Notes to self about this template folder

• Make controls all the things
  • Be pedantic about the makefile, especially as one R script may produce many plots and serial objects
• As a result, all R scripts in scripts/ are written to be run from the top level directory

When starting out a new report:

1. In ~/some-folder-path/PhD/Projects/, run mk-dated-dir a good title
2. cp -R 0000_project-folder-template/ yyyy-mm-dd_a-good-title/
3. cd yyyy-mm-dd_a-good-title/
4. Run ./run-me-after-copy.sh which will:
   • trash ./*/.gitignore to get rid of the .gitignore files that only exist so the directory structure of this template is preserved
   • trash ./.git to remove this git repository and initialise one for the project
   • trash README.md - should only exist in the original template directory and is just notes to myself
   • trash $0 delete itself
5. Change the name of the report-template.rmd file to something sensible
   • Change the corresponding entry in the .gitignore for the .tex file
6. Change the corresponding value of WRITEUP in the makefile
7. Change any other filenames containing report-template, ./run-me-after-copy.sh will tell you what to change.
8. Comment in/out the relevant things in the .gitignore

Log/

• Use futile.logger throughout the scripts to keep an eye on things
  • Note that there is a performance penalty for this - should not be used in tight loops

package/

• Sometimes, the scripts/common will be insufficient, and things should really be put in an R package
• In this case:
  • Uncomment the relevant line in the .gitignore
  • Initialise a repository inside package/packagename/
  • Add as a submodule to the top level git repo
    • Not easy: git submodule add ./package/pkgname/ ./package/pkgname-sub
    • Can't avoid duping the submodule
    • At least makes it easy to make make depend on the submodule and then rerun everything as a result of an update
    • Whole idea seems a little over-the-top
  • Make will need to know when git submodule update is run (TODO: Figure this out)

plots/

• Store as .png until one needs the .pdf, uses less of my time overall
• Use matching subdirectory names where possible

rds/

• Dead dove do not eat
• Again try to match subdirectory names with plots/, and sometimes scripts

scripts/

• For all varieties of scripts, usually R, sometimes C++ or sh
• Try to use subdirectories for smaller topics
• The scripts/common folder exists as a kind of mid way point between one huge R file and a package
  • The is useful for when I'm not sure if the code will be reused, or the best way to solve some problem I am likely to preform a number of times, without introducing the nonzero amount of overhead that packages introduce (particularly with Stan and C++ components that are to be reused)

tex-input/

• The report-template.rmd file makes liberal use of \input{tex-input/\\d{4}-a-tex-file.tex}, for syntax highlighting / editor reasons
• The files here also tend to be in subdirectories, which I name after the section of the document
• I name the file as: 0010-appropriate-filename.tex and the next tex file in the section of the report is named 0020-another-appropriate-file.tex
  • The label is then (fig|eqn|tbl|alg):appropriate-filename, so that I can easily remember the label names
    • Sometimes this conflicts because you have tex-input/example-one/0010-stage-one-target-posterior.tex, and tex-input/example-two/0010-stage-one-target-posterior.tex
    • Latex will helpfully tell you that you have nonunique labels defined
    • Append some identifier to the label? Unsure
  • The zero padding on either side of the filename numbering is so that I can easily insert new files without renaming everything the ordering
• The tex-input/pre.tex file contains all the things that rmarkdown / knitr put into the header of the resulting report-template.tex file