include tooltip in v1 defintion

[tomwayson]

It was part of the v0.x definition and I think we want this to be abstracted rather than relying on overrides.

I think for v1 we'll want this at the series level rather than the root.

[tomwayson]

FYI https://github.com/Esri/cedar/blob/v0.x/src/cedar.js#L74-L77

[tomwayson]

Seems like in v0 you can't turn off tooltips!

Go to http://jsfiddle.net/tomwayson/dox5o6w4/ and try setting "tooltip": false, on the definition, or passing tooltip: false to .show()

That's kinda awful, and I don't think we should continue that in v1x.

[tomwayson]

Also, I'm not keen on the name tooltip, this isn't some confusingly named text input - it's data, maybe popup?

[tomwayson]

FYI here are the current tooltip settings.

Serial/pie charts:

cedar/packages/cedar-amcharts/src/render/render.ts

Line 99 in 99e774f

graph.balloonText = `${graph.title} [[${spec.categoryField}]]: <b>[[${graph.valueField}]]</b>`

x/y charts:

cedar/packages/cedar-amcharts/src/render/render.ts

Lines 117 to 126 in 99e774f

	graph.balloonText = `<div>${series.category.label}: [[${series.category.field}]]</div><div>${series.value.label}: [[${series.value.field}]]</div>`
	// bubble
	if (spec.type === 'xy' && series.size) {
	graph.valueField = series.size.field
	graph.balloonText = `${graph.balloonText}<div>${series.size.label}: [[${graph.valueField}]]</div>`
	} else {
	delete graph.valueField
	}
	}

[tomwayson]

After a discussion w/ the Hub team we're going to stick w/ the name "tooltip"

[tomwayson]

We should use .visible to control whether or not to show tooltips for consistency with legend (#394), and default to true.

[tomwayson]

The crux of this issue is figuring out how to do token replacement in the string templates.

v0 allows for injecting attribute values using {} as the delimiter: Ex:

{Number_of_SUM} Students in {Type}

However, in v1, b/c there can be multiple series, so you could not specify a single top level tooltip that used hardcoded field names and labels like that. Ideally we'd let consumer's inject values from the series. For example, if you were to try and express the current default tooltip using template literals it would effectively be:

${series.value.label} ${data[series.category.field]}: <b>${data[series.value.field]}</b>

However v1's joined datasets complicate this. For a joined dataset it's effectively:

${series.value.label} ${data.categoryField}: <b>${data[series.value.field + '_' + datasetIndex]}</b>

I'm not even sure we could support adlib expressions that we could capture all that nested nastiness.

An alternative is that we can start w/ tooltip being defined at the series level instead of at the definition.tooltip level. Then consumers could safely use hardcoded field names and labels in the templates. The problem is that if you wanted to turn tooltips off (visible = false), you'd have to do it for each series. Long term we'd want to support tooltip at both at the top level and at the series level, and merge them w/ Object.assign() when generating the AmCharts tooltip. We could do that now and then later add the ability to inject series properties so that you could better handle most cases w/ a single top-level tooltip.

[tomwayson]

#411 helps w/ this a bit, b/c the format for joined field names is now: ${series.source}_${series.value.field} (again using template literals), so at least everything we'd need to do token replacement is available either on data (row) or the series itself. So the joined template would now be:

${series.value.label} ${data.categoryField}: <b>${data[series.source + '_' + series.value.field]}</b>