compiler:pp -s and compiler:pp -v dump should have colors

[Hywan]

@jubianchi did a good POC with it. It is a great idea. We should do that. Maybe @jubianchi could explain more.

---

Want to back this issue? Post a bounty on it! We accept bounties via Bountysource.

[jubianchi]

let me show an example :

This is a custom colorization I made manually in one of my slides.

What should be done:

• Choose a color for namespace names
• Choose a color for token names
• Choose a color for token values
• Choose a color for rules
• Use those color in the token sequence (-s) output
• Use those color in the dump visitor (-v dump) output
• Use those color in the trace (-t) output

[Hywan]

Why not using automatic color for the namespace and for tokens?

[jubianchi]

@Hywan what do you mean by "automatic color" ?

[Hywan]

@jubianchi Automatic color selection, in a palette. Each token or namespace without a color has a color attached automatically, and it remains attached. This way, this is easy to track the token by its color instead of its name.

[Hywan]

@jubianchi For instance, regarding your example, T_NUMBER will be blue, T_OP_PLUS will be orange, T_VAR will be red, but T_OP_EQUAL will not be orange but pink instead. All token value will have the color associated to its token name. Same for namespace.

[jubianchi]

@Hywan good idea! let me try something when I get some time ;)

[1e1]

Moreover, I would like something like highlight_code for a beautiful display ;)

[jubianchi]

I experimented with some algorithms to colorize the output and it seems hard to do it automatically:

• we have to generate random color with reasonable distance, or
• we have to hardcode a color palette which is large enough to deal with with huge grammars, but
• we would have to take care of user's terminal background (color cannot be the same on light and dark backgrounds)

If we choose to generate random color on-the-fly, we will have to add too much code (IMHO) to compute random colors, deal with distances between them, etc.

If we choose to hardcode a palette, we would have to hardcode two palettes: one for light backgrounds and one for dark ones. Or we have to find a palette working with both backgrounds.

[jubianchi]

example palette (80 colors):

[1e1]

Can we specify in an other file, which color to join for each token.
Perhaps a %filename%.pp-lut relative to %filename%.pp

[1e1]

Or shall we think about a "check style" tool?
An extra file setup the coding style and the token style (text color, line break, spacing, etc)
Perhaps there is the same thought.

[Hywan]

@1e1 I don't understand what you're talking about. We are talking about colorizing the output of some compiler:pp command, not to colorize a PP grammar.

@jubianchi Let me ask some colleague designers. Having a pre-defined palette sounds good for me. Else we would have to start a Hoa\Color library and we don't have time for this right now. Your proposed palette seems excellent. What is the source?

[jubianchi]

@Hywan I randomly generateed 256 colors and manually removed those not fitting in both backgrounds

[1e1]

@Hywan
I thought it is to highlight a rule written from a PP file.

[Hywan]

@jubianchi Good work.

[Hywan]

@ghn points me this https://github.com/mbostock/d3/wiki/Ordinal-Scales#categorical-colors. These palettes have not been designed to work on both light and black backgrounds, but we can maybe find some inspiration there.

[jubianchi]

Example output with the 80 colors palette:

For each new token we shift a color from the palette and use it.

[Hywan]

How do you choose colors in your palette? Are colors already ordered by distance (in which space?)? Do you choose colors by the maximum distance each time?

[jubianchi]

How do you choose colors in your palette?

For each new token we shift a color from the palette and use it.

Are colors already ordered by distance (in which space?)?

No.

Do you choose colors by the maximum distance each time?

No.

[jubianchi]

And here is an example trace output:

[Hywan]

Sounds good. Colors are a little bit too “agressive”, thoughts?

[jubianchi]

@Hywan yes the colors are not optimal :/

[Hywan]

But this is an excellent start! I am going to ask for a “““palette flatter”””

[Hywan]

@ghn provides me this other resource: http://kentor.me/posts/generating-pastel-colors-for-css/. @jubianchi, it can help you a lot I guess.

[Hywan]

And so this: https://randomcolor.llllll.li/ (a PHP port exists, for inspiration).

[jubianchi]

@Hywan I searched yesterday for libraries doing good colors randomization and I found https://randomcolor.llllll.li/ was really promising but to do things well, we should port it.

[Hywan]

@jubianchi Isn't it possible to develop a very minimal algorithm?

[flip111]

@jubianchi do you want to post your code in a github repository somewhere? Maybe some people can take it as a starting point and continue your work if you don't longer want it.