The stats for plugins don't seem to accurately reflect version information. For example, with questionnaire, according to the stats, a third of the sites are using Moodle 3.5 and 3.6. This means that they are likely using the 3.5 release of questionnaire. There have been four releases of questionnaire 3.5. Yet the downloads per month don't show any 3.5 releases. This can't be right?
That wouldn't explain why there is no data for the 3.5.* release. But, I just released a new version (3.5.4), and now there is data. Probably something to do with a delay between when I released and the data.
But this also brings up the previous problem. Because I released 3.5.4, all of the data for 3.5.1, 3.5.2 and 3.5.3 are removed.
The latest 3.5.x version of your plugin is 3.5.4 released on 21 March 2019. The downloads chart does not show figures for the current month. The download stats for the 3.5.4 will appear on the 1st April 2019.
I can understand that in your particular case, it would be useful to also show the download stats for the previous 3.5.x versions but such a behavior contradicts other plugin's requirements. Displaying all the versions makes the chart unreadable. So yeah, it's not ideal.
Out of curiousity, is my case different? Wouldn't other release new versions of their plugins for the same releases of Moodle? If I'm different, what is the best way to do this?
Is there data available that would show the downloads for a particular version of Moodle? That is, show the downloads for all versions of the plugin that indicate it supports 3.5? If the data is there, I could see that being very useful.
Is there data available that would show the downloads for a particular version of Moodle?
IIRC, we have data associated with a particular plugin version. So we probably could have one chart per Moodle version - though it would mean that one plugin version can appear in multiple charts (as it can support multiple Moodle versions).
BTW, are the download stats that important for you? I personally found them pretty irrelevant, given that sites have multiple channels to obtain the plugin code. To me, the only significant stats is the actual number of sites having the plugin installed.
The stats per plugin version are very useful, yes. The actual number of sites having the plugin installed is interesting, but does not tell me what versions they are using. It's also dependent on the site registration data (I believe), so is probably missing a lot of sites. The downloads likely represents more of the reality, assuming most sites install their plugins from the database rather than checking out the repo. Plus I can tell whether users are updating their plugins.
I see the problem now with the "one chart per Moodle version" though... What I'm really looking for is data that shows how many uses of each release tree there are. And I can see even grouping then together will be inaccurate, as the same site may have downloaded multiple releases within a release tree...
What I was hoping for was something similar to the doughnut graph for "by Moodle version" as a "by plugin version". Maybe not possible.
It's also dependent on the site registration data (I believe), so is probably missing a lot of sites.
No. These stats are not related to the sites registration. They are analyzed based on incoming requests for available updates. That feature is independent on whether the site is registered or not. You are right that many sites may have available updates notifications disabled, for sure.
At the moment, we do not have data on "sites per plugin version" available.