Notes

labnotebook

Updates to configuration to be more portable and lightweight:

  • Drop APIs for Mendeley and Github in favor of just pulling from atom/rss feeds. Doesn’t require a platform-specific API or authentication, and has proven more robust, if sometimes less flexible and fancy.
  • Drop git-modified and sha timestamps. It’s nice having these on the pages and included in the rdfa, but the information is accessible from the git repository and from Github. The overhead time in computing these values and successfully including them wasn’t worth it.
  • However, add a sha for the vita page using a simple filter.

  • Wishing I could just embed svgs into md output as XML, instead of as binary. See full description of issue: yihui/knitr#754.

  • Installed vim-pandoc’s dev versions, vim-pantondoc and vim-pandoc-syntax via pathogen; seem to be working rather nicely.

reproducible paper example

from richfitz mwpennel and co: A manuscript with examples executed by makefile (or R source script) and checked and built by travis: richfitz/wood. Wow. Particularly nice jobs in:

  • handling downloading of raw data from original (Dryad-based) sources. Makefile seems to handle doing this in a cache-friendly way that is lighter on Travis network.
  • providing the cleaned, process data as seperate data files in clearly marked output directory. Still, the output would be more modular if done in some more generic format than R’s binary serialization as rds files. They do also provide data on Dryad, which is presumably the processed data in a more sensible format, though their data DOI, 10.5061/dryad.v7m14, is still embargoed in advance of publication at the time of writing.
  • Travis checks not only that code runs, but that it downloads and that the manuscript and figures can be complied (by LaTeX), with supplement knitted and compiled as html.
  • in a very nice touch, Travis pushes html and pdf results back to Github on gh-pages.

Some very minor nitpicks:

  • manuscript isn’t strictly a literate (knitr) document, though figures are being produced on the fly from code by make. May be a sensible choice (though misses a few opportunities for dynamic inline values).
  • license not declared, but already coming soon richfitz/wood#8
  • discussion of citation of code: richfitz/wood#11
  • trouble in executing locally, see richfitz/wood#12

rfishbase