Relocate first tick when falls inside a rangebreak

[archmoj]

Fixes#4722,
this PR improves tick positioning by moving the first tick when it falls inside a rangebreak.

Demos: before vs after.

@plotly/plotly_js

[nicolaskruchten]

So here the ticks are all at 16:00 which is weird

[alexcjohnson]

What's going on in test/image/baselines/axes_breaks-gridlines.png ? The ticks seem to have moved to the last day of the month, which means "Mar 2015" is actually at March 31, ie essentially where an April 1 tick should have been, so it looks like we're labeling April as March.

This seems consistent on the autorange view, but if you zoom in a bit to 1-month ticks things get worse: some ticks seem to be on the last day of the month, others on the first, so sometimes you get two identical ticks:

I guess I'm nervous about doing this at the level of tick0 in general - particularly if you have some funny non-periodic breaks, shifting tick0 could result in other ticks that don't need to move at all getting moved. Seems to me the solution needs to be about moving each individual tick out of a break if it falls within one.

This PR also still has the odd spacing issue for automatically-determined tick0 and dtick values. See test/image/baselines/axes_breaks-weekends-weeknights.png - to solve that we'll likely need to adjust this algorithm when there are hourly ticks > 1 h along with hour rangebreaks.

[archmoj]

Fixes#4722,
this PR improves tick positioning by moving the first tick when it falls inside a rangebreak.

Demos: before vs after.

@plotly/plotly_js

[nicolaskruchten]

So here the ticks are all at 16:00 which is weird

[alexcjohnson]

What's going on in test/image/baselines/axes_breaks-gridlines.png ? The ticks seem to have moved to the last day of the month, which means "Mar 2015" is actually at March 31, ie essentially where an April 1 tick should have been, so it looks like we're labeling April as March.

This seems consistent on the autorange view, but if you zoom in a bit to 1-month ticks things get worse: some ticks seem to be on the last day of the month, others on the first, so sometimes you get two identical ticks:

I guess I'm nervous about doing this at the level of tick0 in general - particularly if you have some funny non-periodic breaks, shifting tick0 could result in other ticks that don't need to move at all getting moved. Seems to me the solution needs to be about moving each individual tick out of a break if it falls within one.

This PR also still has the odd spacing issue for automatically-determined tick0 and dtick values. See test/image/baselines/axes_breaks-weekends-weeknights.png - to solve that we'll likely need to adjust this algorithm when there are hourly ticks > 1 h along with hour rangebreaks.

[archmoj]

Fixes#4722,
this PR improves tick positioning by moving the first tick when it falls inside a rangebreak.

Demos: before vs after.

@plotly/plotly_js

[nicolaskruchten]

So here the ticks are all at 16:00 which is weird

[alexcjohnson]

What's going on in test/image/baselines/axes_breaks-gridlines.png ? The ticks seem to have moved to the last day of the month, which means "Mar 2015" is actually at March 31, ie essentially where an April 1 tick should have been, so it looks like we're labeling April as March.

This seems consistent on the autorange view, but if you zoom in a bit to 1-month ticks things get worse: some ticks seem to be on the last day of the month, others on the first, so sometimes you get two identical ticks:

I guess I'm nervous about doing this at the level of tick0 in general - particularly if you have some funny non-periodic breaks, shifting tick0 could result in other ticks that don't need to move at all getting moved. Seems to me the solution needs to be about moving each individual tick out of a break if it falls within one.

This PR also still has the odd spacing issue for automatically-determined tick0 and dtick values. See test/image/baselines/axes_breaks-weekends-weeknights.png - to solve that we'll likely need to adjust this algorithm when there are hourly ticks > 1 h along with hour rangebreaks.

[archmoj]

Fixes#4722,
this PR improves tick positioning by moving the first tick when it falls inside a rangebreak.

Demos: before vs after.

@plotly/plotly_js

[nicolaskruchten]

So here the ticks are all at 16:00 which is weird

[alexcjohnson]

What's going on in test/image/baselines/axes_breaks-gridlines.png ? The ticks seem to have moved to the last day of the month, which means "Mar 2015" is actually at March 31, ie essentially where an April 1 tick should have been, so it looks like we're labeling April as March.

This seems consistent on the autorange view, but if you zoom in a bit to 1-month ticks things get worse: some ticks seem to be on the last day of the month, others on the first, so sometimes you get two identical ticks:

I guess I'm nervous about doing this at the level of tick0 in general - particularly if you have some funny non-periodic breaks, shifting tick0 could result in other ticks that don't need to move at all getting moved. Seems to me the solution needs to be about moving each individual tick out of a break if it falls within one.

This PR also still has the odd spacing issue for automatically-determined tick0 and dtick values. See test/image/baselines/axes_breaks-weekends-weeknights.png - to solve that we'll likely need to adjust this algorithm when there are hourly ticks > 1 h along with hour rangebreaks.

[archmoj]

Fixes#4722,
this PR improves tick positioning by moving the first tick when it falls inside a rangebreak.

Demos: before vs after.

@plotly/plotly_js

[nicolaskruchten]

So here the ticks are all at 16:00 which is weird

[alexcjohnson]

What's going on in test/image/baselines/axes_breaks-gridlines.png ? The ticks seem to have moved to the last day of the month, which means "Mar 2015" is actually at March 31, ie essentially where an April 1 tick should have been, so it looks like we're labeling April as March.

This seems consistent on the autorange view, but if you zoom in a bit to 1-month ticks things get worse: some ticks seem to be on the last day of the month, others on the first, so sometimes you get two identical ticks:

I guess I'm nervous about doing this at the level of tick0 in general - particularly if you have some funny non-periodic breaks, shifting tick0 could result in other ticks that don't need to move at all getting moved. Seems to me the solution needs to be about moving each individual tick out of a break if it falls within one.

This PR also still has the odd spacing issue for automatically-determined tick0 and dtick values. See test/image/baselines/axes_breaks-weekends-weeknights.png - to solve that we'll likely need to adjust this algorithm when there are hourly ticks > 1 h along with hour rangebreaks.

[archmoj]

Fixes#4722,
this PR improves tick positioning by moving the first tick when it falls inside a rangebreak.

Demos: before vs after.

@plotly/plotly_js

[nicolaskruchten]

So here the ticks are all at 16:00 which is weird

[alexcjohnson]

What's going on in test/image/baselines/axes_breaks-gridlines.png ? The ticks seem to have moved to the last day of the month, which means "Mar 2015" is actually at March 31, ie essentially where an April 1 tick should have been, so it looks like we're labeling April as March.

This seems consistent on the autorange view, but if you zoom in a bit to 1-month ticks things get worse: some ticks seem to be on the last day of the month, others on the first, so sometimes you get two identical ticks:

I guess I'm nervous about doing this at the level of tick0 in general - particularly if you have some funny non-periodic breaks, shifting tick0 could result in other ticks that don't need to move at all getting moved. Seems to me the solution needs to be about moving each individual tick out of a break if it falls within one.

This PR also still has the odd spacing issue for automatically-determined tick0 and dtick values. See test/image/baselines/axes_breaks-weekends-weeknights.png - to solve that we'll likely need to adjust this algorithm when there are hourly ticks > 1 h along with hour rangebreaks.

[archmoj]

Fixes#4722,
this PR improves tick positioning by moving the first tick when it falls inside a rangebreak.

Demos: before vs after.

@plotly/plotly_js

[nicolaskruchten]

So here the ticks are all at 16:00 which is weird

[alexcjohnson]

What's going on in test/image/baselines/axes_breaks-gridlines.png ? The ticks seem to have moved to the last day of the month, which means "Mar 2015" is actually at March 31, ie essentially where an April 1 tick should have been, so it looks like we're labeling April as March.

This seems consistent on the autorange view, but if you zoom in a bit to 1-month ticks things get worse: some ticks seem to be on the last day of the month, others on the first, so sometimes you get two identical ticks:

I guess I'm nervous about doing this at the level of tick0 in general - particularly if you have some funny non-periodic breaks, shifting tick0 could result in other ticks that don't need to move at all getting moved. Seems to me the solution needs to be about moving each individual tick out of a break if it falls within one.

This PR also still has the odd spacing issue for automatically-determined tick0 and dtick values. See test/image/baselines/axes_breaks-weekends-weeknights.png - to solve that we'll likely need to adjust this algorithm when there are hourly ticks > 1 h along with hour rangebreaks.

[archmoj]

Fixes#4722,
this PR improves tick positioning by moving the first tick when it falls inside a rangebreak.

Demos: before vs after.

@plotly/plotly_js

[nicolaskruchten]

So here the ticks are all at 16:00 which is weird

[alexcjohnson]

What's going on in test/image/baselines/axes_breaks-gridlines.png ? The ticks seem to have moved to the last day of the month, which means "Mar 2015" is actually at March 31, ie essentially where an April 1 tick should have been, so it looks like we're labeling April as March.

This seems consistent on the autorange view, but if you zoom in a bit to 1-month ticks things get worse: some ticks seem to be on the last day of the month, others on the first, so sometimes you get two identical ticks:

I guess I'm nervous about doing this at the level of tick0 in general - particularly if you have some funny non-periodic breaks, shifting tick0 could result in other ticks that don't need to move at all getting moved. Seems to me the solution needs to be about moving each individual tick out of a break if it falls within one.

This PR also still has the odd spacing issue for automatically-determined tick0 and dtick values. See test/image/baselines/axes_breaks-weekends-weeknights.png - to solve that we'll likely need to adjust this algorithm when there are hourly ticks > 1 h along with hour rangebreaks.