Timestamped index to buffer

[jayvdb]

#64 (comment) mentions merging stdout and stderr to reflect that both typically appear on the same user console, interleaved.

Another similar feature is timestamping the activity occurring on the "console", so those timestamps can be shown to the user, or timing info used in test logic.

[bskinn]

Would this make more sense as a logging feature, tied/related to #43?

I guess the logging would derive from some sort of (timestamp, message) semantic structure to the content that's handled by the mocked stdfoos, though -- no reason not to expose that structure through the API in some fashion.

[jayvdb]

I've tried tightly coupling stdio with logging once before, in pywikibot, and it is working there (or was last I looked, but I am not active there for over a year). It has some pleasing benefits, like this feature being 'free', but it required some nasty initialisation sequences. I suspect that it might be less troublesome here when there isnt a real console attached, and also require less hacks because of no py2 support. Worth trying again; I'll do a POC based on pywikibot and we can evaluate it.

[bskinn]

Seems to me the "easiest" way to do it is to have the streams actually represented by an underlying list of ~annotated chunks of text, as was done with one of the entities over at coala-utils. The stream behaviors would then have to be implemented as an interface layer over this data structure.

Logging could then occur around the various interactions with the data structure... read/write/etc.

I think dealing with arbitrary seeks within a given "message unit" could be hairy, if such were desired to be implemented, though.