Repository navigation
setDT not always sets ".internal.selfref" #6864
Description
Reading #6862 I found this behaviour:
ds = copy(datasets::iris)
x = setDT(get("ds"))
attr(ds, ".internal.selfref") # <pointer...>
ds = copy(datasets::iris)
x = setDT(get0("ds")) # note get0
attr(ds, ".internal.selfref") # NULL
Both ds's set as data.table's but in the second form, ds lacks selfref.
AFAIK setalloccol() is shallow copying ds, but then setDT fails to reassign it to the symbol ds because not recognizing 'get0' bindimg.
Arguably there is no easy way to detect the binding for every possible call that returns a data.table bound to a symbol, so I think setDT should warn when it is unable to assign it back -if a shallow copy is made-. or if ds lacks selfref before returning.
However, x does have it, and thus identical(ds, x) fails in second case.
Related perhaps #6754
Activity
MichaelChirico commentedon Mar 12, 2025
I think in this case we can continue the kludge pretty easily -- there's already logic for
get(), it was just written beforeget0()existed (? or at least was more well-known):data.table/R/data.table.R
Line 2990 in 3afc750
Note also #6702
venom1204 commentedon Mar 13, 2025
Hi @MichaelChirico @rikivillalba
Can I work on this issue .
As suggested by Michael, I'll update the existing logic in setDT() to handle get0() similarly to get():
Is this approach working good or is there anything else you would like to suggest?
Thanks!
MichaelChirico commentedon Mar 14, 2025
Let's get #6802 finished first, since it is likely to generate conflict with that change
venom1204 commentedon Jun 24, 2025
hi @MichaelChirico now as the issue #6702 is finished can i work on this issue
i have proposed solution for it.
thanks for your time
(textproc/R-data.table) Updated 1.16.4 to 1.18.4, Fix build against 4…