Support for period: auto, automatic statistic period

[dbuezas]

What's Changed

• Automatic statistic period support by @dbuezas in Automatic statistic period support #94

Full Changelog: v1.4.8...v1.5.0

---

This discussion was created from the release Support for `period: auto`, automatic statistic period.

[dbuezas]

Idea by @Zingel1986

[dbuezas]

but there's little that can go this wrong

"It was at this moment he knew he f'd up"

@r-jean-pierre I found the bug and it is totally my fault. I'm working on the fix

[dbuezas]

v1.6.3 fixes it.

[dbuezas]

In case you wonder what went wrong:
https://github.com/dbuezas/lovelace-plotly-graph-card/pull/121/files

[r-jean-pierre]

Glad you find what was working wrong.
So, I changed nothing in my just yaml, updated plotly-graph to 1.6.3, and even rebooted the machine, but it's still weird.
But: maybe I need to wait several minutes to have some fresh new data (?), my sensors are quite lazy

This what I just get for now:

it seem that now the Humidity(max) in orange took the latest value of the Temperature(mean)

Hope i'm not raising a false alarm...

[dbuezas]

Uhm. Can you check tomorrow and let me know if this is indeed not just your sensor acting up? :)
I was able to reproduce the previous bug you reported and then fix it, so this is quite unexpected to me.

Thanks for following up on this!

[netweaver1970]

I don't know if it's related or not, but I have sometimes my time axis going totally crazy, when changing time ranges using the buttons, a bit of scrolling left and right (and also periods changing as part of the mapping), ending up in all kind of years, e.g. the year 2130, sometimes even further in the future.

Also, I have sometimes disappearing data series (they disappear also from the legend). But I can't tell how to provoke the issue. That's why I didn't make an issue yet. Both issues don't always happen.
But when doing a manual refresh, all data series 'come back' and the time range is back in 2022, as expected.
Pretty vague, I know. But it's happening often enough to make me write about it.

[dbuezas]

If you catch it, please share a screenshot :)

[r-jean-pierre]

So, after 24H I start to have some tiny insights. I vote for a combination of "last value" + different y-axis (forced or automatically found by the card).

In a few words, I have exactly the same end-of-line behaviour, when it seems that one sensor is affected to the wrong trace. What was wrong yesterday is "correct" in the today graph, but today my latest values are creating unexpected end of curves for Nov 7th.

I'm gonna test and start from scratch with another type of sensor (previously xiaomi, and now with tahoma heater integration):

So far so good after 24h of observation:

I will introduce the advanced features one after one to recreate the more complex graphs and see if I can deduce something

[dbuezas]

This should be fixed right in the latest release of a couple of minutes ago.
The reason was this oversight:

here

This logic cleans up the data out of what is visible, but keeps the points right out of the visible range. The algorithm goes through all the points and keeps track of the last one out of range from the left and from the right.
My mistake was where I declared the variable, which meant that in the specific situation of one entity not having a cached value out of visible range would "inherit" the datapoint of the last entity.

Thanks for all the screenshots and detailed reports! ❤️

Let me know if it works as expected now

[r-jean-pierre]

Hey,

I think it's solved indeed!

[Balou067]

Hi I’m not sure if, with the ‘period” feature, also the start of the x-axis of the values displayed should be set/fixed? So like, if one hour is displayed, the x-axis starts at the start of this hour, by a day being displayed, the start of the x-axis is at midnight of this day, for a week being displayed, the x-axis starts at midnight of the first day of the week and so on. Or, if not intended to be part of the ‘period’ feature, how can that be done?

[dbuezas]

Not part of period. Feel free to post an issue with the feature request, it will be picked it sooner or later :)