Version 0.4.0

[ressy]

I've just released version 0.4.0. This version is primarily just bug fixes and housekeeping to keep things running smoothly with newer R, RStudio, and the other R packages we use here. The below are some additional notes and ideas for future directions.

RStudio-less Installs

The CHIIMP documentation says RStudio is required, but strictly speaking it's only need to get Pandoc installed. Installing Pandoc alone without RStudio seemed to work fine during testing here, so this will probably be explicitly described in later versions. This could make setup a little faster and more streamlined for anyone not interested in the R aspect.

R Version Support

(This whole section is really only relevant those trying to use CHIIMP as an a R package.)

I made an effort to thoroughly test this version on a wide array of operating systems, R versions, and RStudio versions, including the latest available R 3 (3.6.3). In total I tested with:

• Windows 11 / Mac OS 10.15 / Ubuntu 20.04.4 LTS
• R 3.6.3 / R 4.2.1
• RStudio 2021.09 / RStudio 2022.07 / standalone Pandoc 2.19.2

I had originally intended to keep CHIIMP working with R 3 for as long as it was showing up as a default in common OS installs (for example Ubuntu 20.04 LTS is still providing 3.6.3 via Canonical's repos) but other R packages are dropping support faster than I realized, so that could be difficult. My initial thinking was that R 4's compatibility-breaking changes (e.g. stringsAsFactors) meant that there would still be wide support for the previous major version for some time, but evidently not. For our part we're already using R 4, and if users don't have a strong opinion on this, we'll drop official support for 3.6.3 next time.

CHIIMP Versioning and Compatibility

So far every version of CHIIMP has maintained backwards compatibility in terms of its use as a standalone program; the example dataset and configuration should still work in 0.4.0 as they did for 0.1.0. I don't think we've broken compatibility in the R package's public functions either, but that may be coming soon for some much-needed reorganization. Either of those kinds of changes would be noted in the NEWS file and announcements.

While versions are below 1.0, I will still try to ensure the behavior as a standalone program stays stable, and delay those kinds of compatibility-breaking changes, if any, until 1.0 and subsequent major versions. In the meantime, minor version number changes (like 0.4.0 to 0.5.0) may change R package behavior, while patch version number changes (0.4.0 to 0.4.1) should not change anything user-facing except for bug fixes. If and when CHIIMP actually reaches 1.0 that compatibility rule for minor versions should apply to the R package's public functions along with its standalone behavior, like you'd expect with proper semantic versioning.

Possible Future Changes

• I'm considering deprecating the prepare_dataset feature (that automatically builds a sample table from filename patterns, rather than using a ready-made CSV file). We haven't used that at all in quite some time and have found explicit tables of sample attributes to be much more reliable. This feature will most likely be deprecated in the next version and removed in 1.0, unless there's demand for it.
• There are some long overdue requests for optional trimming of adapters and primers and support for ambiguity characters in primers. I'm still hoping to address those, hopefully in the next minor release.
• The nested configuration options are proving hard to keep track of and difficult for users to manage, so a simplification is probably in order. In particular editing YAML files to change program options has been a stumbling block for our users, and since the ability to manage CSV spreadsheets is already required anyway, one obvious possibility is to support a simple two-column key=value CSV format for configuration options as an alternate format to the YAML.