JE Meeting 02/12/19

Research update & questions

• JE: text has a horizonal preference, doesn’t work horizontally. make time go vertically, so we have room. can stack text in such a way that newer text overlays older text. Another way is only show the value at the cursor and just show that there are values elsewhere. Numeric values can be hinted at height of bar. Complex values don’t have good hints. Should at least mockup the idea of rotated 90 degrees (definitely before coding it!), because it breaks that dataflow diagram expectation but it may work ok.
• JE: I’m not sure how important visually correlating timelines are when you have a cursor that puts a red dot at the same time t and connects them and you can adjust the cursor red dot on all timelines in synchronous, plus an inspector so show value at cursor
• JE: scrubbing window just like video editing app, except we’ve exploded the timeline in space, but they are synchronized so they should all scroll and zoom together. My cursor idea is just like the movie app vertical like.
• JE: different kinds of tick marks could help make it seem like it’s moving, like a grid, below the line, or just wait till they click on things and then they’ll see them move. subtle tick mark could do it. play with it. could use color…

• JE: layout is a black art… have some boxes, divide up the boxes, translate everything into the layout box… layout of graphs is a dark art… graph layout: there are thousands of papers on this, and packages such graphviz that give you choices on which algorithms to use. try to use a package that gives graph layout things, to minimize line overlap as well as various weights on keeping certain things lined up, and other heuristics, etc.

• Doing individual visualizations
• I’ll probably want the arrows to stay stationary and have the events move leftwards (to simulate the arrow moving), and give you the ability to scrub time somehow. I can then probably delete the pan and zoom stuff. I wonder if I’ll need some background context to make it clear the arrow is “moving”, such as a tickmarks on the timelines.
• I am curious how I’ll do layout algorithms for this diagrams in the general case. I shot a message to Josh Horowitz to ask what he did for Pane because he mentioned it was difficult but he came up with something good. I’m mostly curious about the laying out of boxes and arrows at a high level but this will also be interesting even at the smaller levels, of calculating pixels relatively for all things. One such thing that’s particularly important is laying out the text labels for Events, particularly when Events occur close together in time. The tooltip solution I have now is meager at best.
• Make a sharper distinction between how I treat Behaviors and Events. This will become more apparent for pull-based Behaviors… I should also strike up a discussion with Simon about how his push-based Behaviors are really Dynamics (from Reflex).

Salon de Refuge paper

• go to >programming<? JE yes!
• dates (april 1-4)? JE probably go for the whole week, but P/X and Salon minimally, and meet people.
• you going? JE maybe

Future directions

• keep up the (x,y,t) => color? do GUIs with text with that?
  • JE: this hasn’t been done. language of combinators on these functions, such as overlay one on top of another. another such as next to each other. then add
• focus more on dataflow (lucid dataflow?) vs control flow? more on IO monad considered harmful?
  • JE: there is some support for this idea that the IO monad and algebraic effects are the new hotness
• more on denotational semantics? (vs declarative landin, etc)
• • JE: people do operational semantics these days, which is purely syntax, standard reduction semantics, written as inference rules, such as Cyrus Omar’s paper or anyone, really
    • TODO read: https://www.cs.ox.ac.uk/files/3228/PRG06.pdf
    • JE: declarative means timeless (JE’s definition)