The (aggregated) history of every CRAN package ever
Mark Padgham, rOpenSci
ropensci.org
mark@ropensci.org
The (aggregated) history of every CRAN package ever
Mark Padgham, rOpenSci
pkgstats
Search "ropensci pkgstats"
Extract summary statistics of R package structure and functionality. Not all statistics of course, but a good go at balancing insightful statistics while ensuring computational feasibility.
What
statistics
?
What
statistics
?
- > 90 statistics
DESCRIPTIONstatistics- Authors, contributors
- URL(s)
- Dependencies
- Languages, files and lines-of-code
- "Objects" (or "functions" in R)
- Relations between objects
- Including across all languages (R → C; js → C++)
- External calls to all R functions
What
am I going to talk
about
?
- CRAN submissions
- Package metadata
- Package structure
- Documentation
- Functions
- Dependencies
- Stability
Cran submissions
Annual
submissions
New / total
submissions
No significant change
Package MetaData
Authors & Contributors
Why are there
two lines?
How to define CRAN in any one year?
- Snapshot at end of year
- Includes packages from prior years
- Average of all uploads in that year
Authors & Contributors
Packages with
URLs
Top 10
URLs
Top 10
Translations
Package Structure
Numbers of
Files
Numbers of
Functions
Numbers of
Functions
per File
Numbers of
Lines of Code
Lines of Code
per
Function
Documentation
Numbers of
Files
Numbers of
Vignettes
Function & parameter documentation
Package functions
Function
Call
s
Function
Call
Networks
With time:
- Function call networks become denser
- And constently more centralised
>With package maturity:
- Function call networks become far denser
- Enormously more centralised (>5×)
- And contain more discrete clusters
Package Dependencies and Network Structure
CRAN
package
dependencies
snapshots
uploads
CRAN
network
metrics
stability
Coupling in
stability
I =
ne
na + ne
n = N umbers of function calls not packages
Coupling in
stability
- More stable:
- Packages used by others
- Packages don't use others
- Less stable:
- Packages not used by others
- Packages only use others
Coupling in
stability
Imports by other packages
Summary
Summary
Getting Better:
- Numbers of authors & contributors
- Use of URLs (code hosting)
- Numbers of test files
- Numbers of functions
- Numbers of vignettes
- Centrality of function networks
Summary
Neutral:
Improvement through package maturity only
- Ratio of new/total submissions to CRAN
- Size of packages
- Function & parameter documentation
- Integration of function networks
- Proportion of non-imported packages (~60%)
- Average centrality of whole CRAN network
- Coupling stability & use by other packages
Summary
Getting Worse:
- Translations
- Use of base R functions in favour of contributed
- Numbers of data files
Conclusion: Update intervals
pkgstats
- Data updated and published daily
- Please use!
- Search "ropensci pkgstats"
- Use download function , or from GitHub
-
mark@ropensci.org