Correcting the order of traces

[13bzhang]

This is a pull request to address #225 "Ordering of traces is wrong". This seems to be a problem with a lot of geoms. I am attempting a general first-cut solution (pretty drastic try) by reversing the order of the traces. I want to see how it fares in the plotly test tables.
@cpsievert @chriddyp

[13bzhang]

Ok, maybe that is too drastic of a change to reverse the order of the traces since a lot of the tests depend on the order of the traces from the old code. Let me think this over.

[13bzhang]

After realizing how dumb my initial solution was, I decided to reverse the order of the traces only when there is a problem. Right now the two cases where I have detected a problem include:

1. geom_area
2. geom_density when position = "stack"

So I added the following lines to trace_generation.R:

  # reverse the traces in the following cases:
  # geom_area
  # geom_density with position = stack
  if (g$geom == "area" | 
        g$geom == "density" & l$position$.super$objname == "stack"){
    traces <- rev(traces)
  } else{
    traces
  }

I also included tests for the two cases that I list above to test-ggplot-area.R and test-ggplot-density.R, respectively. I supect there are other cases of these trace orders out of whack. I will keep looking for them. If you find them, let me know.

@cpsievert @chriddyp

[cpsievert]

The message below was automatically generated after build https://travis-ci.org/ropensci/plotly/builds/65741975

On TravisCI, commit 495ba85 was successfully merged with 6ff8831 (master) to create 5660dda. A visual testing table comparing 6ff8831 with 5660dda can be found here:
http://ropensci.github.io/plotly-test-table/tables/5660ddaf3015e927d1c47960946647ad13973176/index.html

[13bzhang]

Hit a snag because the dataset for one of the new tests required the package dplyr. I rewrote the code so the data gets prepped using base R instead. Good ol' aggregate.

[cpsievert]

The message below was automatically generated after build https://travis-ci.org/ropensci/plotly/builds/65745397

On TravisCI, commit 6afd57e was successfully merged with 6ff8831 (master) to create 2c09c51. A visual testing table comparing 6ff8831 with 2c09c51 can be found here:
http://ropensci.github.io/plotly-test-table/tables/2c09c512f388905bc6cb79815db4581e990aa046/index.html

[cpsievert]

The message below was automatically generated after build https://travis-ci.org/ropensci/plotly/builds/66496333

On TravisCI, commit a892b5d was successfully merged with 6ff8831 (master) to create e0f0fae. A visual testing table comparing 6ff8831 with e0f0fae can be found here:
http://ropensci.github.io/plotly-test-table/tables/e0f0faeaa8a3165b7620956f3d5c903d0bb8e17b/index.html

[cpsievert]

The message below was automatically generated after build https://travis-ci.org/ropensci/plotly/builds/96085859

Detected 3 differences ->

http://cpsievert.github.io/plotly-test-table/bdcba39/area-traces_order
http://cpsievert.github.io/plotly-test-table/bdcba39/density-traces_order
http://cpsievert.github.io/plotly-test-table/bdcba39/scatterplots-jitter

[cpsievert]

^^^ much better than the test table, huh @chriddyp? :)

(that jitter test is a false positive, just ignore it, I'll fix it later)

[cpsievert]

The message below was automatically generated after build https://travis-ci.org/ropensci/plotly/builds/96121548

Detected 3 differences ->

http://cpsievert.github.io/plotly-test-table/0696ba9/area-traces_order
http://cpsievert.github.io/plotly-test-table/0696ba9/density-traces_order
http://cpsievert.github.io/plotly-test-table/0696ba9/scatterplots-jitter