MacPorts Statistics (was Re: usage numbers for macports vs. homebrew?)

[Jeremy Lavergne]

On Mar 19, 2014, at 15:16, Craig Treleaven <ctreleaven at cogeco.ca> wrote:

> As I understand it, the statistics process arose in GSOC 11, right?  Is it pretty much as described at:
> https://trac.macports.org/wiki/MacPortsStatisticsGSoC2011
> There is a (short) thread about the proposal here:
> https://lists.macosforge.org/pipermail/macports-dev/2011-March/014384.html
> which suggests it is inspired by Debian's Popcon tool ("Popularity Contest”).

Inspired, sure. But that doesn’t mean we’re going for 1:1.

> Is the above accurate?  I understand that Popcon tracks the last access time (atime) for files in a port and that our implementation does not.
> 
> From what I can see, our tool will not gather whether a port was Requested or not.  I think that's a key data point.  Some libs will get installed very frequently as dependencies but will seldom be requested.

At the time of GSoC11, I don’t think MacPorts even had the notion of requested ports.

Nothing prevents this from being added at any later date.

> Also, this is implemented as a weekly launchd job.  What happens if there is a problem somewhere between the users machine and the server receiving the data?  Does the user job abort gracefully?  Try again?

Port uses curl to submit, and then catches and prints any errors. Finally returns 0.
But don’t take my word for it: the code would reveal the truth there.

http://trac.macports.org/changeset/82920/

> What is the user wants to get rid of MacPorts and deletes /opt/local/*. My experience is that launchd will _never_ stop trying to run the missing job and will spam the Console log with complaints about it.

Are you saying no one ever should use launchd?
It doesn’t sound like this is our problem.

> Do we have a sample statistics page for a port/all ports?

I believe Clemens has a test box.