output | ||
---|---|---|
|
March 2nd: 1st deliverable.March 31st: Q1 report.May 4th: 2nd deliverable.- June 29th: 3rd deliverable.
- June 30th: Q2 report.
- August 31st: 4th report. End of the project.
Required by the R Consortium in the last communication about the Q2 report.
- Review the badge program and start the application process of achieving a badge (and inform the R Consortium). For
quantities
, and by extension, forunits
anderrors
. - Provide feedback using the survey.
Currently, there are parsers implemented in quantities
: parse_quantities
, parse_errors
, parse_units
. As a possible follow-up, we could try to integrate these parsers into readr
. Two possible lines of action:
- Submit those readers as a PR to
readr
.readr
should depend onquantities
(which is not on CRAN yet),errors
andunits
. - Work on
readr
to open its APIs (R and C++) to external parsers. Pros: adds value toreadr
without changing its dependencies. Cons: a little bit more work.
Either way, it requires approval from readr
's maintainer (Jim Hester).
The idea is to write a vignette (or several vignettes) describing how to work with quantities in the most typical workflows: R base, tidyverse, data.table. To that end, we should
- Identify all the most common high level operations: filtering, ordering, transformation, aggregation, grouped operations, (un)pivoting, joining... more? Then identify these operations for each workflow:
- R base:
[
andsubset
,order
,transform
,aggregate
,by
... - tidivyerse:
filter
,arrange
,transmute
andmutate
,summarise
,group_by
... - data.table: ... :)
- R base:
- For each workflow and operation, show usage examples, identify incompatibilities, find workarounds.
- Output:
- A set of working use cases.
- A set of incompatibilities and possible future action lines to solve them.
- follow, and refer to BIPM conventions/best practices
- implement in pillar methods:
- units in column header
- remove/replace <> with []; follow `units_options` that also sets plot output
Fitting linear models and extraction of regression coeficients as quantities.
- To be explored; viability TBD.
- Discuss support for heterogeneous units.
- Understand performance issues reported in
sf
in order to:- Define a reference benchmark (separate repo). It should be useful for
errors
too. - Further investigate the
xptr_cache
branch (units system enclosed in a C++ class):- Enables unit caching.
- The isolation provided centralises memory management.
- Enables an easy extension to support multiple units systems (long-term goal).
- Define a reference benchmark (separate repo). It should be useful for
- Support samples as input. Store the whole sample instead of a single value and an associated error. Operations work directly on these samples, so that every kind of correlation would be captured.
- More general alternative to the default Taylor-based mechanism.
- Representation? (matrix vs. list of vectors).
- Length of samples, resampling.
- use logo