int64 type in python, got numeric type in R

[ssh352]

First thank the authors for such a useful package!

I have a python function. It returns a data frame, in python the column is int64. I use reticulate to call this function in R and save the data in a data.table. I check the column type and it is now numeric. In this process some digits of the column values are changed. Any idea how to avoid this problem?

[eddelbuettel]

You can't as there is no (native) int64 in R, only a class wrapping a (64-bit) numeric giving int64 semantics.

[ssh352]

You can't as there is no (native) int64 in R, only a class wrapping a (64-bit) numeric giving int64 semantics.

Thank you Dirk, so perhaps I should convert the column from int64 to string in python first, pass it to R to avoid loss of precision, then use bit64 package as.integer64 to convert back to int64?

[eddelbuettel]

There is no loss of precision of int64 when the bit64 package is used to represent it. But I forget the details -- you may need to turn it on, or maybe it happens, or ... Maybe check the unit tests or docs for reticulate to see how it is used?

[kevinushey]

There's a package adding (optional, opt-in) bit64 support that I intend to review soon:

#590

[colearendt]

Related to #323