ggalluvial

This patch fixes a bug introduced in v0.12.0 to is_lodes_form(), which introduced errors when an alluvium appears in multiple panels of a faceted plot. See issue #65 for details.

This patch fixes a bug introduced to stat_flow() in v0.12.0. See issue #64 for details.

In addition to upkeep and tidying, this release includes three significant upgrades:

• more intuitive parameters for ordering strata and lodes
• additions and documentation for computed variables
• several curve alternatives to x-splines

See NEWS.md for a comprehensive list of changes.

In response to ggplot2 v3.2.0, which removes the plyr dependency, the dependency has been removed from ggalluvial as well.

This patch corrects an error in self_adjoin() that returns an empty data frame when x is continuous-valued. A new example in stat_stratum() illustrates the debugged functionality.

This patch corrects outdated error throwing at negative observations in alluvia-form data and includes new examples using negative observations.

Prompted by use in the wild, this release extends functionality to handle negative-valued observations (y) analogous to their handling by geom_bar(). Additionally, the internals are synchronized across the statistical transformations and several stat parameters are reconfigured or expanded. See NEWS.md for details. Finally, basic tests of the stats (based on output y values) and the geoms (using vdiffr) are added.

Several outstanding problems with various parameters have been resolved, including inconsistent naming conventions, incompatibility with ggplot2, lack of defaults, unnatural behavior, and deprecation (previously deprecated params are now defunct). Also, a new vignette details how strata and lodes are ordered at each axis.

Some bugs are fixed in the positioning of lodes without incident flows and in the aesthetic reordering of flows. sessioninfo is suggested rather than devtools.

The self_adjoin() function, invoked by geom_flow(), is exported, in order to make the internals more transparent. Also, several long-deprecated parameters are now defunct.

Vertical scale transformers like scale_y_continuous() don't recognize the weight aesthetic parameter, so alluvium weights should now be passed to the y aesthetic. See the NEWS for details.

It's bothered me for a while that the alluvial data functions don't accept unquoted variable names. Through a combination of reading and experimentation, i've now gotten them to. I also simplified the naming convention for these functions and deprecated some unnecessary functions and parameters. There may be additional kinks to work out before this version proceeds to CRAN; please let me know if you encounter them!

This version introduces parameters diffuse, distill, and discern to the to_lodes() and to_alluvia() functions that convert alluvial data between the wide and long formats, in order to more intuitively handle auxiliary variables (i.e. variables that are not used as axes in the diagram). Other changes resolve CRAN checks, debug layer internals, and improve documentation.