bch441-work-abc-units/.utilities.R

171 lines
4.7 KiB
R

# .utilities.R
#
# Miscellaneous R code to suppport the project
#
# Version: 1.2
# Date: 2017 09
# Author: Boris Steipe
#
# V 1.2 update database utilities to support 2017 version of JSON sources
# V 1.1 2017 updates for ABC-units
# V 1.0 First code
#
# ToDo:
# Notes:
#
# ==============================================================================
# ====== SCRIPTS =============================================================
source("./scripts/ABC-dbUtilities.R")
# ====== SUPPORT FUNCTIONS =====================================================
objectInfo <- function(x) {
# Function to combine various information items about R objects
#
# Input: an R object
# Value: none - prints information as side-effect
cat("object contents:")
print(x, digits = 22) # print value at maximal precision
cat("\nstructure of object:\n")
str(x)
if (! is.list(x)) { # Don't use cat() if x is a list. cat() can't handle lists.
cat("\nmode: ", mode(x), "\n")
cat("typeof: ", typeof(x), "\n")
cat("class: ", class(x), "\n")
}
# if the object has attributes, print them too
if (! is.null(attributes(x))) {
cat("\nattributes:\n")
attributes(x)
}
# Done
}
biCode <- function(s) {
# Make a 5 character "biCode" from a binomial name by concatening
# the uppercased first three letter of the first word and the first
# two letters of the second word. If there is only one word, we take the
# first five characters from that. Outputs are padded with "." if necessary.
# NAs in input are preserved.
# Parameters:
# s chr vector of binomial species names
# Value: chr vector of biCodes, same length as s, NAs are preserved
b <- character(length(s))
s <- gsub("[^a-zA-Z ]", "", as.character(s)) # remove all non-alphabetic
# characters except space
s <- toupper(s)
for (i in seq_along(s)) {
x <- unlist(strsplit(s[i], "\\s+"))
if (length(x) == 0) { # empty string
x <- c("", "")
} else if (length(x) == 1) { # only one string
x <- c(substr(x, 1, 3), substr(x, 4, 5)) # 3 + 2 with whatever is there
}
x <- paste0(x[1:2], "...") # pad strings
b[i] <- paste0(substr(x[1], 1, 3), substr(x[2], 1, 2))
}
b[is.na(s)] <- NA # recover NAs from input
return(b)
}
pBar <- function(i, l, nCh = 50) {
# Draw a progress bar in the console
# i: the current iteration
# l: the total number of iterations
# nCh: width of the progress bar
ticks <- round(seq(1, l-1, length.out = nCh))
if (i < l) {
if (any(i == ticks)) {
p <- which(i == ticks)
p1 <- paste(rep("#", p), collapse = "")
p2 <- paste(rep("-", nCh - p), collapse = "")
cat(sprintf("\r|%s%s|", p1, p2))
flush.console()
}
}
else { # done
cat("\n")
}
}
waitTimer <- function(t, nIntervals = 50) {
# pause and wait for t seconds and display a progress bar as
# you are waiting
t <- as.numeric(t)
if (t < 0.1) {return(invisible())}
increment <- t / nIntervals
bar <- "----:----|" # One module for the progress bar:
bar <- rep(bar, ceiling(nIntervals / 10)) # repeat,
bar <- unlist(strsplit(bar, "")) # split into single characters,
bar <- bar[1:nIntervals] # truncate,
bar <- paste(bar, collapse="") # and collapse.
cat(sprintf("\nWaiting: |%s\n |", bar))
for (i in 1:(nIntervals - 1)) {
Sys.sleep(increment)
cat("=")
}
Sys.sleep(increment)
cat("|\n\n")
return(invisible())
}
# ====== PDB ID selection ======================================================
selectPDBrep <- function(n) {
# Select n PDB IDs from a list of high-resolution, non-homologous, single
# domain, single chain structure files that represent a CATH topology
# group.
# Parameters n num number of IDs to return.
# Value: char PDB IDs
# Note: the list is loaded from an RData file in the data directory
load("./data/pdbRep.RData") # loads pdbRep
if (n > length(pdbRep)) {
stop(sprintf("You can select no more than %d IDs.", length(pdbRep)))
}
set.seed(as.numeric(Sys.time()))
return(sample(pdbRep, n))
}
# ====== DATA ==================================================================
# 10 species of fungi for reference analysis.
# http://steipe.biochemistry.utoronto.ca/abc/index.php/Reference_species_for_fungi
REFspecies <- c("Aspergillus nidulans",
"Bipolaris oryzae",
"Coprinopsis cinerea",
"Cryptococcus neoformans",
"Neurospora crassa",
"Puccinia graminis",
"Saccharomyces cerevisiae",
"Schizosaccharomyces pombe",
"Ustilago maydis",
"Wallemia mellicola"
)
# [END]