tbl_cube()thanks to Hadley Wickham PR #102.
Now depends on R 3.5.0.
Removed obscure numeric ‘what’ logic for
Fixed tidync method for
Found a really bad bug in
hyper_array() (#92), now fixed. Axis order in transforms was sometimes reversed, which caused garbled results from
hyper_tibble(). The effect is to ruin any ggplot2 figures for some source files. It’s very likely that no other system yet uses the
hyper_array() format so impact is low.
tidync is now part of the rOpenSci project.
Fixes in tests and examples to avoid version-incapable NetCDF problems on Solaris.
A number of improved tests and documentation fixes.
Deal with warnings from tidyr version > 0.8.3.
New class ‘tidync_data’ for output of
hyper_array(), no underlying change to the object which is simply a list of arrays from each variable, and axis transforms stored in an attribute.
Old deprecated function
axis_transforms() now Defunct.
FIRST RELEASE, tidync was greatly improved via help from the rOpenSci review process.
hyper_grids() to report available grid names.
A printing error of dimension value ranges is now fixed, thanks to James Goldie (#84).
Now supports ‘NC_CHAR’ type, by exploding these into the array size expected.
Breaking change: when using
tidync$gridit’s now expected that this must be
tidyr::unnest()ed in order to expand out the grid list per variable, in line with https://github.com/hypertidy/ncmeta/issues/26.
hyper_array function now stores the relevant transforms table as an attribute
transforms so that objects can be constructed directly from the native array output.
Now dependent on ncmeta >= 0.0.2, partly to avoid crashing on invalid source or file strings
hyper_index() and incorporated that into
hyper_filter(), there’s now only one delay-capable class which is ‘tidync’
hyper_filter() now uses a selection idiom, to record the state of the axis rather than explicitly filter it. This means we can have more flexibility on what the axis transform tables can be used for, and removes some unwieldy handling code. All the available axes are on the object from first contact, which means we can program against the entire space in the source which will help for complex mapping scenarios.
hyper filter() print now handles the case of char-type coordinate values by setting the min and max to NA_real_
Various improvements and fixes for the print method for tidync
Support coordinate-less dimensions has been added, there is new information in the print summary about which dimensions are a “coord_dim” and this results in the axis transform tables using the index as the coordinate value.
This version sees a new model where activation is on ‘grids’, effectively a space composed of dimensions. In addition to the variables, dimension, and attributes entities we add ‘grid’ defined by a set of dimensions, and ‘axis’ which is an instance of a particular dimension as used by a variable.
Sources without recognizable variables now gracefully handled, with help from ncmeta.