Dev/cohort assess

[emilliman5]

A minimal working cohort implementation.

Additions:

1. cohort_ref class and constructor.
2. cohort_assess wrapper
3. cohort_score wrapper
   1. missing summarize_scores
4. 3 cohort assessments
   1. dependency conflicts (missing or version mismatch)
   2. dependency tree bottlenecking
   3. namespace export conflicts

Many of the helper functions (class coercion, formatters, etc) were copy/pasted from pkg_* versions to get to a working example as well as to enforce/maintain a cohort_* vs pkg_* nomenclature.

Things left outstanding:

 1. `summarize_scores` for `cohort_metrics`
 2. assessment dispatch -- cohort assessments will get execututed against `pkg_refs`
 3. handle cases where underlying `pkg_ref` slot is `NA`/missing when executing assessments
 4. `cohort_score` result formatting
 5. unit testing
 6. vignette -- inprogress
 7. lots of other little things (documentation helpers, class coercion and formatting, etc.)
 8. more assessments
 9. Create better metric for reverse dependency assessment (remove igraph dependency, find metric that better captures dependency bottlenecks)

A brief example:

#Define a cohort of pkgs
pkgs <- c("broom", "cli", "crayon", "dbplyr", "dplyr", "forcats", 
          "ggplot2", "haven", "hms", "httr", "jsonlite", "lubridate",
          "magrittr", "modelr", "pillar", "purrr", "readr", "readxl",
          "reprex", "rlang", "rstudioapi", "rvest", "stringr", "tibble",
          "tidyr", "xml2")

#Create a list of pkg_refs
pr <- lapply(pkgs, pkg_cran)

#Construct a cohort_ref
cohortRef <- cohort_ref(pr, "recommended")

#Run cohort assessments
cohortAssess <- cohortRef %>% cohort_assess()

#Score cohort
cohortScore <- cohortAssess %>% cohort_score()

[emilliman5]

A design question to consider.

how to dispatch assessments that are specific for pkg_ref vs cohort_ref. We could modify all_assessments() and get_assessments() to select based on s3 classes. this would also require stricter requirements about assessment method class dispatch. another option would be to use a slightly different assessment naming convention (e.g. instead of assess_has_bug_url, assess_cohort_has_bug_url) however, if an assessment is applicable to both pkg_ref and cohort_ref this would cause some duplication

[codecov-commenter]

Codecov Report

Merging #248 (b45ba67) into master (19d429e) will decrease coverage by 5.44%.
The diff coverage is 18.01%.

@@            Coverage Diff             @@
##           master     #248      +/-   ##
==========================================
- Coverage   57.80%   52.36%   -5.45%     
==========================================
  Files          63       68       +5     
  Lines         986     1119     +133     
==========================================
+ Hits          570      586      +16     
- Misses        416      533     +117     

Impacted Files	Coverage Δ
R/assess_reverse_dependencies.R	30.76% <0.00%> (-69.24%)	⬇️
R/cohort_assess.R	0.00% <0.00%> (ø)
R/cohort_metric.R	0.00% <0.00%> (ø)
R/cohort_ref_class.R	0.00% <0.00%> (ø)
R/cohort_score.R	0.00% <0.00%> (ø)
R/pkg_metric.R	38.88% <ø> (ø)
R/pkg_ref_class.R	87.40% <0.00%> (-1.40%)	⬇️
R/assess_has_dependency_conflict.R	14.28% <14.28%> (ø)
R/assess_exported_namespace.R	69.23% <20.00%> (-30.77%)	⬇️
R/assess_dependencies.R	67.56% <66.66%> (+3.93%)	⬆️
... and 6 more

📣 Codecov can now indicate which changes are the most critical in Pull Requests. Learn more

[elimillera]

Looks really good. I updated the branch to develop to give us a better idea of how this works along with other changes

[elimillera]

I think we can remove the commented code here. I agree with the option to error out if there isn't anything in the cohort

[elimillera]

Looks really good. I updated the branch to develop to give us a better idea of how this works along with other changes