#### Regression tests for GRAPHICS & PLOTS -- requiring strict PDF equality pdf("reg-plot.pdf", paper="a4r", encoding ="ISOLatin1.enc", compress = FALSE, useDingbats = TRUE) ## since we supply the font metrics, the results depend only on ## the encoding used: Windows is different from Unix by default. options(warn = 1) # print as they occur plot(0) # this should remain constant str(par(c("usr","xaxp","yaxp"))) stopifnot(all.equal( par(c("usr","xaxp","yaxp")) , list(usr = c(0.568, 1.432, -1.08, 1.08), xaxp = c(0.6, 1.4, 4), yaxp = c(-1, 1, 4)))) ### Test for centring of chars. All the chars which are plotted should ### be centred, and there should be no warnings about ### font metrics unknown for character `?' par(pty="s") plot(c(-1,16), c(-1,16), type="n", xlab="", ylab="", xaxs="i", yaxs="i") title("Centred chars in default char set (ISO Latin1)") grid(17, 17, lty=1) known <- c(32:126, 160:255) for(i in known) { x <- i %% 16 y <- i %/% 16 points(x, y, pch=-i) } par(pty="m") ## PR 816 (label sizes in dotchart) ### Prior to 1.2.2, the label sizes were unaffected by cex. dotchart(VADeaths, main = "Death Rates in Virginia - 1940", cex = 0.5) dotchart(VADeaths, main = "Death Rates in Virginia - 1940", cex = 1.5) ## killed by 0 prior to 1.4.0 and in 1.4.1: t1 <- ts(0:100) ## only warnings about values <= 0 plot(t1, log = "y") plot(cbind(t1, 10*t1, t1 - 4), log="y", plot.type = "single") stopifnot(par("usr")[4] > 3) # log10: ylim[2] = 1000 ## This one needs to be looked at. ## lty = "blank" killed the fill colour too. plot(1:10, type="n") polygon(c(1, 3, 3, 1), c(1, 1, 3, 3), col="yellow", border="red", lty="blank") rect(6, 6, 10, 10, col="blue", border="red", lty="blank") ## in 1.5.0 all omit the fill colours. with(trees, symbols(Height, Volume, circles=Girth/24, inches=FALSE, lty="blank", bg="blue")) ## in 1.5.0 ignored the lty. ## axis() and par(mgp < 0) lt <-"31" # in R x <- seq(-2,3, len=1001) op <- par(tck= +0.02, mgp = -c(3,2,0)) plot(x, x^2 - 1.2, xaxt = "n", xlab="", type ='l', col = 2, main = "mgp < 0: all ticks and labels inside `frame'") x <- -2:3 lines(x, x^2 - 1.2, type ="h", col = 3, lwd=3) axis(1, pos = 0, at=-1:1, lty = lt, col=4)## col & lty work only from R 1.6 par(op) axis(1, pos = 0, at=c(-2,2,3), lty = lt, col=4) mtext(side=1,"note the x-ticks on the other side of the bars") ## plot.table(): explicit xlab and ylab for non-1D plot(UCBAdmissions)# default x- and y-lab plot(UCBAdmissions, xlab = "x label", ylab = "YY")# wrong in 1.5.1 ## axis suppression plot(tt <- table(c(rep(0,7), rep(1,4), rep(5, 3))), axes = FALSE) plot(tt, xaxt = "n") ## wrong till (incl.) 1.6.x ## legend with call lo <- legend(2,2, substitute(hat(theta) == that, list(that= pi))) stopifnot(length(lo$text$x) == 1) ## length() was 3 till 1.7.x plot(ecdf(c(1:4,8,12)), ylab = "ECDF", main=NULL) ## ylab didn't work till 1.8.0 plot(1:10, pch = NA) # gave error till 1.9.0 points(1:3, pch=c("o",NA,"x"))# used "N" try(points(4, pch=c(NA,FALSE)))# still give an error ## 'lwd' should transfer to plot symbols legend(1,10, c("A","bcd"), lwd = 2:3, pch= 21:22, pt.bg="skyblue", col = 2:3, bg = "thistle") ## (gave an error for 2 days in "2.0.0 unstable") x <- 2^seq(1,1001, length=20) plot(x, x^0.9, type="l", log="xy") ## gave error 'Infinite axis extents [GEPretty(1.87013e-12,inf,5)]' for R 2.0.1 plot(as.Date("2001/1/1") + 12*(1:9), 1:9) ## used bad 'xlab/ylab' in some versions of R 2.2.0(unstable) ## dotchart() restoring par() Opar <- par(no.readonly=TRUE) ; dotchart(1:4, cex= 0.7) Npar <- par(no.readonly=TRUE) ii <- c(37, 50:51, 58:59, 63) stopifnot(identical(names(Opar)[ii], c("mai","pin","plt","usr","xaxp","yaxp")), identical(Opar[-ii], Npar[-ii])) ## did not correctly restore par("mar") up to (incl) R 2.4.0 ## plot.function() [n=11, ... : since we store and diff PS file !] plot(cos, xlim=c(-5,5), n=11, axes=FALSE); abline(v=0) ## did *not* plot for negative x up to R 2.5.1 plot(sin, -2,3, xlim=c(-5,5), n=11, axes=FALSE, xlab="")# plot from -2 axis(1, at=c(-2,3), tcl=-1); axis(1, at=c(-5,5)) ## (from,to) & xlim should work simultaneously plot(cos, -7,7, n=11, axes=FALSE) ## gave wrong ylab in R 2.6.0 plot(cos, -7,7, ylab = "Cosine cos(x)", n=11, axes=FALSE) ## partial matching of 'ylab'; mapping [0,1] (not [-7.7]): ## margins chosen to avoid rouding error showing to 2dp. op <- par(mar=c(5,4.123,4,2)+0.1) plot(gamma, yla = expression(Gamma(x)), n=11, yaxt="n") par(op) ## plot.ts(x, y) could get the labels wrong in R <= 2.6.0: x <- ts(1:5);x1 <- lag(x, 2); plot(x1, x, axes=FALSE) # adding a curve in log scale : curve(5*exp(-x), 0.1, 100, n = 3, log="x", ylab="", axes=FALSE) curve(5*exp(-x), add=TRUE, n = 3, col=2,lwd=3) ## should fully overplot; wrong default xlim in 2.6.1 ## (and *slightly* wrong up to 2.6.0) ## Axis() calls via plot() {[xy]axt to keep *.ps small} x <- as.Date("2008-04-22 09:45") + (i <- c(0,4)) plot(x, xaxt="n")# not ok in 2.6.2, nor 2.7.0 plot(x, i, yaxt="n")# ok in 2.6.2 and 2.7.0 plot(i, x, xaxt="n")# ok in 2.6.2 and not in 2.7.0 ## table methods should be bypassed: dotchart(table(infert$education)) ## failed in 2.12.[12] ## cex as "..." in "high level" function hc <- hclust(dst <- dist(c(1:2, 5)), method="ave") plot(hc, cex = 2, axes=FALSE, ann=FALSE) ## cex was not used in 3.0.[01] ## axis.Date() and axis.POSIXct() with reversed 'xlim' toD <- as.Date("2016-08-19"); dates <- c(toD - 10, toD) plot(dates, 1:2, xlim = rev(dates), ann=FALSE, yaxt="n", frame.plot=FALSE) ## failed to label the dates in R <= 3.3.1 ## axis.Date() with various data types: x <- seq(as.Date("2022-01-20"), as.Date("2023-03-21"), by = "days") plot(data.frame(x, y = 1), xaxt = "n") axis.Date(1) axis.Date(3, at = "2022-04-01") axis.Date(3, at = as.Date("2022-07-01")) axis.Date(3, at = as.POSIXct(as.Date("2022-10-01"))) axis.Date(3, at = as.POSIXlt(as.Date("2023-01-01"))) axis.Date(3, at = as.integer(as.Date("2023-04-01"))) ## automatically extends the format: axis.Date(1, at = "2022-02-15", tck = -0.05, mgp = c(3,2,0)) ## axis.POSIXct() with various data types (2 minutes) x <- as.POSIXct("2022-10-01") + c(0, 60, 120) plot(data.frame(x, y = 1), xaxt="n") axis.POSIXct(1) axis.POSIXct(3, at = "2022-10-01 00:01") axis.POSIXct(3, at = as.Date("2022-10-01")) axis.POSIXct(3, at = as.POSIXct("2022-10-01 00:01:30")) axis.POSIXct(3, at = as.POSIXlt("2022-10-01 00:02")) axis.POSIXct(3, at = as.numeric(as.POSIXct("2022-10-01 00:00:30"))) ## automatically extends format (here: subseconds): axis.POSIXct(3, at = "2022-10-01 00:00:30.25", mgp = c(3,2,0)) ## axis.POSIXct: a few days, extending the format days <- seq(as.Date("2022-10-01"), as.Date("2022-12-21"), by="days") x <- as.POSIXct(as.character(days)) plot(data.frame(x, y = 1), xaxt="n") axis.POSIXct(1, x) axis.POSIXct(1, x, at = as.Date("2022-10-12"), mgp = c(3,2,0), tck = -0.04) axis.POSIXct(3, x, at = as.POSIXct("2022-10-15")) axis.POSIXct(3, x, at = as.POSIXlt("2022-10-15"), mgp = c(3,2,0)) axis.POSIXct(1, x, at = "2022-11-01 23:00", mgp = c(3,2,0), tck = -0.04) axis.POSIXct(3, x, at = "2022-11-01 06:00") axis.POSIXct(3, x, at = as.numeric(as.POSIXct("2022-12-01"))) ## axis() -- labels only written when there's room plot2 <- function(at, wait=FALSE) { plot1 <- function(x,y) plot(x,y, type="n", xlab="", ylab="", tck=0, frame.plot=FALSE) plot1( at, at); axis(3, at= at, tck=0) plot1(-at,-at); axis(3, at=-at, tck=0) if(wait) { mtext("Click here to advance!", line=-2, cex=1.5, col=2); locator(1) } } at <- c(7:15, 2*(8:15), 5*(7:15), 10*(8:15)) op <- par(mfrow=2:1, mgp = c(1.5, 0.6, 0), mar = .1+c(2,2,2,1), lab = c(20,20,7), las = 1) # las=1: all horizontal => y-axis perpendicular interAct <- dev.interactive() if(interAct) { xMar <- c(1,3)/2 ; nP <- 10 } else { xMar <- c(1,3)*3/2; nP <- 3 } ## Now increasing margins ==> decreasing plot area ==> shrinking plot ## (*is* device dependent [here have "a4r"-pdf]) : for(n in 1:nP) { par(mar = par("mar") + xMar) plot2(at, n < nP && interAct) } par(op) ##----- pairs() using verInd & horInd ---------- mpairs <- function(..., p=3, npp=2, main) { # p=4, npp=7 was Chris Andrews' example x <- matrix((1:(p*npp))/npp, ncol=p, dimnames=list(NULL, paste0("x", 1:p))) ## -> x \in [1/npp,.., p] <==> ceiling(x) \in {1,2,...,p} ## the panel function mP <- function(x,y, ...) { ## really made for the 'x = ...' above j <- ceiling(mean(y)) # in 1:p points(x,y, col=ceiling(mean(x)), pch= 3*j, cex=((p-1.5)*j+1)/p, ...) g <- 1:(p-1); abline(h=g, v=g, lty=3, col="gray33") } if(missing(main)) main <- if(...length()) sub("mpairs", '', deparse(sys.call())) lim <- range(x) k <- max(2, npp-2) ## bug in R(?): par(lab = c(1,1,7)) giving nonsense! op <- par(lab = c(k,k,7), oma=c(1,0,0,0)); on.exit(par(op)) pairs(x, panel=mP, xlim=lim, ylim=lim, main=main, cex.main=1, ...) } mpairs() # 4x4 matrix of scatterplots if(interAct) dev.set(if(length(dev.list()) < 2) dev.new() else dev.prev()) mpairs(horInd=1:2,verInd= 1:3 ) # 3x2 matrix: upper left [WRONG in R <= 3.5.0] mpairs(horInd=3:2) # 2x4: middle rows swapped mpairs( verInd=c(3,1)) # 4x2: cols 3,1 ## ## now all with 'row1attop=FALSE'======= if(interAct) dev.set(dev.prev()) mpairs(row1attop=FALSE) # 4x4 matrix of scatterplots -- perfect in R <= 3.5.0 if(interAct) dev.set(dev.next()) ## a version of those above, with 'row1attop = FALSE' mpairs(horInd=2:1,verInd=3:1 , row1attop=FALSE) # 3x2: *swapped* upper left mpairs(horInd=c(3,1) , row1attop=FALSE) # 2x3: swapped outer rows mpairs( verInd=c(3,2), row1attop=FALSE) # 3x2: cols 3,2 ## axis() when in subnormal range (x is subnormal if 0 < x < .Machine$double.xmin): (.min.exp.subnormal <- with(.Machine, double.min.exp + double.ulp.digits)) # -1074 (ry <- range(y74 <- 2^(.min.exp.subnormal + 0:50))) # all subnormal plotNchk <- function(y) { plot(y, log = "y", xlab="", ylab="", xaxt="n") # (as small pdf as possible) ry <- range(y) xp <- par("yaxp") ## and indeed yaxp do cover the range(y) : stopifnot(xp[1] <= ry[1], ry[2] <= xp[2], xp[3] == 1) invisible(xp) } if(interactive()) # not regularly, where pdf is stored plotNchk(y74) # gives 3 warnings; 1. from pretty(): "very small range" plotNchk(y74[1:8]) # 3 warnings *and* no error anymore plotNchk(y74[1:2]) # (ditto) ## dotchart(*, pch=., groups=*) -- PR#16953 ## dotchart(*, ylab=.) for groups; g <- rep(1:3, each=2) dotchart(VADeaths[1:2, 1:3], color=g, pch=g, ylab = "Grouping: {Urbanity . Gender} x Age", xaxt="n", frame.plot=FALSE) ## now pch and colors match groups; ## ylab placement; group (row) labels show again ## non-integer mgp[3] -- PR#18194 par(mar = c(5, 5, 5, 5)) plot.new() plot.window(xlim = c(0, 5), ylim = 0:1) if(dev.interactive()) { box(lty = 3) ## 'axis' puts line and labels in right place when 'mgp[3]' is integer mgp_bottom <- c(4, 2.5, 1) # 'mgp[1]' is arbitrary axis(1, at=c(1,4), mgp = mgp_bottom) mtext(c("labels", "line"), side = 1, line = mgp_bottom[2:3]) } ## Now, 'axis' puts line & labels in right place also when 'mgp[3]' is noninteger: mgp_top <- c(4, 2.5, 0.9) axis(3, at=c(1,4), mgp = mgp_top) mtext(c("labels are here", "line"), line = mgp_top[2:3]) ## They were one line too high in R <= 4.1.1