Allow hiding nil group entries or allow simple bar chart color assignment

[forddavis]

Description

I would really like an option to hide nil bars. Perhaps this is an unconventional use case, but I am trying to show intensity of a variable over time on a bar chart.

I could also achieve this if it were possible to style individual bars and fully customize the legend, however it doesn't appear to be possible to assign colors to bars within a single dataset either.

Steps to Reproduce

data = [
  [0.5, "00:00", "#07A6FF", "nominal"],
  [0.64, "00:20", "#FFA500", "marginal"],
  [0.75, "00:40", "#FFA500", "marginal"],
  [1.0, "01:00", "#FF0000", "failed"]
]

intervals = data.map(&:second)
intensities = data.map(&:last).uniq

g = Gruff::Bar.new
colors = ["#07A6FF","#FFA500", "#FF0000"]

g.theme = {
  colors: colors,
  marker_color: '#dddddd',
  font_color: 'black',
  background_colors: 'white'
}

g.labels = intervals

intensities.each do |intensity|
   g.data(intensity, data.map { |i| i.first if i.last == intensity })
end

This results in wonky spacing because it draws the nil datapoints for each set. But I only want to draw one set per label.

Additionally there is no way to assign a color to each bar without assigning separate datasets.

Would it be possible to add support for hiding empty bars and their spacing?

Apologies if I'm missing something. I thought for sure it could be done with the library in it's current state since it is very capable, but I've spent a good amount of time tinkering with this and can't figure it out with existing methods.

System Configuration

• ImageMagick version: ImageMagick 7.1.1-8 Q16-HDRI aarch64 21129
• RMagick version: 5.2.0
• Gruff version: 0.22.0
• Ruby version: 2.7.2
• Environment (Operating system, version and so on): Mac Ventura 13.3.1

[Watson1978]

Unfortunately, it does not have such the features...
If we support it, we may need to create a new class instead of Gruff::Bar...

[forddavis]

Thank you for the response @Watson1978. Is there any interest in this feature? I could potentially make a PR.

It seems like an option g.collapse_empty_columns defaulting to false would be potentially useful, although I could see a separate class being warranted if you expect that to be a major deviation from current behavior.

[Watson1978]

The internal that it is not distinction is made between nil and 0.
It wouldn't handle properly when you really need 0, currently implementation...