37
|
1 # load sparse matrix package
|
|
2 suppressPackageStartupMessages(library('Matrix'))
|
|
3
|
|
4 # access a numeric column
|
|
5 get_numeric <- function(table, column_key) {
|
|
6 column <- as.numeric(column_key)
|
53
|
7 column_data <- suppressWarnings(as.numeric(as.character(table[column][[1]])))
|
37
|
8 return (c(column_data))
|
|
9 }
|
|
10
|
|
11 # access a label column
|
|
12 get_label <- function(table, column_key) {
|
|
13 column <- as.numeric(column_key)
|
55
|
14 column_data <- as.character(table[column][[1]])
|
|
15 return (c(column_data))
|
37
|
16 }
|
|
17
|
|
18 # inflate three columns into matrix
|
|
19 matrify <- function (data) {
|
|
20 if (ncol(data) != 3)
|
|
21 stop('Data frame must have three column format')
|
|
22 plt <- data[, 1]
|
|
23 spc <- data[, 2]
|
|
24 abu <- data[, 3]
|
|
25 plt.codes <- levels(factor(plt))
|
|
26 spc.codes <- levels(factor(spc))
|
|
27 taxa <- Matrix(0, nrow=length(plt.codes), ncol=length(spc.codes), sparse=TRUE)
|
|
28 row <- match(plt, plt.codes)
|
|
29 col <- match(spc, spc.codes)
|
|
30 for (i in 1:length(abu)) {
|
|
31 taxa[row[i], col[i]] <- abu[i]
|
|
32 }
|
|
33 colnames(taxa) <- spc.codes
|
|
34 rownames(taxa) <- plt.codes
|
|
35 taxa
|
|
36 }
|
|
37
|
|
38 # flatten data.frame into three column format
|
|
39 flatten <- function(my_matrix) {
|
|
40 summ <-summary(my_matrix)
|
|
41 summ <- data.frame(i=rownames(my_matrix)[summ$i], j=colnames(my_matrix)[summ$j], x=summ$x)
|
|
42 summ
|
|
43 }
|
|
44
|
|
45 # wrapper
|
|
46 wrapper <- function(table, columns, options) {
|
|
47
|
|
48 # initialize output list
|
|
49 l <- list()
|
|
50
|
|
51 # get number of columns
|
|
52 n = length(columns)
|
|
53
|
|
54 # consistency check
|
|
55 if (n %% 3 != 0) {
|
|
56 print ('heatmap::wrapper() - Data not consistent (n mod 3 != 0)')
|
|
57 return (l)
|
|
58 }
|
|
59
|
|
60 # create index sequence
|
|
61 index = seq(1, n, by=3)
|
|
62
|
|
63 # get keys
|
|
64 keys = names(columns)
|
|
65
|
|
66 # loop through blocks
|
|
67 for (i in index) {
|
|
68 # create columns
|
|
69 ci <- get_label(table, columns[keys[i]])
|
|
70 cj <- get_label(table, columns[keys[i+1]])
|
|
71 cx <- get_numeric(table, columns[keys[i+2]])
|
|
72
|
|
73 # create a frame from columns
|
|
74 my_frame <- data.frame(ci=ci, cj=cj, cx=cx)
|
|
75
|
|
76 # create matrix out of the frame
|
|
77 my_matrix <- matrify(my_frame)
|
|
78
|
|
79 # create/cluster matrix
|
|
80 row_order <- hclust(dist(my_matrix))$order
|
|
81 col_order <- hclust(dist(t(my_matrix)))$order
|
|
82
|
|
83 # reorder matrix
|
|
84 my_matrix <- my_matrix[row_order, col_order]
|
|
85
|
|
86 # transform back to three columns
|
53
|
87 my_flatmatrix = flatten(my_matrix)
|
37
|
88
|
|
89 # append to result list
|
|
90 l <- append(l, list(my_flatmatrix$i))
|
|
91 l <- append(l, list(my_flatmatrix$j))
|
|
92 l <- append(l, list(my_flatmatrix$x))
|
|
93 }
|
|
94
|
|
95 # return
|
|
96 return (l)
|
|
97 }
|