Plotly: add_trace in a loop

Question

I'm trying to add_trace ad each loop, but I get only one plot with multiplies lines on over each other.

mean <- -0.0007200342
sd   <- 0.3403711
N=10
T=1
Delta = T/N

W = c(0,cumsum( sqrt(Delta) * rnorm(N, mean=mean, sd=sd)))
t <- seq(0,T, length=N+1)

p<-plot_ly(y=W, x=t)

for(i in 1:5){

  W <- c(0,cumsum( sqrt(Delta) * rnorm(N, mean=mean, sd=sd)))
  p<-add_trace(p, y=W)

}
print(p)

My guess is a scoping issue. When defining `y=W`, `W` is first found within the environment of the plot. Directly using `p<-add_trace(p, y=c(0,cumsum( sqrt(Delta) * rnorm(N, mean=mean, sd=sd))))` resolves the issue, but I imagine you are looking for a more general solution. — Axeman, Dec 09 '15 at 19:01
Also interested in a solution here..seems to be a general plotly issue? I mean aforementioned solution works for this simple example, but not if you have big dataframes. As soon as you have soom loop variable or anything in the plotly call, it only takes the last loop variable... — MichiZH, Dec 10 '15 at 13:10

score 7 · Accepted Answer · answered Jan 22 '16 at 19:11

7

The plot_ly and add_trace functions have an evaluation = FALSE option that you can change to TRUE, which should fix the scope issues.

answered Jan 22 '16 at 19:11

Perry Kuipers

121
6

score 5 · Answer 2 · answered Mar 08 '16 at 14:57

5

Use evaluate = TRUE in add_trace.

answered Mar 08 '16 at 14:57

TheDimLebowski

344
1
3
12

score 2 · Answer 3 · answered Dec 10 '15 at 13:51

Nasty, but works:

mean <- -0.0007200342
sd   <- 0.3403711
N=10
T=1
Delta = T/N

W = c(0,cumsum( sqrt(Delta) * rnorm(N, mean=mean, sd=sd)))
t <- seq(0,T, length=N+1)


for(i in 1:5){

  W <- c(0,cumsum( sqrt(Delta) * rnorm(N, mean=mean, sd=sd)))

  assign(paste("W_",i,sep=""),W)
  assign(paste("Name_", i, sep=""), paste("Name",i,sep=""))
  if(i==1){
    pString<-"p<-plot_ly(x = t, y = W_1, name='W1')"
  } else {
    pString<-paste(pString, " %>% add_trace(x=t, y =",  eval(paste("W", i, sep="_")),", name=", eval(paste("Name", i, sep="_")), ")", sep="")
  }

}
eval(parse(text=pString))
print(p)

score 1 · Answer 4 · answered Oct 30 '16 at 16:57

I'd do this like this:

mean <- -0.0007200342
sd   <- 0.3403711
N=10
T=1
Delta = T/N

# a list with the trace Y values
Ws <- lapply(
  1:15,
  function(idx){
    c(0,cumsum( sqrt(Delta) * rnorm(N, mean=mean, sd=sd)))
  } 
)

# this could be a list with the trace X values, but is just a seq
t <- seq(0,T, length=N+1)

# a list with plotly compliant formatted objects
formattedW <- lapply(
  seq_along(Ws),
  function(idx, datasetY, datasetX){
    return(list( x = datasetX, y = datasetY[[idx]], type="scatter", mode = 'lines+markers'))
  },
  datasetX = t,
  datasetY = Ws
)

# Reduce the list of plotly compliant objs, starting with the plot_ly() value and adding the `add_trace` at the following iterations
Reduce(
  function(acc, curr){
    do.call(add_trace,c(list(p=acc),curr))
  },
  formattedW,
  init=plot_ly()
)

score 0 · Answer 5 · answered Oct 19 '22 at 09:24

I had exactly the same problem. Besides, evaluate = TRUE is not a valid attribute for scattering plots. I solved it by unlisting the vector into the for loop.

for(i in 1:5){

  W <- c(0,cumsum( sqrt(Delta) * rnorm(N, mean=mean, sd=sd)))
  p<-add_trace(p, y=unlist(W))

}

I know this answer is late, but maybe it will be useful to someone else.

score -3 · Answer 6 · edited Feb 09 '17 at 14:08

-3

It is described here : http://www.r-graph-gallery.com/129-use-a-loop-to-add-trace-with-plotly/

save your plot in a variable, and then do add_trace :

p <- plotly(...)

p<- add_trace(p, ...)

edited Feb 09 '17 at 14:08

Alexander Farber

21,519
75
241
416

answered Feb 09 '17 at 13:47

Shiv

1

Plotly: add_trace in a loop

6 Answers6

Linked