legends render on top of each other

[ijlyttle]

I didn't see this in my quick scan of open issues:

It seems that each legend wants the same top-left attachment point within the view, so end up being printed on top of each other.

Here's an example:

ggvis(mtcars, props(x= ~mpg, y= ~disp, fill= ~factor(cyl), size= ~hp)) + 
  layer_point()

Here's the session info (probably time to update to the new version for me):

> sessionInfo()
R version 3.0.2 (2013-09-25)
Platform: x86_64-apple-darwin10.8.0 (64-bit)

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] ggvis_0.1.0.99    devtools_1.4.1.99 ggplot2_0.9.3.1   stringr_0.6.2    
[5] reshape2_1.2.2    lubridate_1.3.3   magrittr_1.0.0    dplyr_0.1.2      
[9] knitr_1.5.22     

loaded via a namespace (and not attached):
 [1] assertthat_0.1     bitops_1.0-6       caTools_1.16       colorspace_1.2-4  
 [5] dichromat_2.0-0    digest_0.6.4       evaluate_0.5.1     formatR_0.10      
 [9] grid_3.0.2         gtable_0.1.2       httpuv_1.2.3       httr_0.2          
[13] labeling_0.2       MASS_7.3-29        memoise_0.1        munsell_0.4.2     
[17] parallel_3.0.2     plyr_1.8.1         proto_0.3-10       RColorBrewer_1.0-5
[21] Rcpp_0.11.0        RCurl_1.95-4.1     RJSONIO_1.0-3      scales_0.2.3      
[25] shiny_0.9.1        tools_3.0.2        whisker_0.3-2      xtable_1.7-1   

Thanks!

[ijlyttle]

I see that this issue is alluded-to in the axes-legends vignette, and I found a workaround for the time-being:

ggvis(mtcars, props(x= ~mpg, y= ~disp, fill= ~factor(cyl), size= ~hp)) + 
  layer_point() + 
  guide_legend("size", orient = "left")

I'll leave it to Hadley et al. to decide if this should remain an issue.

I'll take my question on other workarounds to the google group.

[wch]

It looks like there's a bug in creating the legend section of the spec. This should work, but doesn't: ggvis(mtcars, props(x =~ mpg, y =~ disp, fill =~ factor(cyl), size =~ hp)) + layer_point() + guide_legend("size", properties = list(legend = props(y := 150)))

The problem is that in the spec, there should be a section that looks like this:

"legends" : [
  {
    "size" : "size",
    "orient" : "right",
    "properties" : {
      "legend" : {
        "y" : {
          "value" : 100
        }
      }
    },
    "title" : "hp"
  },
 ...
]

But instead the "properties" subsection looks like this, with an extra "update" in there:

  "properties" : {
    "legend" : {
      "update" : {
        "y" : {
          "value" : 150
        }
      }
    }
  }

[ijlyttle]

Hi Winston,

It looks like you've got the issue isolated.

Thanks,

Ian

[hadley]

Same as #145

[wch]

Updated example:

mtcars %>% ggvis(x= ~mpg, y= ~disp, fill= ~factor(cyl), size= ~hp) %>% 
  layer_points()

It would be nice to be able to automatically position the legends so they don't overlap. I suspect we won't be able to position them exactly, but that we'll have to use a heuristic and hope that there's no overlap.

Here's how to manually position the legend. It's way too verbose as-is:

mtcars %>% ggvis(x= ~mpg, y= ~disp, fill= ~factor(cyl), size= ~hp) %>% 
  layer_points() %>%
  add_legend(size = "size", properties = legend_props(legend = list(y = 100)))

[yingchen69]

Hi, I tried a little change, but had no success

mtcars %>% 
  ggvis(x= ~mpg, y= ~disp, fill= ~factor(cyl)) %>% 
  layer_points(shape= ~factor(gear)) %>%
  group_by(gear) %>%
  layer_model_predictions(model = "lm", strokeDash = ~factor(gear)) %>%
  add_legend(shape = "shape", properties = legend_props(legend = list(y = 100)))

Instead,

        Guessing formula = disp ~ mpg
        Error in add_legend(`mtcars %>% ggvis(x = ~mpg, y = ~disp, fill = ~factor(cyl)) %>%           layer_points(shape = ~factor(gear)) %>% group_by(gear) %>%     layer_model_predictions(model =  "lm", strokeDash = ~factor(gear))`,  : 
        unused argument (shape = "shape")

How can I fix this?

Thanks a lot,

Ying

[ijlyttle]

Hi Ying,

This may just be a matter of a misnamed arguement in add_legend(): shape -> scales

Does this work for you?

mtcars %>% 
  ggvis(x= ~mpg, y= ~disp, fill= ~factor(cyl)) %>% 
  layer_points(shape= ~factor(gear)) %>%
  group_by(gear) %>%
  layer_model_predictions(model = "lm", strokeDash = ~factor(gear)) %>%
  add_legend(scales = "shape", properties = legend_props(legend = list(y = 100)))

[yingchen69]

Hi Ian, it works! Thanks a lot for the help!

Ying

[yingchen69]

Hi,

Sorry I got another issue. This method seems not work if I have add_tooltips() in the same ggvis plot.

mtcars %>% 
  ggvis(x= ~mpg, y= ~disp, fill= ~factor(hp)) %>% 
  layer_points(shape= ~factor(gear)) %>%
  add_tooltip(function(df) df$disp) %>%
  group_by(gear) %>%
  layer_model_predictions(model = "lm", strokeDash = ~factor(gear)) %>%
  add_legend(scales = "shape", properties = legend_props(legend = list(y = 200)))

It works for the first second, then the bottom legend jumps to top and overlaps with the top legends. One more trick?

Thanks a lot,

Ying

[ijlyttle]

This is more of a workaround than a fix.

Try adding set_options(duration = 0) to the end of your ggvis pipe.

[yingchen69]

Hi Ian, thanks a lot. It works!

Ying

[justacec]

A was looking at this issue because I was running into it as well. Although I see the fix to manually place the y position of the legend, how do you know what the y value should be. My legend can shrink and grow. Is it possible to distribute the legends horizontally all cascading from the top? That would leave a lot of undesirable white space but it would avoid the issue of manual placement.

[voxnonecho]

For reference, StackOverflow user experiencing the same problem: http://stackoverflow.com/questions/30916337/legends-on-ggvis-graph-are-overlaping-when-using-tooltip/