Re: [tracecompass-dev] UI for external analyses

[Alexandre Montplaisir <alexmonthy@xxxxxxxxxxxx>]

On 2016-02-15 06:09 PM, Matthew Khouzam wrote:
> On 16-02-15 05:37 PM, Alexandre Montplaisir wrote:
> > Hi all,
> >
> > I am currently working on finishing up the integration between Trace
> > Compass and the lttng-analyses project [1]. For those not familiar
> > with it, this is a set of scripts that can be run on LTTng traces to
> > extract statistics, graphs etc. in different categories (I/O,
> > scheduling, CPU usage, and so on). The integration with Trace Compass
> > will allow calling these scripts directly from the UI, and reporting
> > the resulting in tables or charts in Trace Compass.
> >
> > One thing I am wondering is how to present this new feature in the UI.
> >
> > Some background information first:
> > Trace Compass has the notion of "analysis", which means "something
> > that runs on a trace". Typically those analyses first create some
> > indexes, statesystems, etc., and then show their results in views
> > (like the Control Flow view, CPU Usage view, etc.). These views are
> > dynamic, which mean they will follow time range selections and update
> > what they show accordingly.
> > However, "external" analyses like the lttng-analyses function a little
> > bit differently. They do not run automatically on a trace, instead the
> > user decides to run them, either on the whole trace or on a given time
> > range. The results are then shown in static tables or graphs. Unlike
> > the "dynamic" views from before, these tables will not update their
> > contents. They just show the static results of the analysis, which
> > we've been calling "report".
> >
> > So it goes like this:
> > (standard) analysis -> dynamic views
> > external analysis -> report -> static tables/graphs
> >
> > Note that subsequent runs of a given external analysis would produce
> > separate reports. It could be useful to have each one of those reports
> > remain available, until they are explicitly deleted.
> >
> >
> > In my current prototype, I just display all available external
> > analyses under the trace, just like standard analyses are (see
> > attached screenshot). They can then be invoked by right-click on them
> > and choosing "Run External Analysis" from the contextual menu. The
> > results are displayed in non-singleton views, one per table of output
> > of the analysis.
> >
> > This is not ideal, because it's confusing which analysis are supposed
> > to be run manually, and which are just there for listing their
> > outputs. A potential alternative would be to have a new "Run External
> > Analysis" option, from the context menu of the *trace*, which would
> > open a dialog to handle all available external analyses.
> >
> > Also, we'd probably need a new name instead of "External Analysis".
> > These may not even be external eventually, they could be defined in
> > Java, shipped with Trace Compass, and be able to be called on demand
> > by the user.
> I think analysis snapshot would be more descriptive.
>
> Think about a profiler, you take snapshots and verify their results.
>
> Also, you will need to find a way to manage them, you will have a lot of
> snapshots, maybe they should be auto archived or something... I know,
> v2.0. :)
>
> Do you have a git branch we can play with? it can help us get a better
> feel for the feature than a simple screenshot.

Sure! Here's a branch of the TC prototype:
https://github.com/alexmonthy/tracecompass/tree/feature-python-analyses-integration
(there are new plugins, so you will have to import them in the workspace 
if you launch it from Eclipse)

You will also need the lttng-analyses tree:
https://github.com/lttng/lttng-analyses
clone and run:
$ sudo ./setup.py install

There are packages in the Ubuntu PPA and such, but these are too old, 
the -mi scripts are presently only in master, and will be in 0.4.

Then the analyses should show up under any LTTng kernel trace, 
right-click one of them and select "Run External Analysis". It will run 
the analysis on the currently selected time range, or the whole trace if 
there is no time range selection.

Once done it should open its results in new views. There is a menu in 
each of these views to create histogram charts from the contents (it may 
not make sense for all views though). Also, if a table row represents a 
time stamp or time range, clicking it should select that time range in 
the other views.

Cheers,
Alex

> > Thoughts?
> >
> >
> > Thanks,
> > Alex
> >
> >
> > [1] https://github.com/lttng/lttng-analyses