Mccalluc/better ids

[mccalluc]

This provides more informative row labels in the heatmap. I have a few concerns:

• Search is just doing a linear scan of the labels: if the labels get longer, search will slow down proportionally. I had tried using Whoosh, but the indexing took a very long time. Might be worth trying to find an intermediate solution.
• Manually testing it, with some frequency, things just don't work. I've thought these were race conditions in Dash that were outside my control, but I'm not really confident about that. With these changes, I feel like it's even more frequent. Maybe we should understand those before pushing the framework even harder.
• Estimating the width of the labels: Right now I'm just assuming 10 pixels per letter, but if we're putting more in there, that estimate could be wide of the mark. Is it worth getting actual character widths?

I had begun it thinking it'd be an easy way to get a new version out, but there are enough changes here that I don't think that would be a good idea. I'll go back and make a trivial branch, and hopefully the travis rebuild will work.

[mccalluc]

• table not working on demo data
• ok to do left substring search, but could be a later pr
• there may be a lot of text columns: just take the last one

[mccalluc]

@gmnelson / @scottx611x (and adding @jkmarx, new policy) If something is a slightly risky upgrade, how would you all feel about having 2 visualizations active in Refinery. It would also make it easier to compare performance one against the other.

(I still need to address the feedback from Geoff.)

[gmnelson]

@mccalluc If the 2 visualizations were named sufficiently different so that users were not confused as to which to use, I don't foresee this presenting an issue right now, given the limited usage. I can discuss this more, if needed, after the meeting tomorrow.

[mccalluc]

@jkmarx : Is this improved enough to merge? (Thanks for pushing back!)