A cognostic that is entirely NA causes a cryptic error message.

[jrounds]

kv = lapply(1:10, function(i) list(k = i, v= i))
ddo0 = ddo(kv)
cogFn = function(i) list(cog1 = 1, cog2 = NA)
t = tempfile()
vdb_conn = vdbConn(t, autoYes=TRUE)
makeDisplay(ddo0, panelFn=function(v) plot(1,1), cogFn=cogFn, conn= vdb_conn, name="break things")

Error in data.frame(name = nms[i], tmp, stringsAsFactors = FALSE) : 
  arguments imply differing number of rows: 1, 0

If you are like me you are like what?

The bug bites here:
https://github.com/tesseradata/trelliscope/blob/1a651ad231546b592236cb7251fa31d1f7f1adc6/R/cognostics.R#L413

The issue is that when the column of cog is NA this line is empty:
tmp <- attr(x[[i]], "cogAttrs")

when it is empty the row count is mismatched on the next line:
data.frame(name = nms[i], tmp, stringsAsFactors = FALSE)

Two ideas present themselves:

1. Catch and give a stop with a decent error message. Difficultly: easy.
   if(all(is.na(...))) stop("At least one cognostic of _name_ must not be NA")
2. Make it work. Technically not broken? Maybe it is. Difficulty: little harder. Involves checking tmp and also making sure everything down streams of it still works with all NA. Maybe not worth it. An NA column has no decision-able information it is just taking up space.

I think it is worth fixing because you really need to back up and think "what did I do last?" to figure this one out. In my case I added a bunch of cognostics and one was always NA.
#1 sounds safest to me. Just catch it and give it a decent message. Want a pull request?