crash in analyse_FadingMeasurement()

[mcol]

Using the object object built in test_analyse_FadingMeasurement.R, I can crash the function with a simple call:

  # Create artificial object ------------------------------------------------
  l <- list()
  time <- 0
  for(x in runif(3, 120,130)) {
    irr  <- set_RLum(
        "RLum.Data.Curve",
        data = matrix(c(1:x, rep(1,x)), ncol = 2),
        originator = "read_XSYG2R",
        recordType = "irradiation (NA)",
        curveType = "simulated",
        info = list(
          startDate = format(Sys.time() + time, "%Y%m%d%H%M%S"),
          position = 1)
      )

    lum  <- set_RLum(
        "RLum.Data.Curve",
        data = matrix(c(1:40, exp(-c(1:40)/ x * 10)), ncol = 2),
        originator = "read_XSYG2R",
        recordType = "IRSL",
        curveType = "measured",
        info = list(
          startDate = format(Sys.time() + time + x + 30, "%Y%m%d%H%M%S"),
          position = 1)
      )

    time <- time + x + 60
    l <- c(l, irr, lum)
  }

object <- set_RLum("RLum.Analysis", records = l, originator = "read_XSYG2R")
analyse_FadingMeasurement(object, signal.integral = 1:2, background.integral = 3)

# Error in Tx_data[[x]] : subscript out of bounds

The crash occurs in this bit of code when x = 2 as our object has length(Lx_data) == 2 and length(Tx_data) == 1):

Luminescence/R/analyse_FadingMeasurement.R

Lines 395 to 398 in 759a114

	LxTx_table <- merge_RLum(.warningCatcher(lapply(1:length(Lx_data), function(x) {
	calc_OSLLxTxRatio(
	Lx.data = Lx_data[[x]],
	Tx.data = Tx_data[[x]],

@RLumSK What's the best approach? Stopping if length(Lx_data) != length(Tx_data), working only up to the longest common length, or something else again?

[RLumSK]

The best is here to throw a warning when the lengths differ and use the shortest common length.

[mcol]

Using the shortest common length, the error moves further down:

Error in data.frame(g_value, TC = tc, G_VALUE_2DAYS = g_value_2days[1],  : 
  arguments imply differing number of rows: 1, 0

That occurs because approx() fails, so T_0.5.predict and T_0.5.interpolated are NULL and the T_0.5 data.frame cannot be constructed correctly:

Luminescence/R/analyse_FadingMeasurement.R

Lines 591 to 613 in 759a114

	T_0.5.interpolated <- try(approx(x = LxTx_table[["LxTx_NORM"]],
	y = LxTx_table[["TIMESINCEIRR_NORM"]],
	ties = mean,
	xout = 0.5), silent = TRUE)
	if(inherits(T_0.5.interpolated, 'try-error')){
	T_0.5.predict <- NULL
	T_0.5.interpolated <- NULL
	}else{
	T_0.5.predict <- stats::predict.lm(fit_predict,newdata = data.frame(x = 0.5),
	interval = "predict")
	}
	T_0.5 <- data.frame(
	T_0.5_INTERPOLATED = T_0.5.interpolated$y,
	T_0.5_PREDICTED = (10^T_0.5.predict[,1])*tc,
	T_0.5_PREDICTED.LOWER = (10^T_0.5.predict[,2])*tc,
	T_0.5_PREDICTED.UPPER = (10^T_0.5.predict[,2])*tc
	)

Would it be OK in case approx() fails to generate directly such a data frame?

T_0.5 <- data.frame(
        T_0.5_INTERPOLATED = NA,
        T_0.5_PREDICTED = NA,
        T_0.5_PREDICTED.LOWER = NA,
        T_0.5_PREDICTED.UPPER = NA
    )

The tests we have pass with that change, but I don't think that we actually test the output in such cases.

[RLumSK]

Yes, this is a good solution!

[mcol]

Another problem came up in this snippet of code:

Luminescence/R/analyse_FadingMeasurement.R

Lines 696 to 699 in 759a114

	legend.pos = "outside",
	main = expression(paste(T[x], " - curves")),
	mtext = plot_settings$mtext
	)

The evaluation of the expression() term generated this error:

Error in `plot_RLum.Analysis(object = object, subset = NULL, ...)`: object 'x' not found

After a lengthy search, it turns out that using bquote() around the expression makes the error go away and the plot title appears correctly.

[mcol]

Fixed by #192.