Gantt performance is really mediocre for huge traces.

[dosimont]

Query duration is too long compared to what I get with Ocelotl.
In the case of Nancy 700 lu C, I succeed to load a focus on the temporal perturbation, but in about 15 minutes (compared to 5 minutes for Ocelotl). You really must use an iterator and only one query, even if the events are not loaded in the exact time order (this is not a critical problem since the Gantt is not usable for an analysis while it is not completely loaded).

[dosimont]

Since the query management does not depend completely on the graphical API, you should take this into account for the new Gantt.

[generoso]

Query management does not depend at all on the graphical API. On the contrary, the existing Gantt plugin has already been designed in a modular way: the data loading part and the data visualization part can be provided using the extension point mechanism.

We can take advantage of this modular architecture to rapidly prototype the different data loading mechanism proposed by @dosimont in the existing Gantt (providing a new loader class).

Of course, if the results are satisfying (especially for huge traces), consider this aspect when developing the new Gantt (issue #164).