• 1 layer faceting ( x – cross )
• 1 layer faceting ( + – concat )
• multi-layer faceting
• labels

• chart support — given an array of binned data e.g., [1,1,1,1,1,1,2,2,2,2,3,3,3,4,4,5,5,5,5,5,5,5] count unique values and plot histogram. (Zening)
• Extend Datavore to call binning — see vizrec code (Ham)

• make import vegalite, clusterfck works

Some Related Post:
deepak1556/gulp-browserify#42

bar.x-avg_Cost__Total_$-Q.y-bin_Speed_IAS_in_knots-Q
<
point.x-avg_Cost__Total_$-Q.y-bin_Speed_IAS_in_knots-Q

Specifically, only apply bins to all of them, or aggregation to all of them or do nothing (raw).

Currently if we have *(Q1) and *(Q2) selected, we get all the following

• Q1 x Q2
• AVG(Q1) x AVG(Q2)
• BIN(Q1) x BIN(Q2)
• AVG(Q1) x BIN(Q2)
• BIN(Q1) x AVG(Q2)
• We shouldn’t include the latter two unless the user explicitly specifies that they want BIN.
• for BIN(Q1) x BIN(Q2), we should automatically add count to the table (so we get heat map or table of circular sizes..)

Question:

• What should we do if users specify MIN(Q1), *(Q2)?

should we generate BIN and MIN or BIN and AVG for Q2

• quant
• time

Rank

• raw > avg, sum > bin ?

• Support Year only

** Advance **

• Support Multi Level Expansion (Year -> Month -> Day)

Most of the time it’s less likely to be the top one.

• adjust the ranking
• add pruning rule

• Start by prototyping symbol map of USA (by states) separately

Asking for “major_genre”, “imdb_votes”, “production_budget” takes 6 sec

(maybe due to the fact that some visualizations are being rendered too.)

Generate

• aggregate
• bin

Prune

• no measures d3196df
• aggregate with no dimensions d3196df

Effect__Amount_of_damage (O)
When__Phase_of_flight (O)
avg Cost__Total_$ (Q)

is

bar.y-Effect__Amount_of_damage-O.color-When__Phase_of_flight-O.x-avg_Cost__Total_$-Q

worse than

point.x-Effect__Amount_of_damage-O.y-When__Phase_of_flight-O.size-avg_Cost__Total_$-Q

• genAggregate pruning

e.g., when having QxQxO

• expand projections
• think about criteria for projections

• allows row/col without x,y with text.

Encode more important information more effectively

• selected field should be considered more important than field automatically added.

Generate some tables like:

univariate

• Q
• T
• O —> becomes OxT

bivariate

• QxQ
• QxT
• QxOs
• QxOb
• Qx#
• TxOs
• TxOb
• Tx#
• OxO —> becomes OxOx#
• Ox#

trivariate

• Q x Q x Q
• Q x Q x T
• Q x Q x Os
• Q x Q x Ob
• Q x Q x #
• Q x T x Os
• Q x T x Ob
• Q x T x #
• Q x Os x Os
• Q x Os x Ob
• Q x Ob x Ob
• Q x Os x #
• Q x Ob x #
• T x Os x Os
• T x Os x Ob
• T x Ob x Ob
• T x Os x #
• T x Ob x #
• Os x Os x #
• Ob x Os x #
• Ob x Ob x #

Note: Ob = high cardinality ordinal, Os = low cardinality ordinal

(We can further improve ranking with the non-pruned set after the deadline)

• make sure we don’t recommend bar

For example,

bar.y-Effect__Amount_of_damage-O.x-avg_Cost__Repair-Q.size-bin_Cost__Total_$-Q

• 2D Plot

• Zooming and Panning

bar.y-Effect__Amount_of_damage-O.size-bin_Cost__Repair-Q.x-avg_Cost__Total_$-Q

• x,y, row, col

depends on vega/vega-lite#24

mouseover marks, show underlying data’s information.

Right now we have to many similar, redundant results.

Interactions

• rect for selections

• sort icons
  

bar.y-Effect__Amount_of_damage-O.x-avg_Cost__Repair-Q.color-bin_Cost__Total_$-Q