This repository has been archived by the owner on Jan 8, 2021. It is now read-only.
forked from NCAConsortium/Qualification-Requirements
-
Notifications
You must be signed in to change notification settings - Fork 0
/
report.helpers.R
218 lines (208 loc) · 7.67 KB
/
report.helpers.R
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
#' Generate a list of text grammatically correctly
#'
#' @param x a vector to be put into a list.
#' @param oxford.comma Should the Oxford comma be used?
#' @return A character string of the vector input separated by commas and the
#' word "and" as appropriate.
#' @export
comma.and <- function(x, oxford.comma=TRUE) {
ret <- as.character(x)
if (length(ret) == 0) {
warning("comma.and was given zero-length input")
ret <- ""
} else if (length(ret) == 1) {
## Do nothing
} else if (length(ret) == 2) {
## Only put an "and" between the inputs
ret <- paste(ret, collapse=" and ")
} else {
## Put commas between most of the entries, put "and" between the last two,
## and put an Oxford comma in before "and" if requested.
thisand <- " and "
if (oxford.comma) {
thisand <- ", and "
}
ret <- paste(paste(ret[-length(ret)], collapse=", "),
ret[length(ret)], sep=thisand)
}
ret
}
library(dplyr)
listtable <- function(data, listdef, name.col="Description", count=1) {
ret <- data.frame(stringsAsFactors=FALSE)
tmprow <- data.frame(Description=NA, stringsAsFactors=FALSE)[,c(),drop=FALSE]
name.col.current <- paste(name.col, count, sep=".")
for (i in seq_along(listdef)) {
if (is.list(listdef[[i]])) {
subresult <- listtable(data, listdef[[i]], name.col=name.col, count=count + 1)
tmp.wrapper <- data.frame(X=names(listdef)[i],
stringsAsFactors=FALSE)
names(tmp.wrapper) <- name.col.current
subresult <- cbind(tmp.wrapper, subresult)
ret <- bind_rows(ret, subresult)
} else if (is.character(listdef[[i]])) {
tmprow[,names(listdef)[i]] <- data[[listdef[[i]]]]
}
}
if (ncol(tmprow) > 0) {
ret <- bind_rows(tmprow, ret)
}
## Sort the columns of the output
depth.name.count <- length(grep(paste0("^", name.col), names(ret)))
if (depth.name.count > 0) {
depth.names <- paste(name.col, count:(count + depth.name.count - 1), sep=".")
} else {
depth.names <- c()
}
ret <- ret[,c(depth.names, setdiff(names(ret), depth.names)), drop=FALSE]
attr(ret, "name.col") <- name.col
class(ret) <- c("listtable", class(ret))
ret
}
summary.listtable <- function(object, ..., indent=" --- ", na.string="",
name.col=attr(object, "name.col")) {
name.cols <- grep(paste0("^", name.col), names(object), value=TRUE)
data.cols <- setdiff(names(object), name.cols)
ret <- data.frame()
if (length(name.cols) == 0) {
## Fully recursed
ret <- object[,data.cols]
} else {
ret <- data.frame()
## Operate on the first level of grouping and then recurse for subsequent levels
for (n in unique(object[[name.cols[1]]])) {
mask.n <- object[[name.cols[1]]] %in% n
if (identical(NA, n)) {
header <- data.frame()
content <- object[,data.cols]
} else {
header <- data.frame(X=n, stringsAsFactors=FALSE)
names(header) <- name.col
content <-
summary.listtable(object[mask.n, c(name.cols[-1], data.cols)], ...,
indent=indent, na.string=na.string, name.col=name.col)
}
## Clean up the name.col
if (name.col %in% names(content)) {
mask.na <- is.na(content[[name.col]])
if (sum(mask.na) == 1) {
content[mask.na, name.col] <- header[[name.col]]
if (any(!mask.na)) {
content[!mask.na, name.col] <- paste0(indent, content[!mask.na, name.col])
}
} else {
content[mask.na, name.col] <- na.string
content[,name.col] <- paste0(indent, content[,name.col])
content <- dplyr::bind_rows(header, content)
}
} else { ## name.col isn't in the names
if (nrow(content) == 1) {
content <- dplyr::bind_cols(header, content)
} else {
content[,name.col] <- paste0(indent, content[,name.col])
content <- dplyr::bind_rows(header, content)
}
}
ret <- dplyr::bind_rows(ret, content)
}
}
class(ret) <- c("summary.listtable", setdiff(class(ret), "listtable"))
attr(ret, "name.col") <- name.col
ret
}
as.data.frame.summary.listtable <- function(x, ..., digits=3, na.string="") {
name.col <- attr(x, "name.col")
charcount <- nchar(x[[name.col]])
x[[name.col]] <- sprintf(paste0("%-", max(charcount), "s"), x[[name.col]])
for (n in names(x)) {
if (is.numeric(x[[n]])) {
mask.na <- is.na(x[[n]])
x[[n]] <- PKNCA::signifString(x[[n]], digits=digits)
x[[n]][mask.na] <- na.string
}
}
class(x) <- "data.frame"
x
}
subkable <- function(x) {
charcount <- pmax(sapply(as.list(x), FUN=function(x) max(nchar(x))),
nchar(names(x))) + 1
colsep <- c("+", rep(c("|", "+"), nrow(x) + 1))
ret <- colsep
for (i in seq_len(ncol(x))) {
spformat <- paste0("%-", charcount[i], "s ")
ret <- paste0(ret,
## Add the header
c(strrep("-", charcount[i] + 1),
sprintf(spformat, names(x)[i]),
# Separate the header from the content
strrep("=", charcount[i] + 1),
# Add the content
rbind(sprintf(spformat, x[[i]]),
strrep("-", charcount[i] + 1))),
colsep)
}
knitr::asis_output(paste(c(ret, "\n"), collapse="\n"))
}
# Technique from http://htmlpreview.github.io/?https://github.com/wilkelab/cowplot/blob/master/inst/doc/shared_legends.html
plot_grid_one_legend <- function(...) {
args <- list(...)
## Extract the legend from the first plot.
grobs <- ggplotGrob(args[[1]])$grobs
legend <- grobs[[which(sapply(grobs, function(x) x$name) == "guide-box")]]
## Concatenate the plots together
ret <- list()
for (i in seq_along(args)) {
ret[[i]] <- args[[i]] + theme(legend.position="none")
}
ret[[i + 1]] <- legend
ret
}
# Font Color Format (from http://stackoverflow.com/questions/29067541/rmarkdown-how-to-change-the-font-color)
colortext <- function(x, color, outputFormat) {
if (missing(outputFormat))
outputFormat <- knitr::opts_knit$get("rmarkdown.pandoc.to")
if(outputFormat == 'latex')
paste("\\textcolor{",color,"}{",x,"}",sep="")
else if(outputFormat == 'html')
paste("<font color='",color,"'>",x,"</font>",sep="")
else
x
}
colortextbg <- function(x, color, alphapct=20, outputFormat) {
if (missing(outputFormat))
outputFormat <- knitr::opts_knit$get("rmarkdown.pandoc.to")
if(outputFormat == 'latex') {
# Requires \usepackage{color}
paste("\\colorbox{",color, "!", alphapct, "}{",x,"}",sep="")
} else if(outputFormat == 'html') {
color <- col2rgb(color, alpha=TRUE)
color["alpha",] <- color["alpha",] * alphapct/100
paste("<font style='background-color: ",
paste0("#", paste0(as.hexmode(color), collapse="")), "'>",
x,"</font>",sep="")
} else {
x
}
}
knit_print.sessionInfo <- function(x, ...) {
if (missing(x))
x <- sessionInfo()
packagename <- function(x) {
sha <- ifelse("GithubSHA1" %in% names(x),
paste(", GithubSHA1:", x$GithubSHA1),
"")
sprintf("%s(%s%s)", x$Package, x$Version, sha)
}
otherpkg <- sapply(x$otherPkgs, FUN=packagename)
namespacepkg <- sapply(x$loadedOnly, FUN=packagename)
ret <-
c(paste("*", c(x$R.version$version.string,
paste("Platform:", x$platform))),
paste("* Locale:", paste(strsplit(x$locale, ";")[[1]], collapse=", ")),
"* Packages",
paste(" * Base:", paste(x$basePkgs, collapse=", ")),
paste(" * Attached:", paste(otherpkg, collapse="; ")),
paste(" * Namespaces (not attached):", paste(namespacepkg, collapse="; ")))
knitr::asis_output(paste(ret, collapse="\n"))
}