd = function(x, k, center = runmean(x,k),
endrule=c("sd", "NA", "trim", "keep", "constant", "func"),
align = c("center", "left", "right"))
{
endrule = match.arg(endrule)
align = match.arg(align)
dimx = dim(x) # Capture dimension of input array - to be used for formating y
x = as.vector(x)
n = length(x)
if (k<3) stop("'k' must be larger than 2")
if (k>n) k = n
y = double(n)
.C("runsd", as.double(x), as.double(center), y, as.integer(n),
as.integer(k), NAOK=TRUE, DUP=FALSE, PACKAGE="caTools")
y = EndRule(x, y, k, dimx, endrule, align, sd, na.rm=TRUE)
return(y)
}
#==============================================================================
EndRule = function(x, y, k, dimx,
endrule=c("NA", "trim", "keep", "constant", "func"),
align = c("center", "left", "right"), Func, ...)
{
# Function which postprocess results of running windows functions and cast
# them in to specified format. On input y is equivalent to
# y = runFUNC(as.vector(x), k, endrule="func", align="center")
# === Step 1: inspects inputs and unify format ===
align = match.arg(align)
k = as.integer(k)
k2 = k%/%2
if (k2<1) k2 = 1
yIsVec = is.null(dimx) # original x was a vector -> returned y will be a vector
if (yIsVec) dimx=c(length(y),1) # x & y will become 2D arrays
dim(x) <- dimx
dim(y) <- dimx
n = nrow(x)
m = ncol(x)
if (k >n) k2 = (n-1)%/%2
k1 = k-k2-1
idx1 = 1:k1
idx2 = (n-k2+1):n
# === Step 2: Apply different endrules ===
if (endrule=="NA") {
y[idx1,] = NA
y[idx2,] = NA
} else if (endrule=="keep") {
y[idx1,] = x[idx1,]
y[idx2,] = x[idx2,]
} else if (endrule=="constant") {
y[idx1,] = y[k1+1+integer(m),]
y[idx2,] = y[n-k2+integer(m),]
} else if (endrule=="trim") {
y = y[(k1+1):(n-k2),] # change y dimensions
} else if (endrule=="func" || !yIsVec) {
for (j in 1:m) {
for (i in idx1) y[i,j] = Func(x[1:(i+k2),j], ...)
for (i in idx2) y[i,j] = Func(x[(i-k1):n,j], ...)
}
}
# === Step 3: Adjust aligment if needed ===
if (endrule!="trim") {
if (align=="left") {
y[1:(n-k1),] = y[(k1+1):n,]
if (endrule=="keep") {
y[(n-k1+1):n,] = NA
} else if (endrule=="func" || !yIsVec) {
for (j in 1:m) for (i in (n-k+1):n) y[i,j] = Func(x[i:n,j], ...)
}
} else if (align=="right") {
y[(k2+1):n,] = y[1:(n-k2),]
if (endrule=="keep") {
y[1:k2,] = NA
} else if (endrule=="func" || !yIsVec) {
for (j in 1:m) for (i in 1:k) y[i,j] = Func(x[1:i,j], ...)
}
}
} # if (endrule!="trim")
# === Step 4: final casting and return results ===
if (yIsVec) y = as.vector(y);
return(y)
}
|