inst/response_to_reviewers.md

In ropensci/citesdb: A High-Performance Database of Shipment-Level 'CITES' Trade Data

Response to reviewers

Thanks to both reviewers for their comments and issues that have helped us to make a more user-friendly package.

Changes implemented

• We have changed the installation method in the README to use devtools.

• We have added information to the cites_status table indicating the location of the database files on disk.

• We now clear out and disconnect the database tables after updating, which should trigger disk cleanup and avoid doubling database size.

• Thanks for the reviewers for sticking with us through a long and merry chase of a bug that was preventing package-building in https://github.com/ecohealthalliance/citesdb/issues/1, which had its origins in a missing token for use with the rcites package as well as a low-level database lock issue. We now cache this the rcites information to avoid having to make remote calls in vignette-building, and also have resolved the DB locking conflict.

• We have modified our linter tests to avoid the false positives shown in https://github.com/ecohealthalliance/citesdb/issues/2.

• We have renamed the internal function check_status() to cites_check_status() to be less generic.

• We have made more elaborate help and examples for the cites_db(), cites_shipments(), and metadata functions to illustrate their use and distinguish between dplyr- and DBI-based workflows.

Changes not implemented / justifications

• We have opted not to use the glue package and instead stick with the base R paste() functions in the interest of limiting dependencies. We believe this is a minor trade-off.

• The low test coverage shown in https://github.com/ecohealthalliance/citesdb/issues/4 is due to tests skipped on CRAN. Setting the environment variable to NOT_CRAN=true shows that our test coverage is 65%. This is lower than typical, but as we note in the issue above, this is largely due to the verbose code for interacting with the RStudio connection pane, which cannot be tested except in an interactive session. Other code coverage is greater than 90%. We believe that all core functionality is tested, including important edge-cases and conditions not reflected in the coverage statistic, such as error handling in multiple sessions and changing up upstream data sources.

A final note: we have learned from the maintainers of MonetDBLite that it will not be returning to CRAN (its current iteration fails on R-devel), but they are working on a successor embedded database package that will replace it and go to CRAN later this year. So, for now, we will host this package on GitHub and replace the database back-end and send to CRAN when the successor package is ready. We've added a note to the README showing that users need package build tools for the current version.

ropensci/citesdb documentation built on May 13, 2023, 1:53 a.m.