R package: funGraphs

A simple package to map out dependencies between functions in a directory / package using rlang (in contrast to other regex-based packages). This helps analyse packages and refactor code.

(The package works best for packages with < 50 functions. With larger packages, the diagram would contain too many nodes to be useful.)

---

News

21 Jan, 2019

• Added offline support. Previously the package uses the latest version of the svg.js components by default, but graphs fail to load after the recent update to v2.7 / 3.0. Now everything goes back to v2.6.6 and allows offline use.

14 Oct, 2018

• Added support to updating source graph from svg file: update_graph_from_svg (see Usage 1).

13 Oct, 2018

• [Hot-fix] Fixed problems with multiple functions having the same name, e.g. ".onLoad", in a package and function names having the disallowed symbol "|". See ?remove_duplicates and ?remove_vbar for more detail if you ran into troubles.
• Added a wrapper function to simplify usage: build_pkg_graph (see Usage 1).

---

Usage 1

# At the package directory
g <- funGraphs::build_pkg_graph()

# It's unlikely the graph-arrangement algorithm can get to exactly what you want; you can move the nodes around as you see fit. Once you are done, you can click 'save', then copy the generated svg string into a file and save it as a SVG file, e.g. "test.svg". To load a SVG file, copy the svg string into the textbox provided, then click 'load'. Note that this is read-only. If you wish to continue modifying the graph, use 'update_graph_from_svg' instead:

library(magrittr)
g %>% 
  update_graph_from_svg("test.svg") %>% 
  start_app()

Usage 2

library(funGraphs)
library(magrittr)

# At the package directory
g0 <- build_graph_from_dir("R/")    # Analyse directory and build 

# `prepare_graph` uses `igraph::layout_with_sugiyama` which works well
# with small (<20) number of nodes
g0 %>% 
  prepare_graph() %>%               # Add plotting parameters
  start_app()                       # Display app

# `prepare_graph_Rgraphviz` uses the default in `Rgraphviz` which
# works well for larger number of nodes.
g0 %>% 
  prepare_graph_Rgraphviz() %>%     # Add plotting parameters
  start_app()                       # Display app

---

Example 1: R package 'funGraphs'

From the diagram, we see that build_graph_from_dir, start_app and prepare_graph_Rgraphvizare at the top level. Indeed, they form the interface of the package. (Note that nodes with the same color are from the same file.)

---

Example 2: R package 'glue'

From the diagram, we see that glue_collapse and glue_data (and also as_glue.default, [.glue, [[.glue) separate the high-level calls from the low-level functions. From the grouping and naming of the functions, one would guess that

• The light yellow functions provide the core functionality,
• The bright green ones are the low-level utility functions,
• The dull green, blue, purple ones form the high-level usage functions.

(Minor manual modification, and removal of %||% and an extra .onLoad)