ENH: use etelemetry to check for a newer version available upon running a command or importing .api

[yarikoptic]

It already helped me to realize that I forgot to "release" 0.2.* on Github. Now they are released, so I guess it takes etelemetry service some time to update its knowledge. Hopefully tomorrow we should see the effect.

Attn @satra who might provide recommendation and also I wonder if it would be worth trying to generalize such check_available_version and ship it within etelemtry-client itself, so users did not have to reimplement similar logic. Could have signature like

check_available_version(github_project, current_version, info, debug)

where info and debug will be the callable (with the same signature as of the logger) to use. What do you think @satra ?

TODOs:

• [ ] Wrap the entire body to be ran in a separate thread. Even though it might not be taking long, IMHO it might be of benefit to cmdline apps to not spend time waiting

[satra]

generalize such check_available_version and ship it within etelemtry-client itself, so users did not have to reimplement similar logic.

i think that would be good.

[yarikoptic]

generalize such check_available_version and ship it within etelemtry-client itself, so users did not have to reimplement similar logic.

i think that would be good.

oh, and you have a similar one in nipype (which you pointed out in https://github.com/nipy/heudiconv/issues/392) https://github.com/nipy/nipype/blob/master/nipype/__init__.py#L69 which also adds ability to prescribe some versions as bad_versions, i.e. carrying a critical bug I guess, so then info level should become a warning.

[kyleam]

From a user perspective, I very much dislike a tool or library reaching out to a server each time it is invoked, for activity unrelated to anything I explicitly requested or wanted. Please consider making this opt-in or at least opt-out.

[satra]

From a user perspective, I very much dislike a tool or library reaching out to a server each time it is invoked, for activity unrelated to anything I explicitly requested or wanted.

@kyleam - don't you do that every time you use the browser or any app on your phone?

[kyleam]

@kyleam - don't you do that every time you use the browser

Using a browser is a very specific request by me to connect to servers. We're talking about a local python library/command-line tool that does not even require an internet connection for many operations. I should be able to use a local tool on my computer without it phoning a server.

[satra]

but your apps are local and they do tell you updates are available. i think providing a way for developers to communicate with their users to notify bad versions, updates is a good thing.

i understand the philosophy, and ET does allow a way to disable it and handles no connection uses, but i am saying that most apps today communicate with servers. i don't necessarily see that as concern if we ensure that people know that the app (even if it's a python tool on the command line) will communicate.

[yarikoptic]

From a user perspective, I very much dislike a tool or library reaching out to a server each time it is invoked, for activity unrelated to anything I explicitly requested or wanted. Please consider making this opt-in or at least opt-out.

I believe there is a possibility to opt-out via env variable (NO_ET). Unfortunately, due to necessity to provide funding agencies with proof of user base, we did agree (during last EAB meeting) that opt-in would be our way to go.

[kyleam]

@satra

but your apps are local and they do tell you updates are available.

[ Sorry, I ignored the smartphone part of your reply since I don't use one. :/ ]

Even if you assume that your users are comfortable with their smartphones behaving this way (most of them probably are), that doesn't mean that they expect or prefer their, say, Debian systems to behave that way. And given that reproman isn't being used on smartphones, the latter is much more relevant.

i don't necessarily see that as concern if we ensure that people know that the app (even if it's a python tool on the command line) will communicate.

I agree with that. But as this PR stands (it's of course still work-in-progress) it doesn't look like we do anything to ensure that users know this or advertise how they can disable it.

[kyleam]

@yarikoptic

I believe there is a possibility to opt-out via env variable (NO_ET).

Thanks.

[yarikoptic]

superseded with smaller #564