(some?) eval expressions have issues with :=

[eantonya]

dt1 = data.table(a = 1, key = 'a')
dt2 = data.table(c = 1, fn = list(quote(5*a)), key = 'c')

# this works
dt1[, eval(dt2[a]$fn[[1]], .SD)]
[1] 5

# this doesn't
dt1[, n := eval(dt2[a]$fn[[1]], .SD)]
#Error in eval(expr, envir, enclos) : object 'a' not found

Relevant SO post - http://stackoverflow.com/a/30899285/817778

[jangorecki]

This is very similar to #1180 but looks to be resolved already

dt1 = data.table(a = 1, key = 'a')
dt2 = data.table(c = 1, fn = list(quote(5*a)), key = 'c')
dt1[, n := eval(dt2[a]$fn[[1]], .SD)][] 
#       a     n
#1:     1     5

To be closed by adding unit test to prevent future regression.

[mattdowle]

Thanks @eantonya. Bisect testing showed it was fixed in v1.9.8 (Nov 2016), likely bug fix 16 for #880 :

16. DT[, LHS := RHS] with RHS is of the form eval(parse(text = foo[1])) referring to columns in DT is now handled properly. Closes #880. Thanks to tyner.

PR #4361 adds your example to the test suite for completeness.