- mostly everything in js?
- interfaces in R/Python/Julia?
defaults could be specified as follows (should be specified in a JSON file, maybe extend Vega format?) for sharing we could store this file in a git repos (on github? http://fitzgen.github.io/github-api/)
list(
lines=aplot(first_mapping=aes(x=year, y=life.expectancy, color=region),
geom_line(data=WorldBank, size=4, first_mapping=aes(clickSelects=country)), # 4 pixels -- not passed through a scale.
geom_tallrect(data=years), # needs year_xmax and year_xmin columns,
# ignores aes(color) since there is no color column
# aes(x) is plot-specific but aes(clickSelects) is layer-specific.
points=aplot(first_mapping=aes(x=birth.rate, y=life.expectancy, color=region, size=population),
# aes generate axes or guides.
geom_point(data=WorldBank, showSelected(year), encode=vars(color, size)),
# points always take x and y, this one also takes optional data color, size
geom_text(data=WorldBank, showSelected(year, country), first_mapping=aes(label=country)), # takes x, y, label
geom_text(data=years, showSelected(year), x=0.5, y=1, first_mapping=aes(label=sprintf("year = %d")))
)
)but minimally we just specify the data sets and maybe some common scales… actually we could specify just a href to the data set
animint(data=list(years=years, WorldBank=WorldBank),
scales=list(region.colors=region.colors))and then on the web page we could interactively define the linked plots by filling in the following 2 tables (1 for each plot).
| scale=region.colors | |||||||||||
| data | geom | x=year | y=life.expectancy | stroke=region | fill | stroke-width | opacity | group | label | clickSelects | showSelected |
|---|---|---|---|---|---|---|---|---|---|---|---|
| / | < | < | |||||||||
| WorldBank | line | encoded | encoded | encoded | - | 4 | 1 | country | - | country | [] |
| (not selected) | - | - | - | same | - | same | 0.5 | - | - | - | - |
| years | tallrect | encoded | - | “black” | “black” | 1 | 0.5 | - | - | year | [] |
| (not selected) | - | - | - | same | same | same | 0 | - | - | - | - |
and for the scatterplot,
| scale=region.colors | |||||||||||
| data | geom | x=birth.rate | y=life.expectancy | stroke | fill=region | size=population | stroke-width | opacity | label | clickSelects | showSelected |
|---|---|---|---|---|---|---|---|---|---|---|---|
| / | < | < | |||||||||
| WorldBank | circle | encoded | encoded | “black” | encoded | encoded | 1 | 1 | - | country | [year] |
| (not selected) | - | - | - | same | same | - | 0 | 1/2 | - | - | - |
| WorldBank | text | encoded | encoded | - | “black” | 12 | - | 1 | country | (none) | [year, country] |
| years | text | 0.5 | 1 | - | “black” | 12 | - | 1 | yearLabel | (none) | [year] |
The main idea is that there are two kinds of aesthetics:
- some plot-specific, each need an axis/legend to decode: x, y, stroke, fill, size, stroke-width, opacity.
- some geom-specific: clickSelects, showSelected, group, label.
size aesthetic controls <text style=”font-size:
x=year with geom line assumes that WorldBank has a year column and with geom tallrect looks for columns named year_min and year_max in years.
x, y, size have 3 kinds of values:
| unit type | x | size | constant? |
|---|---|---|---|
| data | year 1996 | 1e8 people | |
| canonical | [0, 1] | 30px circle diameter | yes |
| svg | x=243 (px) | r=15 (px) |
but stroke-width, stroke, fill, opacity only have 2 kinds of values
| unit type | stroke | constant? |
|---|---|---|
| data | region Europe | |
| svg | “red” | yes |
if we specify constants for these aesthetics, they should be interpreted as canonical (x, y, size) or svg (stroke-width, stroke, fill, opacity) values.
scales take units from the first row and give units in the second row.