rbindlist raises malformed factor

[jangorecki]

recent devel

rbindlist(list(data.table(a=as.factor(1:2), b=as.factor(2:3)),
          list(a=as.factor(3L), b=as.factor(4:5))))
#Error in as.character.factor(x) : malformed factor

error could be improved

on 1.12.2 it segfaults

[MichaelChirico]

Actually this still segfaults for me:

library(data.table)
# data.table 1.12.3 IN DEVELOPMENT built 2019-07-07 12:06:15 UTC using 4 threads (see ?getDTthreads).  Latest news: r-datatable.com
rbindlist(list(data.table(a=as.factor(1:2), b=as.factor(2:3)), list(a=as.factor(3L), b=as.factor(4:5))))

 *** caught segfault ***
address 0x0, cause 'memory not mapped'

Traceback:
 1: rbindlist(list(data.table(a = as.factor(1:2), b = as.factor(2:3)),     list(a = as.factor(3L), b = as.factor(4:5))))

R version 3.6.0 (2019-04-26)
Platform: x86_64-apple-darwin15.6.0 (64-bit)
Running under: macOS Mojave 10.14.5

Matrix products: default
BLAS:   /Library/Frameworks/R.framework/Versions/3.6/Resources/lib/libRblas.0.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/3.6/Resources/lib/libRlapack.dylib

locale:
[1] C/UTF-8/C/C/C/C

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

loaded via a namespace (and not attached):
[1] compiler_3.6.0

[MichaelChirico]

The segfault is caused here:

data.table/src/rbindlist.c

Lines 444 to 453 in 0ba9144

	for (int k=0; k<n; ++k) {
	SEXP s = thisColStrD[k];
	if (s!=NA_STRING && -TRUELENGTH(s)!=k+1) { nohop=false; break; }
	}
	if (nohop) memcpy(targetd+ansloc, INTEGER(thisCol), thisnrow*SIZEOF(thisCol));
	else {
	int *id = INTEGER(thisCol);
	for (int r=0; r<thisnrow; r++)
	targetd[ansloc+r] = id[r]==NA_INTEGER ? NA_INTEGER : -TRUELENGTH(thisColStrD[id[r]-1]);
	}

I don't 100% follow the logic of what's going on here, but I believe the break is meant to break out of a different loop than it's breaking (or there's meant to be two breaks).

If I change the memcopy section to also break, we still get the malformed factor error, but no segfault:

if (nohop) {
  memcpy(targetd+ansloc, INTEGER(thisCol), thisnrow*SIZEOF(thisCol));
  break; // <- new break
}

---

^ red herring explanation, but the line is indeed where the segfault is triggered

[mattdowle]

Recycling is missing in that section. Needed to recycle the length-1 as.factor(3L) in this example. Will fix ...