Data Science. TileDB. Open Source. Quant Research. R. C++. Debian. Linux. Adjunct Clinical Professor, University of Illinois. Lots of coffee. And some running.
Chicago, IL, USA
Joined March 2007
- Tweets 5,361
- Following 350
- Followers 10,149
- Likes 7,646
Hey @espn here is another product idea: "data science battle royale" between base #rstats, #tidyverse, #pandas
... because friends still don't let friends use E@$el for real data work.
📢 EXCEL ESPORTS IS COMING TO @ESPN! This Friday, August 5, tune in for "Excel Esports: All-Star Battle" on #ESPN8: The Ocho.
We are beyond excited about this and hope that Excel Esports will be the most-viewed show! 😎
The show hosted by @MrExcel and @OzExcel
#excel
1
2
5
Sure:
a) `svn blame` (or ditto `git` if you use a mirror)
b) Look into actual release tarballs
#rstats
1
2
Glad you like it. We built on top of it with a package for the special case of showing students possible errors in their answers when using an autograder, see
cran.r-project.org/package=t…
for more. As a by-product this extends it for use the `tinytest` test runner we use. #rstats
2
7
I like `diffobj` by @BrodieGaslam for this. It has a number of options; a vanilla use (where my default theme makes the color difference a little hard to see) is below: `diffobj::diffObj(M, N)`.
See cran.r-project.org/package=d… for more.
#rstats
2
3
1
36
There *is* a difference between installing from a carefully maintained repository such as CRAN, versus closing your eyes and trusting a remote URL via `install_github`.
AFAIK we have not been had yet in #Rstats but the attack vector is obviously there.
I am uncovering what seems to be a massive widespread malware attack on @github.
- Currently over 35k repositories are infected
- So far found in projects including: crypto, golang, python, js, bash, docker, k8s
- It is added to npm scripts, docker images and install docs
5
21
Ah, flim, my bad, missed that. You may not have any control over the main manual (apart from conditional snippets in the Rd files).
1
2
Sounds like a use case for a pre-made pdf (that you control, including XeTeX if you need it) that is that wrapped inside an Rnw shim. I use that in a few packages in order to "freeze" pdfs in a failsafe manner. The original write-up by @markvdloo is at markvanderloo.eu/yaRb/2019/0…
1
5
That 2018 post may be outdated as you can now say
library(dplyr, include.only=c("mutate", "data_frame", "%>%"))
and if you then check for, say, `filter()` you see it is still the one from `stats` as `dplyr` was not imported as a whole, but only the declared parts. #rstats
3
7
31
r2u is at its core an `apt` repo, so `apt install` it is. Via `bspm` it is used by `install.packages()`.
Which means all ops NOT using `install.packages()` (i.e. RStudio GUI + internal, renv, rig, ...) do NOT benefit. Similarly, no versioned use of install.packages() either.
2
We can do this in five lines without `dockerfiler`:
FROM eddelbuettel/r2u:22.04
COPY DESCRIPTION /tmp
RUN install.r remotes \
&& cd /tmp \
&& Rscript -e 'remotes::install_deps(".", dependencies=TRUE)'
50 seconds, Depends + Suggests for moderately large package #rstats
1
1
2
GIF
Yes! Where rspm goes 'broad' (across OSs) we go 'deep' with r2u and integrate fully with apt. So a) you get all depends, no suprises and b) depends are known to the system so no more libicu upgrade breaking stringi etc. #rstats
See the r2u site for more:
eddelbuettel.github.io/r2u/
1
1
4
Minor reminder: We often make pre-releases of #Rcpp, #RcppArmadillo, ... available in the rcpp-drat repo. If you want to see what will come to CRAN and #Rstats (eventually) please install and test. I run these locally all the time too without woes.
rcppcore.github.io/drat/
3
5
Have you seen r2u with its packages which lets eg `install.packages("haven")` give you binaries with all dependencies?
eddelbuettel.github.io/r2u/
(And I know of no comprehensive complete solution. Also, technically, zlib1g-dev only needed to build, and zlib1g to run.)
1
2
4
I wondered about spokes too, but never found out. I care more about the Conti 'no flats' tires :) Bike is a charm, old (off Craigslist like the Dahon) but with a neat Shimano Ultegra set. Added half-fenders too as I commuted sometimes. It hates the pot holes here though.
1
The picture is fourteen years old (!!) and lacks the more recent daily-errand Breezer. I still ride the Trek when I (somewhat rarely) ride, haven't used or needed the Dahon in some time.
A few older posts are at dirk.eddelbuettel.com/blog/s…
1
5
A few folks aggregate here
strava.com/clubs/rstats-cycl…
(And there is a sibbling #rstats running club at the same base URL ending in just .../rstats )
1
4
16
New this week is #carbonLang as new direction / extension for #Cpp. Talk + disc by @chandlerc1024 are very good. For hybrid #Rstats/C++ folks some parallels on 'continuity + backwards comp' vs 'breaking and new stuff'. No single dominating choice either.
youtube.com/omrY53kbVoA
1
2
9
This cannot get covered in a reply thread here so maybe try to use the
- Writing R Extensions manual
- CRAN Repository Policy
- r-package-devel list
- 18000+ CRAN packages as models
Going with external compilation is the most difficult package setup. But it is possible. #rstats
1
1