WISH: Built-in MD5 checksum calculator, e.g. `tools::md5(x)`

[HenrikBengtsson]

Background

R has a built-in MD5 checksum calculator tools::md5sum(), but it only operates on files. It takes a vector of pathnames (not connections) as input and returns a character string of the same length containing MD5 checksums, e.g.

> files <- dir(R.home(), pattern = "^C", full.names = TRUE)
> files
[1] "C:/PROGRA~1/R/R-3.2.5/CHANGES" "C:/PROGRA~1/R/R-3.2.5/COPYING"
> tools::md5sum(files)
     C:/PROGRA~1/R/R-3.2.5/CHANGES      C:/PROGRA~1/R/R-3.2.5/COPYING
"d45eec95ce49830fdd3277950397dbde" "0cce1e42ef3fb133940946534fcf8896"

Wish / Suggestion

Calculating MD5 checksums is such a common task that it would warrant a core R functions for calculating the checksum for an R object x, e.g. tools::md5(x).

There is an internal src/library/tools/src/md5.c file that implements the MD5 checksum. It even has an internal md5_buffer() function that seems to do exactly this.

See also

• digest package: Provides well-tested function digest::digest(x, algo="md5") for calculating the MD5 checksum for R object x.

[luciorq]

I would love to see an update on that front, is there any way to ping R core members responsible for utils?

[tdeenes]

yup, I tried it ~2.5 years ago on r-devel without success... See the convo here

[yihui]

FYI I asked an R core member in 2019 offline and he said it was not straightforward. Last week I had a precious chance to meet four R core members, and the good news is the larger sample size worked: wch/r-source@c91b845 Hopefully we will be able to use tools::md5sum(bytes = raw()) in the next version of R, which means we can serialize() an object to a NULL connection, get the raw bytes, pass them to md5sum(), and obtain the MD5 checksum.

[yihui]

Again FYI, here comes sha256sum(): wch/r-source@a7c34a7

[HenrikBengtsson]

These are excellent news. Now we just have to wait for R 4.5.0 to be released. I think we can close this wish.