Make it such that sample distribution scripts can be run from anywhere in project

[jaclyn-taroni]

Purpose/implementation

This is the possibly overly complicated thing mentioned #78 (review). It occurred to me that someone might navigate to the sample distribution directory and try to run the Rscripts, i.e.:

cd analyses/sample-distribution-analyses
Rscript 01-filter-across-types.R

This would raise an error because this is supposed to be run from the root of the project and the data, results, etc. directories are hardcoded as such.

This prompted me to try a strategy like what we have in lines 3-8 of ./scripts/run_in_ci.sh

@cbethell you were onto something when you mentioned setwd() in person on Wednesday. Using it is generally considered a bad idea for portability.

By adding the changes I have to the shell and R scripts in this module it ensures:

1. The shell script will always run as it is being called in the directory it lives in = will always be run as if it is in the same directory that the Rscripts it runs are in. The --vanilla addition is probably more important if this is run outside the Docker container.
2. The working directory for the Rscripts will always be the root of the project, as is intended. We can use rprojroot::find_root(rprojroot::has_dir(".git")) to find the root of the project. The criterion is where the .git folder lives. (The irony of using this in conjunction with setwd() is not lost on me.)

Issue

N/A - I don't think this should be a requirement or included as part of the instructions for #77. I was curious and wanted to see if it worked given how this project structure is a bit different from how I usually set up my own projects.

Directions for reviewers

If you are willing to check out my branch, I would try running these scripts using different places as your working directory.

Docker and continuous integration

• The dependencies required to run the code in this pull request have been added to the project Dockerfile.

[jaclyn-taroni]

With 0a26f6a, I've taken out setwd() and used a strategy that is more consistent with the intended use of rprojroot.

[cgreene]

Took a gander at how you were doing this. Nifty approach! It seems like we could point folks at this to ease their use during development. Not sure if you wanted this reviewed yet, so I just markedo ne thing I noticed as I looked at how this was done.

[cgreene]

Suggested change

	# The nonessential output produced in the root directory of this repository when we run `analyses/sample-distribution-analysis/run-sample-distribution.sh`
	# The nonessential output produced in the root directory of this repository when we run `run-sample-distribution.sh`

[cgreene]

I think that since it's in the same directory, the path should only be the script name.

[jaclyn-taroni]

Good idea -- added in 148f376

[jaclyn-taroni]

This is ready for review.

[jaclyn-taroni]

From CircleCI output:

[WARNING]: Empty continuation line found in:
    RUN apt-get update -qq && apt-get -y --no-install-recommends install     libgdal-dev     libudunits2-dev     libmagick++-dev
[WARNING]: Empty continuation line found in:
    RUN apt-get update -qq && apt-get -y --no-install-recommends install     && install2.r --error     --deps TRUE     gdalUtils     leafem     lwgeom     stars     leafpop     plainview     sf     mapview
[WARNING]: Empty continuation line found in:
    RUN apt-get update -qq && apt-get -y --no-install-recommends install     && install2.r --error     --deps TRUE     R.utils     treemap     d3r     hexbin     rprojroot
[WARNING]: Empty continuation lines will become errors in a future release.

What I've read suggested that is due to the presence of comments within the RUN statements, so I moved them in dcd9b23 while I was at it.

[cgreene]

This looks good to me, but I'd appreciate a pair of eyes that are more familiar with R on it. @cbethell or @cansav09 - could one of you take a look this morning too?

[cansavvy]

My comments may be me misunderstanding how to use the changes made to this script.

[cansavvy]

Is this assuming you have set your current directory as the sample-distribution-analysis?

[cansavvy]

With these new additions can we have a sentence or two here that clarifies what this script will or won't assume?

[jaclyn-taroni]

If you are in the directory with the README, you are in the directory where the bash script is and yes I can say that the script will always run as if you are calling it from the directory it lives in. The purpose of the changes are so these scripts will run correctly no matter what your working directory is when you call it.

[cansavvy]

Cool. Then my running of the scripts were appropriate tests then. Just wanted to check that everything it was and wasn't doing was what was expected. Worked for me!

[cansavvy]

LGTM!