getDydmacs causes system failure

[mpadge]

This is part of what autotest does by way of testing all possible forms and values of inputs, with code taken directly from the example for getDydmacs. The lowerLV parameter shown here causes a complete R session crash:

dvn <- scrapeVarCross(dat = DRES, x_order = "sip", x_stem = "PRQC", x_delim1 = "_", x_delim2=".", x_item_num="\\d+", distinguish_1="1", distinguish_2="2")
qual.config.script <-  scriptCFA(dvn, lvname = "Qual", model = "configural")
qual.config.mod <- lavaan::cfa(qual.config.script, data = DRES, std.lv = F, auto.fix.first= F, meanstructure = T)
qual.dmacs <- getDydmacs(DRES, dvn, qual.config.mod, lowerLV = -.Machine$integer.max / 1000)

autotest can handle a lot of things, and will in the future diagnose these kinds of total meltdowns without actually generating them, but at the moment they are simply left to happen. That makes this pretty important to fix!! Thanks!

[jsakaluk]

This is quite interesting that only lowerLV throws the error, because the example in getDydmacs uses function defaults for the lowerLV and upperLV arguments:

qual.dmacs <- getDydmacs(DRES, dvn, qual.config.mod)
The default arguments for lowerLV and upperLV are very similar, -5 and 5, respectively. I'll take a look, but this may be an error where the devil is in the (difficult to spot) details.

[mpadge]

autotest uses defaults just to work out what kind of input it is. In that case, it identifies both as integers, and so tries the full range of possible integer values. lowerIV works okay for -Machine$integer.max by simply rejecting it, and does same for up to 1/100th of that value, but at 1/1000th, it obviously accepts the value as somehow valid and attempts to do something ... evidently destructive. Just try for yourself inputting that value, find the point where it all goes wrong, and, well, fix it. Should be fairly straightforward.

[jsakaluk]

Will do. Thanks :)