X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=R%2Fdist.topo.R;h=230f009e30d633935907327d2010f0846d2c9abd;hb=a2fc961dffe9f9b7994ed880e68c03b2334dc341;hp=fa9e869d7b0eb7b9d4925bd2538f640ea2a8a03f;hpb=52008daf7f708f3fcdc735f22af308dd1a461670;p=ape.git
diff --git a/R/dist.topo.R b/R/dist.topo.R
index fa9e869..230f009 100644
--- a/R/dist.topo.R
+++ b/R/dist.topo.R
@@ -1,9 +1,9 @@
-## dist.topo.R (2010-11-18)
+## dist.topo.R (2012-12-12)
## Topological Distances, Tree Bipartitions,
## Consensus Trees, and Bootstrapping Phylogenies
-## Copyright 2005-2010 Emmanuel Paradis
+## Copyright 2005-2012 Emmanuel Paradis
## This file is part of the R-package `ape'.
## See the file ../COPYING for licensing issues.
@@ -68,29 +68,28 @@ dist.topo <- function(x, y, method = "PH85")
## 'x' is a list of objects of class "phylo" possibly with no class
if (!is.null(attr(x, "TipLabel"))) return(x)
ref <- x[[1]]$tip.label
- if (any(table(ref) != 1))
- stop("some tip labels are duplicated in tree no. 1")
n <- length(ref)
- for (i in 2:length(x)) {
- label <- x[[i]]$tip.label
+ if (length(unique(ref)) != n)
+ stop("some tip labels are duplicated in tree no. 1")
+
+ ## serious improvement by Joseph W. Brown!
+ relabel <- function (y) {
+ label <- y$tip.label
if (!identical(label, ref)) {
if (length(label) != length(ref))
- stop(paste("tree no.", i, "has a different number of tips"))
+ stop(paste("tree ", y, "has a different number of tips"))
ilab <- match(label, ref)
- ## can use tabulate here because 'ilab' contains integers
if (any(is.na(ilab)))
- stop(paste("tree no.", i, "has different tip labels"))
-### the test below does not seem useful anymore
-### if (any(tabulate(ilab) > 1))
-### stop(paste("some tip labels are duplicated in tree no.", i))
-###
- ie <- match(1:n, x[[i]]$edge[, 2])
- x[[i]]$edge[ie, 2] <- ilab
+ stop(paste("tree ", y, "has different tip labels"))
+ ie <- match(1:n, y$edge[, 2])
+ y$edge[ie, 2] <- ilab
}
- x[[i]]$tip.label <- NULL
+ y$tip.label <- NULL
+ y
}
- x[[1]]$tip.label <- NULL
+ x <- lapply(x, relabel)
attr(x, "TipLabel") <- ref
+ class(x) <- "multiPhylo"
x
}
@@ -104,7 +103,7 @@ prop.part <- function(..., check.labels = TRUE)
##
ntree <- length(obj)
if (ntree == 1) check.labels <- FALSE
- if (check.labels) .compressTipLabel(obj) # no need to store cause uncompress later
+ if (check.labels) obj <- .compressTipLabel(obj) # fix by Klaus Schliep (2011-02-21)
for (i in 1:ntree) storage.mode(obj[[i]]$Nnode) <- "integer"
##
## The 1st must have tip labels
@@ -154,24 +153,31 @@ plot.prop.part <- function(x, barcol = "blue", leftmar = 4, ...)
mtext(attr(x, "labels"), side = 2, at = 1:n, las = 1)
}
-prop.clades <- function(phy, ..., part = NULL)
+prop.clades <- function(phy, ..., part = NULL, rooted = FALSE)
{
if (is.null(part)) {
+ ##
+ ## Are we going to keep the '...' way of passing trees?
obj <- list(...)
if (length(obj) == 1 && class(obj[[1]]) != "phylo")
- obj <- unlist(obj, recursive = FALSE)
+ obj <- unlist(obj, recursive = FALSE)
+ ##
part <- prop.part(obj, check.labels = TRUE)
}
- bp <- .Call("bipartition", phy$edge, length(phy$tip.label),
- phy$Nnode, PACKAGE = "ape")
- if (!is.null(attr(part, "labels")))
- for (i in 1:length(part))
- part[[i]] <- sort(attr(part, "labels")[part[[i]]])
- bp <- lapply(bp, function(xx) sort(phy$tip.label[xx]))
+
+ bp <- prop.part(phy)
+ if (!rooted) {
+ bp <- postprocess.prop.part(bp)
+ part <- postprocess.prop.part(part) # fix by Klaus Schliep
+ ## actually the above line in not needed if called from boot.phylo()
+ }
+
n <- numeric(phy$Nnode)
- for (i in 1:phy$Nnode) {
- for (j in 1:length(part)) {
- if (identical(all.equal(bp[[i]], part[[j]]), TRUE)) {
+ for (i in seq_along(bp)) {
+ for (j in seq_along(part)) {
+ ## we rely on the fact the values returned by prop.part are
+ ## sorted and without attributes, so identical can be used:
+ if (identical(bp[[i]], part[[j]])) {
n[i] <- attr(part, "number")[j]
done <- TRUE
break
@@ -181,7 +187,8 @@ prop.clades <- function(phy, ..., part = NULL)
n
}
-boot.phylo <- function(phy, x, FUN, B = 100, block = 1, trees = FALSE)
+boot.phylo <- function(phy, x, FUN, B = 100, block = 1,
+ trees = FALSE, quiet = FALSE, rooted = FALSE)
{
if (is.list(x) && !is.data.frame(x)) {
if (inherits(x, "DNAbin")) x <- as.matrix(x)
@@ -195,6 +202,8 @@ boot.phylo <- function(phy, x, FUN, B = 100, block = 1, trees = FALSE)
}
}
boot.tree <- vector("list", B)
+ if (!quiet) # suggestion by Alastair Potts
+ progbar <- utils::txtProgressBar(style = 3)
for (i in 1:B) {
if (block > 1) {
y <- seq(block, ncol(x), block)
@@ -202,14 +211,21 @@ boot.phylo <- function(phy, x, FUN, B = 100, block = 1, trees = FALSE)
boot.samp <- numeric(ncol(x))
boot.samp[y] <- boot.i
for (j in 1:(block - 1))
- boot.samp[y - j] <- boot.i - j
+ boot.samp[y - j] <- boot.i - j
} else boot.samp <- sample(ncol(x), replace = TRUE)
boot.tree[[i]] <- FUN(x[, boot.samp])
+ if (!quiet) utils::setTxtProgressBar(progbar, i/B)
}
+ if (!quiet) close(progbar)
for (i in 1:B) storage.mode(boot.tree[[i]]$Nnode) <- "integer"
storage.mode(phy$Nnode) <- "integer"
- ans <- attr(.Call("prop_part", c(list(phy), boot.tree),
- B + 1, FALSE, PACKAGE = "ape"), "number") - 1
+
+ pp <- prop.part(boot.tree)
+ if (!rooted) pp <- postprocess.prop.part(pp)
+ ans <- prop.clades(phy, part = pp, rooted = rooted)
+
+ ##ans <- attr(.Call("prop_part", c(list(phy), boot.tree),
+ ## B + 1, FALSE, PACKAGE = "ape"), "number") - 1
if (trees) {
class(boot.tree) <- "multiPhylo"
ans <- list(BP = ans, trees = boot.tree)
@@ -217,6 +233,61 @@ boot.phylo <- function(phy, x, FUN, B = 100, block = 1, trees = FALSE)
ans
}
+### The next function transforms an object of class "prop.part" so
+### that the vectors which are identical in terms of split are aggregated.
+### For instance if n = 5 tips, 1:2 and 3:5 actually represent the same
+### split though they are different clades. The aggregation is done
+### arbitrarily. The call to ONEwise() insures that all splits include
+### the first tip.
+postprocess.prop.part <- function(x)
+{
+ n <- length(x[[1]])
+ N <- length(x)
+ w <- attr(x, "number")
+
+ drop <- logical(N)
+ V <- numeric(n)
+ for (i in 2:(N - 1)) {
+ if (drop[i]) next
+ A <- x[[i]]
+ for (j in (i + 1):N) {
+ if (drop[j]) next
+ B <- x[[j]]
+ if (length(A) + length(B) != n) next
+ V[] <- 0L
+ V[A] <- 1L
+ V[B] <- 1L
+ if (all(V == 1L)) {
+ drop[j] <- TRUE
+ w[i] <- w[i] + w[j]
+ }
+ }
+ }
+ if (any(drop)) {
+ labels <- attr(x, "labels")
+ x <- x[!drop]
+ w <- w[!drop]
+ attr(x, "number") <- w
+ attr(x, "labels") <- labels
+ class(x) <- "prop.part"
+ }
+ ONEwise(x)
+}
+
+### This function changes an object of class "prop.part" so that they
+### all include the first tip. For instance if n = 5 tips, 3:5 is
+### changed to 1:2.
+ONEwise <- function(x)
+{
+ n <- length(x[[1L]])
+ v <- 1:n
+ for (i in 2:length(x)) {
+ y <- x[[i]]
+ if (y[1] != 1) x[[i]] <- v[-y]
+ }
+ x
+}
+
consensus <- function(..., p = 1, check.labels = TRUE)
{
foo <- function(ic, node) {
@@ -260,6 +331,7 @@ consensus <- function(..., p = 1, check.labels = TRUE)
## Get all observed partitions and their frequencies:
pp <- prop.part(obj, check.labels = FALSE)
## Drop the partitions whose frequency is less than 'p':
+ if (p == 0.5) p <- 0.5000001 # avoid incompatible splits
pp <- pp[attr(pp, "number") >= p * ntree]
## Get the order of the remaining partitions by decreasing size:
ind <- sort(unlist(lapply(pp, length)), decreasing = TRUE,