Bug: .() mis-interpreted when being used as plotmath

[MichaelChirico]

I just found out that sometimes .() is needed to create dynamically-labeled plots in base R:

png("~/Desktop/dterror.png")
par(mfrow = c(1, 2))
DT[ , {
  plot(1:10)
  text(2, 9, bquote(R^2 == .(summary(lm(y ~ x))$r.squared)))}]

plot(1:10)
text(2, 9, bquote(R^2 == .(summary(lm(y ~ x, data = DT))$r.squared)))
dev.off()

When used outside of [.data.table, the text is converted (as expected) to the evaluated value of the regression. But [.data.table (I imagine) is detecting this as an alias for list and substituting, leading to the unexpected result on the left.

Workaround for now is to use with(DT, {...}) instead of calling [.data.table but obviously this won't work in situations requiring other arguments of [.data.table.

[jangorecki]

You can try bquote(DT[...]).

[MichaelChirico]

that would make the plot expression much more verbose... the bquote part is
two of maybe 20 lines within j to create the full plot.

if there's no fast way to to put a check on the dot replacement for its
being inside bquote, than so be it I guess... seeing as how this is the
first time I've come across it in ~3 yrs of using data.table, wouldn't be
the that big a deal in the grand scheme of things.

could add a replace_dot argument/option to be shut off in such rare cases?

On Nov 13, 2016 8:33 AM, "Jan Gorecki" notifications@github.com wrote:

You can try bquote(DT[...]).

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#1912 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AHQQdR4wXJhYq96dR-fGNEi45nBvTynaks5q9xGqgaJpZM4KwjmM
.

[jangorecki]

This should be a matter of changing "." value in this line. You could eventually put getOption("datatable.listalias", ".") and control that dynamically, but I don't think use case you mention is enough for putting such "feature." I would go with ?substitute (@MichaelChirico) instead of bquote.

[MichaelChirico]

Hmm, I tried other ways of getting around bquote but never got anything to work. Specifically?

[ecoRoland]

I second this since I occasionally plot in j and even have a package that does so. Just add a check for bquote:

replace_dot_alias <- function(e) {
  # we don't just simply alias .=list because i) list is a primitive (faster to iterate) and ii) we test for use
  # of "list" in several places so it saves having to remember to write "." || "list" in those places 
  if (is.call(e)) {
    if (e[[1L]] == "bquote") return(e) #check for bquote
    if (e[[1L]] == ".") e[[1L]] = quote(list)
    
    for (i in seq_along(e)[-1]) if ((!is.null(e[[i]])) && 
                                               (e[[i]] != "bquote")) #check for nested bquote
       e[[i]] = replace_dot_alias(e[[i]])
  }
  e
}

replace_dot_alias(quote(.(k)))
#list(k)
replace_dot_alias(quote(bquote(.(k))))
#bquote(.(k))
replace_dot_alias(quote(c(.(k), .(l))))
#c(list(k), list(l))
replace_dot_alias(quote(c(.(k), .(bquote(m == .(l))))))
#c(list(k), list(bquote(m == .(l))))

Currently, I have to use parse:

DT[, {
  i <- 42
  parse(text = sprintf("mu == %i", i))
}]