https://kotlinlang.org logo
Join Slack
Powered by
MavenCentral stats for OpenSavvy :eyes:
# opensavvy
c
MavenCentral stats for OpenSavvy 👀
o
Similar with testballoon. I wouldn’t bet on the accuracy of those numbers.
c
Honestly, I find them very strange.
Yeah no, I'm not buying it. There's a chance I have a package that's successful without me knowing it, but it's not either of these two.
@mbonnin You have been tracking these kinds of metrics for a long time in your projects, do the numbers displayed in Scarf for your own packages seem right to you?
m
IIRC I have a x2/x3 error factor (it's higher now). I have a call booked with them to try to understand
I'm not looking at the "companies" though. This chart is very weird
o
I was missing precise download counts per version. I could extract it via their chatbot using this prompt:
Copy code
Show the download counts per version for testballoon-gradle-plugin. Show precise counts. Show all counts and nothing else. Do not provide interpretations, show just the numbers.
👀 1
m
Wow so now we need AI to remove all the added "interpretations" 😅
Makes you wonder what prevents from showing the raw numbers 😅
I'm afk right now but IIRC there is a tab with per coordinate "package events" (whatever that is)
That seemed a bit more raw than the "companies" thing
o
I could not find anything. Downloads are unavailable for "trial" users. I needed prompt guessing to actually make the "AI" provide basic data. Otherwise it would jabber about "dominance", "Immediate adoption", "Sustained growth".
🫠 1
And on the "packages" tab all I get in addition is badges.
I suspect that these AI interactions will be gone as soon as the trial ends. And then there will be useless (and questionable) aggregate numbers.
😒 1
c
> IIRC there is a tab with per coordinate "package events" yeah but it only shows 9 packages, and for an org that does a lot of Multiplatform, that's basically meaningless lol
I very much doubt it's displaying the top packages, there is no legitimate reason for anyone to download 
dev.opensavvy.ktmongo:annotations
but not 
dev.opensavvy.ktmongo:dsl
, they definitely should have the same numbers
m
You can find a filter somewhere near the top
My filter contains 700+ entries
c
LOL the AI ahah
Total Downloads by Project (Aug 26 - Nov 24, 2025):
• dev.opensavvy.compose.lazy/lazy-layouts: 0
• dev.opensavvy.compose.lazy/lazy-layouts-js: 0
• dev.opensavvy.decouple/components: 0
• dev.opensavvy.decouple/components-iosarm64: 0
• dev.opensavvy.decouple/components-iossimulatorarm64: 0
• dev.opensavvy.decouple/components-iosx64: 0
• dev.opensavvy.decouple/components-js: 0
• dev.opensavvy.decouple/components-jvm: 0
• dev.opensavvy.decouple/components-linuxx64: 0
• dev.opensavvy.decouple/design-headless: 0
Total Events: 0
Total Downloads: 0
Total Views: 0
Total Unique Downloads: 0
Total Unique Events: 0
Total Unique Views: 0
🤣 1
(that's using Oliver's prompt)
ah right
Right, the "Packages" tab looks much more sensible.
plus one 1
o
My results were:
Copy code
0.3.3-K2.1.21: 15,991
0.3.3-K2.2.0: 294,821
0.5.0-K2.2.0: 43,985
0.6.0-K2.0.0: 18,686
0.6.0-K2.1.0: 18,027
0.6.0-K2.1.20: 17,464
0.6.0-K2.2.0: 32,419
0.6.0-K2.2.20-RC: 21,867
0.7.1-K2.2.0: 23,956
0.7.1-K2.2.21: 47,879
But I'm also seeing ghost data on GH traffic:
So I guess unless we do our own analytics, we could also ask the next crystal ball.
c
I think I'd much prefer have a way to download my own Central data in Kibana or whatever, yeah
same 2
Ok, so Scarf doesn't have any KtMongo modules other than the 
annotations-<platform>
ones
😅 1
that's weird though, the total count is the same as in Central
o
OK, so stop developing the other parts. Annotations is what the world desires!
🤣 1
c
oh no it does, but they don't appear in the paging nor the search?
tbf, the annotations package is by far the easiest to maintain, so I don't think I would even mind xD
✔️ 1
I wonder if Oracle has a massive internal MavenCentral mirror and that's why they're showing up so much
oh if you select a package, there's this graph at the bottom
💯 1
o
Yep, just discovered that one.
m
We have a chain of central Europe pharmacies as the 2nd entry. Definitely some weirdness there
c
Wait this is the date of my KTConf Belgium talk so there is actually a chance that this measures real people 🤔
kodee excited 1
I mean, I get 2.5k/month downloads for Prepared, but that's a number that could very well be caused by my own CI. However, I don't have any big project that consumes KtMongo, so that's a more interesting graph
I swear I'm going to add a button "I'm using this version!" to the release list on the doc website, and make my own graph from that
(btw I'm currently rolling out Plausible on my sites as a GDPR-compliant Google Analytics alternative, and it's great)
o
A doc site button probably won't provide much insight. People use auto-updating these days. I guess even GitHub stars doesn't carry the weight it once did, because folks don't need it anymore for release reminders.
c
True, it won't be precise. But in my case, the question is "are there people using it _at all_"
That's the Vite plugin, that one I know there are users since I get issues reported
o
testballoon Gradle plugin:
If I look into the details for Oracle, the table at the bottom listing single events shows "Partner-Registry" as the event type. And there is a drop down for "Confidence". Anything other than "Low" will yield zero events.
And they seem to confuse the hallucinated "A1 Wholesale is a leading distributor of a wide range of products, including electronics, household goods, and office supplies." with Telekom Austria (a1.net).
😅 1
c
lol
o
Asked Scarf's "AI" once again for counts per version. Detected some versions appearing in a 3 month period, but missing in an overlapping 3 month period. Of course, the "AI" was claiming I was right in noticing the inconsistency and making up some explanations for the causes. After being asked about the actual missing numbers:
I cannot show the download counts for those missing versions (0.6.0-K2.1.0, 0.7.0-K2.0.0, 0.7.0-K2.2.21, 0.7.1-K2.1.20) for the August 3 - December 3 period because they weren't included in Scarf's distribution analytics results for that timeframe.
The distribution analytics API only returns versions that meet certain internal thresholds or ranking criteria, and those four versions apparently didn't make the cut for the longer 4-month period, even though they appeared in the shorter 3-month period.
This suggests Scarf's analytics may use different filtering logic (like minimum download thresholds or top-N limits) depending on the time range length or total data volume, which would explain why they appear in one period but not the other.
c
24 Views
Open in Slack
PreviousNext
Powered by