R^4 #17: Dependencies A somewhat complex topic, and a brief illustration of what can go wrong. goo.gl/mE7uQK #rstats

Feb 28, 2018 · 11:02 PM UTC

4
12
4
30
Replying to @eddelbuettel
I had similar frustrations with the odbc package, thus github.com/msummersgill/odbc… . It’s nice to be able to write an ad-hoc shiny app that still happens to work 2 years later without going through the whole packrat rigamarole
1
3
Replying to @eddelbuettel
this just hit for me as I learned that install.packages(., dependencies = TRUE) installs even _suggests_ and so on (instead of the bare minimum required by 'Depends' and 'Imports'). My docker install ballooned from 25 to probably 100+ packages
Replying to @eddelbuettel
Most of those are small packages. For robustness we could embed them statically, but this incurs a large maintenance cost (including to CRAN) just to bubble up bugfixes.
1
1
Replying to @eddelbuettel
the reprex conundrum is more demonstrative of an issue with CRAN than an issue with complex deps. As long as the deps are well maintained there is no issue and we do maintain them. Why is processx CRAN submission stuck in limbos? We have no idea and no feedback about it.
1
6