[Rd] summary() dispatch puzzle

[Ben Bolker]

  I'm sorry I haven't boiled this down to a more-minimal example yet,
but ...

  I'm working on an S3 method (tidy.merMod, in the 'broom' package). It
normally handles 'merMod' objects from the lme4 package, but I'm trying
to make it handle 'merModLmerTest' objects from the lmerTest package too.

  The merModLmerTest class inherits (as an S4) class from the merMod
class.  The important difference (for my current purposes) is that
summary.merMod returns an object containing a coefficient table
*without* degrees or freedom or p-values, while the summary method for
merModLmerTest objects returns one *with* both of those columns.

 Because merModLmerTest inherits from merMod, calling tidy(obj) on a
merModLmerTest object does get you into broom:::tidy.merMod.  Within
that function, though, calling summary() appears to call summary.merMod
instead of merModLmerTest, so I don't get the additional columns I want.

  Ideas/directions for further diagnoses?  Is this an S3/S4 confusion,
or something funny about the way the broom and merModLmerTest package
are talking to each other, or ... ?

  cheers
    Ben Bolker


----
library(lme4)
fm1 <- lmer(Reaction~Days+(Days|Subject),sleepstudy)
library(lmerTest)
fm1B <- as(fm1,"merModLmerTest")
coef(summary(fm1B))               ## table with df and p values
coef(lme4:::summary.merMod(fm1B)) ## table without df and p values

is(fm1B,"merMod")  ## TRUE

library(broom)
tidy(fm1B)
## hard to show what's going on without debugging through the function:
## the issue occurs here:
## https://github.com/dgrtwo/broom/blob/master/R/lme4_tidiers.R#L94

I can't replicate this behaviour with a trivial method ...

silly <- function(x, ...) {
    UseMethod("silly")
}
silly.merMod <- function(object, ...) {
    coef(summary(object))
}
silly(fm1B)
environment(silly) <- environment(tidy)
silly(fm1B)