Proposal: new exponentformat which uses "milli" prefix

[ignamv]

Can someone please look over my pull request to see if it's suitable for the main repo?

ignamv#1

[archmoj]

Looking good to me.

[nicolaskruchten]

I see in the constant array SIPREFIXES that k and m are there already... it just seems like a bug to me that exponentformat="SI" doesn't use these.

[nicolaskruchten]

PS @ignamv thanks for the PR! I'm just wondering if we can't just fix this in SI instead of introducing a new mode :)

[alexcjohnson]

The existing SI mode is working as intended. Not using the m prefix - but using all the others - is pretty typical on log axes because the m is too often confusing (vs M) and doesn't shorten the labels much (100m is longer than 0.1). For k it's a little more subtle, we don't use it for 1k but we do for 10k. This I don't feel that strongly about, we could consider it a bug, but I do think there's a logic to the current behavior: the longest pure-power labels you have otherwise are 100k and 100M etc, which are exactly as many characters as 1000, so we defer to readability and use unmodified numbers for just one more power of 10.

All that is to say I think the current behavior makes sense as a default for SI. But I can see wanting to change that and always use a prefix - or, for that matter, keeping the numbers unmodified longer (or using prefixes more extreme than f and T, beyond which we currently switch to power notation - because I didn't deem those prefixes sufficiently well known when this feature was originally written - but that probably would need to be a totally separate attribute).

So what if we make a new attribute to substitute in for the 3 here:

plotly.js/src/plots/cartesian/axes.js

Line 1045 in 2737bf2

if(Math.abs(rangeexp) > 3) {

something like ax.minexponent? For your use case @ignamv you'd set it to 0, but someone who doesn't even want to use k or µ might set it to 6, or something between even.

[ignamv]

So what if we make a new attribute to substitute in for the 3 here

Sounds cleaner! I'll give it a go.

[ignamv]

I've added the new attribute and removed the eng exponentformat. I'm not familiar with this codebase so I based the new code on what I found for the exponentformat attribute.

ignamv#1 (comment)

[alexcjohnson]

@ignamv looks good! Feel free to make a PR to the main repo. It'll need a test or two before it's ready to merge, probably one setting minexponent to zero (the case you care most about) and another setting it large like in ignamv#1 (comment)

This is a good model to start with:

plotly.js/test/jasmine/tests/axes_test.js

Lines 2811 to 2833 in c437cdf

	var textOut = mockCalc({
	type: 'log',
	tickmode: 'linear',
	exponentformat: 'B',
	showexponent: 'all',
	tick0: 0,
	dtick: 1,
	range: [-18.5, 18.5]
	});
	expect(textOut).toEqual([
	'10<sup>\u221218</sup>',
	'10<sup>\u221217</sup>',
	'10<sup>\u221216</sup>',
	'1f', '10f', '100f', '1p', '10p', '100p', '1n', '10n', '100n',
	'1μ', '10μ', '100μ', '0.001', '0.01', '0.1', '1', '10', '100',
	'1000', '10k', '100k', '1M', '10M', '100M', '1B', '10B', '100B',
	'1T', '10T', '100T',
	'10<sup>15</sup>',
	'10<sup>16</sup>',
	'10<sup>17</sup>',
	'10<sup>18</sup>'
	]);

[ignamv]

Thanks! I've added both tests and they seem to work.

[archmoj]

@ignamv Great news! Please move forward and submit a PR.