Chart iteration

[braver]

Incorporates #105 and thus also fixes #102, as well as #101.

• increase vertical size of the chart
• move axes to the top so the grid etc don't cut into them
• improve contrast
• change the hovers (and make them work in Safari etc)
• float the chart next to the package metadata if there is space
• add a legend

The hovers now highlight the week you're on, which matches the title that's displayed. Hovering the right axis highlights the upgrades line. Hovering the legend highlights the matching chart elements.

I must say that I find the upgrade stats a bit hard to believe. A package that has been upgraded halve a year ago that sees on average 70 upgrades per week? As I understand it that means every week 70 users start ST for the first time since May, or have auto_upgrade off and manually request upgrades (who does that?)? Even at > 100k total installs that number feels high to me. 🤷🏻‍♂️ Not saying that it's wrong or broken, but it's not what I'd expect.

[michaelblyons]

I want pretty strongly for the "X installations" legend to be changed. I think it is misleading right now.

• Normally "net installs" vs "total installs" is ambiguous (at least in English) when you just say "installs,"
• But with "removals" right below, I feel it is more likely interpreted as "total installs," which is not the portion of the chart that is that color.

If the total (all-time) installations is also a net value, you can add "net" and it will be decent:

• 140,780 net installations (average 34.2 per week)

But that kind of implies that the average is since package creation. You could instead say

• 140,780 net installations (recent average 34.2 per week)

which is more correct, but I think a better play is just

• Net installations (average 34.2 per week)

since the total number is already in the top-left of the page.

[braver]

Good points! We do actually use net installs everywhere and simply call it “installations”. I was thinking we could perhaps do that here as well, and call what’s now labelled “removals” “installations before removals”? Because “net installs” probably doesn’t mean anything to you unless you’ve been part of these discussions, i.e. even that needs explaining. The net installs number is indeed also elsewhere but you don’t know those were “net” instead of simple all installs because it’s never explicitly specified yet.

[michaelblyons]

Just to be explicit: I like the iteration a lot overall! Any criticism is meant to be constructive.

We do actually use net installs everywhere and simply call it “installations”.

Let's note that in the FAQ. I had always assumed these were gross numbers, including on pc.io.

call what’s now labelled “removals” “installations before removals”

There's probably some phrasing that communicates the right thing. I'm not convinced you have it, yet, but I will still have to think about alternatives.

[braver]

I'm not convinced you have it, yet, but I will still have to think about alternatives.

Same 😊

[braver]

There's not really a best answer here

No, there doesn't seem to be an easy decision here. Maybe move that to an issue or discussion if you want to dig in, I'd like to keep that out of scope of this iteration.

I did 2 things here.

The first move is to revert back to using the gross install count as what we display primarily. It's a higher number, which is never bad 😎 More importantly, we don't have removals from before our new stats, meaning the base install count for most packages that have existed before August is a gross number. In other words, dropping that number by some recent removals is slightly weird, inconsistent. Furthermore, as we've been discussing, I'm not sure I fully trust that removal number to mean what we think it does. This also fixed an untruth if you will in the UI where it previously said "installed X times": it was actually installed more than X, it was just also removed a few times ... but that's not what that says.

Then, in the chart and elsewhere things become easier to explain. We don't have to say all the time "it was installed X, but actually Y etc". No, it was simply installed X and in the chart you can see the net install being lower and you get an idea about how much, but the exact numbers aren't super relevant and we can't really talk about "total net installs" for packages from 2012 anyway.

[kaste]

Love it, absolutely. The look of the legend is great too, but the info...

"recent average" is iirc 53 weeks. Maybe just "Installation: 53 per week, 120.000 in total". We can still switch to recent 4 weeks window which is maybe what we wanted. That's how I remember it.

Do I still have the "ceil" filter checked in? Then: avg_inst | ceil | grouping to not have "74.25 installs per week".

"in total" is still duplicated, esp. when the chart jumps to the left.

gross installs "it's the more "real" number". How I remember it: we wanted to remove the old stats and only keep the "real" numbers (ours, the more recent ones, ditching the ST glory decade). 1M installs without removals, well these just deleted ST or never installed it on their new computer.

The average lines: I think they should be in two colors. Usually they overlap with one portion of the bars:

Obviously the upper average goes through the upper bars, and the lower one goes through the darker orange one. That's the reason I had two colors. (Without any tuning, too low contrast... of course)

The left grey border... I don't see the grid, I don't know where that aligns to. I actually think it floats okay if you don't draw that border.

... as long as it sticks on the right edge. What grid are you seeing here actually.

Actually I love the look.

[kaste]

Really, I would just drop the left border and ceil the average week and ship it.

Then: remove the original downloads counter and move the stars counter to a new place?

[braver]

I couldn’t actually see the two different colors for the averages, the lines are too thin, but will have a look to see what can be done.

Then: remove the original downloads counter and move the stars counter to a new place?

There’s some other potential info in that space too (“unmaintained” etc) so I couldn’t figure out what to do. I don’t really mind repeating the install number, one is just the number with an icon, the other is an explanation.

we wanted to remove the old stats and only keep the "real" numbers

That, or put a greater emphasis on the average rate of installation because that’s more truly “current popularity”…. We should perhaps be able to sort on that rate eventually. Anyway, a topic to revisit some more!

[kaste]

Naturally: "72 per week, 7890 in the last year" (the totals of the chart we are showing) -- if we just had the numbers.

the lines are too thin

They were only visible on hover (but also in my 100 pics, you 🙈) which apparently only worked on Chrome.

[braver]

They were only visible on hover

Ah... 🧐

if we just had the numbers.

Yeah, indeed! Only like 48 more weeks until we've got everything :)

[michaelblyons]

I agree with Kaste on the average lines being two colors. FWIW, the hover effects worked for me on FF except that I didn't have the little spacer between the net and gross installs.

I also agree on dropping the left border, but I don't feel strongly.

"recent average" is iirc 53 weeks. Maybe just "Installation: 53 per week, 120.000 in total". We can still switch to recent 4 weeks window which is maybe what we wanted. That's how I remember it.

Do I still have the "ceil" filter checked in? Then: avg_inst | ceil | grouping to not have "74.25 installs per week".

I think we'll end up with half the packages at "0 per week," but that's fine for now. 😛

[braver]

I also agree on dropping the left border

Without it the chart is floating in nowhere land. There isn't a strong right edge to the page, and neither is there on the chart, so that doesn't hold it in place. (Edit: the border also keeps the chart and the legend together). Hence the border.