TODO.md

colorjam Todo

11dec2024

• Improve blend_color() handling of alpha so it builds over time, use formula: a + (1-a) * b where a,b are alpha values between 0 and 1.

TODO 12jul2024

• blend_colors() - consider preset=getOption("colorjam.blend.preset") so it can be easily customized with a fixed scheme.

• Fix package dependencies so it does not also load jamba upon load.

• color_pie(): make it more similar to jamba::showColors()

  • color_pie(rainbowJam(12)) should show hex color as labels by default, same as jamba::showColors()

  • Consider argument inner_radius to make donut plots.

  • Consider replacing label_radius with combination of

    • label_just for label justification relative to its position, where

      • just=0 means the point is on the left(bottom), and
      • just=1 means the point is on the right(top).

    • label_pos for label position within the inner/outer radius:

      • label_pos=1 positions at the outer radius
      • label_pos=0 positions at the inner radius

• Consider new color display function color_spiral()

  • Alternative to color_pie() to arrange colors around the color wheel with concentric rings. Consider using circlize for spiral display?
  • Goal is to show a lot of colors in an organized way to facilitate picking a color from a pre-defined set.

• Consider adding dependency on farver>=2.1.2 to confirm fixed functions

• Improve the color wheels (angles for color breaks)

  • Revisit the color calculations given farver>=2.1.2 which fixed bug related to color distance calculations, and nearest color calculations.

  • "dichromat2"

    • not enough actual red, too much brown/salmon/hot pink
    • some of this problem is with the C/L sequence not producing a good red

  • "dichromat"

    • too much salmon/orange not enough gold, transition from orange to gold seems too slow
    • using n=6 "mediumpurple1" (H=272), "purple_illusion" (H=279) are too similar.
    • Maybe spread colors in four quadrants: gold, red, purple, blue?

  • "ryb"

    • too much fuschia/hot pink in the color wheel, with n=12 two sections are hot pink
    • with n=36 it looks nice, except too much pink

  • "ryb2"

    • red does not occur reliably after n=4, becomes brown/salmon/pink
    • even with n=36 WHERE IS THE RED? (Haha)

• Consider finally porting the Fibonacci color method.

  • Basic premise: Use magic number phi to represent a Fibonacci spiral around the color wheel.

TODO 24apr2024

• closest_named_color() and closestRcolor()

  • Consider option to return the input/output values in a data.frame:

    • hex color value
    • H,C,L values
    • distance value calculated

  • Consider alternative color distance function.

    • Standard HCL color distance function.
    • Consider farver::compare_colour() with some adjustment for hue difference.

TODO 14mar2024

• DONE. Bug: blend_colors() does not allow changing the preset

  • DONE. h2hw() and hw2h() prioritize preset over h1,h2 which causes blend_colors() to use the global preset, since it does not pass preset="custom" to enable over-riding the default h1,h2 values.

TODO 08jan2024

• Consider new accessible color wheel.

  • Start with red-yellow-blue color wheel.
  • Remove wedges of green/yellow-green.
  • Remove corresponding complement with fuschia/pink.
  • Test remaining wheel for color-blindness friendly distinction.

TODO 14nov2023

• Custom showColors() specifically for sorted colors

  • indicate along the x-axis where the numeric breaks occur
  • most often sorted by Hue, then indicate breaks along x-axis
  • determine numeric values of the first sorted vector, then range, then pretty()

TODO 01oct2023

• rainbowJam()

  • consider Hrange argument to rainbowJam() which would restrict colors to this range of hues. Unclear it if means the virtual hue, or output hue, but probably the virtual hue.

• R-shiny app changes:

  • in h1,h2 scatterplot, editing a point should not change its rank

    • it should be restricted on the x/y axis by the neighboring points
    • help keep points properly aligned, for example preventing identical points from being re-ordered.

  • add label to plotly shapes, using just the row number

  • consider double-clicking to add a point to the h1,h2 plot.

    • Point should be fixed between neighboring points on the x-axis, thus constraining the y-axis value.

  Could use plotly::showModal() to pop-up a confirmation, but later.

  • consider right-clicking to remove a point on the h1,h2 plot. Again, it could use showModal() to confirm, to prevent accidentally removing a point. (There is no convenient "Undo"... but there could be I guess.)

  • Consider "Undo" button

    • When no previous values, the "Undo" button is hidden.
    • Reverts to the previous h1,h2 values, then hides the "Undo" button.
    • Upon editing any value, store previous values, display "Undo" button.

• design2colors() (when ported here)

  • consider changing how class colors are assigned, to ensure each class uses phase values starting at the same place, in the same order. Goal is to have first color in each class using the same phase step.
  • When class colors are assigned, consider assigning group colors using hue splitting instead of lightness splitting. Generate a range of hues so that they do not conflict with other class hues. (Then use new argument Hrange.)

• Consider retiring h2hw(),hw2h(),h2hwOptions()

  • simplify system to use "Virtual Hue" and "Actual Hue"?

TODO 25sep2023

• consider adding launchColorjamShiny() to colorjam.shinyapps.io for testing.
• add unit testing with testthat, currently coverage is via @examples
• consider shinytest2 to test the R-shiny app. Probably not urgent.

TODO 20sep2023

• Consider secret test mode for rainbowJam() that uses HSL instead of HCL

  • likely to show less contrast than HCL, but with better saturation per hue
  • for testing, use C,L values rescaled to HSL range

• Move platjam::design2colors() here?

• Review preset and step

  • After real-world usage, decide whether default colors are visually distinct enough
  • R package pals provides functions to evaluate color palettes
  • Other packages: colorBlindness, colorblindr, shades, ggsci
  • Consider expanding Lrange, Crange to broaden visual range.

TODO 19sep2023

• DONE. sort_colors() default should sort by hue, chroma, luminance.

• DONE. Consider HCL-to-HSL hue mapping

• DONE. remap_colors()

  • convert colors to use a different preset

  • upon changing the hue

    • convert HCL hue to HSL hue
    • convert color internally to HSL to determine the current S,L values
    • adjust to the new hue, then apply the same S,L values
    • upon testing, HSL does not work nearly as well as hoped; HCL was too dark; blending HSL and HCL conversions seemed most effective

TODO 13sep2023

• DONE. Simple R-shiny app

  • visualize color palettes with adjustments

  • select the number of colors

  • edit the step values for Chroma and Luminance

  • edit the phase order of steps

  • choose the preset from known color wheels

  • bonus points: edit the color wheel h1, h2 points:

    • click-and-drag to modify existing points
    • add a new point
    • delete an existing point

• DONE. enhance "preset"

  • DONE. associate with a default_step
  • DONE. associate with direction: 1 or -1 for direction around the color wheel
  • add_colorjam_preset() should validate h1,h2 and direction.

• DONE. Fix approx_degrees() when multiple h1 or h2 values are repeated.

  • The approx() function requires unique x values.
  • Data should be sorted by x,y then uniqueness enforced for x.
  • This bug mainly affects h1,h2 when h2 is used for x.

TODO 07sep2023

• DONE. register a new color wheel

  • for example modifying an existing color wheel with adjust_hue_warp() then saving it as a named preset.
  • use igraph:::.igraph.shapes as a model, they store graph node shapes in that environment, then provide access igraph::shapes() and allow new shapes with igraph::add_shape().

• DONE. port color_complement() into this package.

• make_jam_divergent()

  • When argument linear2 is not supplied, get color_complement(linear1)
  • DONE. Error when using jam_linear and jam_divergent without package prefix when the colorjam package is not formally loaded.

TODO 23sep2022

• Migrate platjam::design2colors() into this package

  • Consider color_list related functions, since design2colors() produces a list of colors and color functions.
  • merge_color_lists() would combine two color lists into one, taking either the union of assigned colors, or the first color function, where relevant.

TODO 22may2022

• Migrate platjam::design2colors() here.

  • Intended to take data.frame with group, subgroup, and create a list of colors for each column, either with named character color vectors named by value, or color functions for numeric columns, as created by circlize::colorRamp2().

• No longer necessary: Implement HSL color space for categorical colors, slowly replacing HCL colorspace in rainbowJam().

TODO

These items are not likely to be on an active todo list:

• Add optional parameter to group2colors() which applies a systematic re-ordering of colors assigned to labels. For example, labels are typically assigned sequential color hue, which may not be ideal for all cases. This method would assign colors to every n-th color.
• Add subgroup2colors() which extends group2colors() by splitting each color into subgroups as needed. In this way, experiment factors can each be assigned colors, while the factor levels are assigned values from a gradient for each color. In future, this function will optionally use a color family (or small range of color hues) to each factor, then will vary the hue, as well as the chroma/luminance for factor levels.
• vals2colorLevels() assigns colors to a numeric range, keeping zero as a mid-point for divergent color scales. Notably allows a "color lens" which applies a scalable adjustment to the color ramp to enhance or reduce the color intensity across the numeric range.

color table viewer (ASCII)

• Perhaps the most frequently used JAM function is (was) a fixed-width colorized table viewer for console output.

  • Similar in theory to the custom tibble print method, this function goes further by colorizing values in each column to help the visual summary.

  • Especially useful for checking experiment design, typos in labels, and numeric ranges. It can, for example, make a text heatmap for numeric values, which can be helpful spotting outliers.

  • It also trims column values so the table fits within one or more screen widths.

  • Similar R package pillar offers colorized formatted data.frame/tibble output.

    • Unclear whether it can provide equivalent capabilities.