3

Hi I want to have a simple function for my local opencpu development server.

getLinearInterpolatedEstimateQuantil <- function(x, type, probs){
  result = quantile(data, type, probs)
  result #print for simpler debug
}

Example Debug input would be something like 

{"type":1,"x":[1,2,3,4,5,6,7,8,9,10],"probs":[0.05,0.25,0.75,0.95]}

But there is a problem: when supplied manually with the sample data like:

debugInput = fromJSON(file("debugInput.json"))

the code compiles.

BUT when I try to access it via http like:

opencpu$browse("library/myLibrary")
curl http://localhost:3469/ocpu/library/predictR/R/getLinearInterpolatedEstimateQuantil/Json -d '{"type":1,"x":[1,2,3,4,5,6,7,8,9,10],"probs":[0.05,0.25,0.75,0.95]}

' -H "Content-Type: application/json"

I only receive as output:

    unused arguments (type = 1, probs = probs)
In call:
getLinearInterpolatedEstimateQuantil(type = 1L, x = x, probs = probs)

So I expect that the parsing has some issues with the arrays?

I hope you can tell me what is wrong with my code.

Edit: I learnt that opencpu performs the json parsing for me. However the code still does not work. (https://www.opencpu.org/posts/scoring-engine/) Edit: still not working Edit: Strange: calling a native function works:

curl http://localhost:5112/ocpu/library/stats/R/quantile/json -d '{"type":1,"x":[1,2,3,4,5,6,7,8,9,10],"probs":[0.05,0.25,0.75,0.95]}' -H "Content-Type: application/json"

however calling my own function results in an error:

curl http://localhost:5112/ocpu/library/predictR/R/getLinearInterpolatedEstimateQuantil/json -d '{"type":1,"x":[1,2,3,4,5,6,7,8,9,10],"probs":[0.05,0.25,0.75,0.95]}' -H "Content-Type: application/json"
unused arguments (type = 1, probs = probs)
In call:
getLinearInterpolatedEstimateQuantil(type = 1L, x = x, probs = probs)

for clarification here my function again:

getLinearInterpolatedEstimateQuantil <- function(x){
  result = quantile(data, type, probs)
  return (result)
}

EDIT again:

library(jsonlite)
myFunction <- function(x, type, probs){
result = quantile(x, type, probs)
return (result)
}

json <- '{"type":1,"x":[1,2,3,4,5,6,7,8,9,10],"probs":[0.05,0.25,0.75,0.95]}'
args <- fromJSON(json)
do.call(myFunction, args)

Results in 

100% 
10 

Warning message:
In if (na.rm) x <- x[!is.na(x)] else if (anyNA(x)) stop("missing values and NaN's not allowed if 'na.rm' is FALSE") :
  Bedingung hat Länge > 1 und nur das erste Element wird benutzt

And 

do.call(stats::quantile, args)

Results in

5% 25% 75% 95% 
  1   3   8  10

Why does the first call result in a warning with different output? Why does the second call work?

zx8754
  • 52,746
  • 12
  • 114
  • 209
Georg Heiler
  • 16,916
  • 36
  • 162
  • 292
  • At a glance: you define a function(x), but your JSON has `type` and `probs` as well; that matches the error message. Shouldn't you declare these parameters? Also, your function currently ignores `x`, but uses `data` which we haven't seen before. – Silly Freak Jun 02 '15 at 22:06
  • Ha nearly there: This is the output of the native call [1, 3, 8, 10] and this the output of my function [10] – Georg Heiler Jun 02 '15 at 22:13

1 Answers1

1

For an RCP with -H "Content-Type: application/json" the top level names in your JSON object must match the parameter names of your function. You can test this as follows:

library(jsonlite)
json <- '{"type":1,"data":[1,2,3,4,5,6,7,8,9,10],"quantil":[0.05,0.25,0.75,0.95]}'
args <- fromJSON(json)
result <- do.call(getLinearInterpolatedEstimateQuantil, args)

So assuming you want to stick with the JSON payload as it is, your function should look like:

getLinearInterpolatedEstimateQuantil <- function(data, quantil, type = 1){

}

For example to call the quantile function directly:

curl http://public.opencpu.org/ocpu/library/stats/R/quantile/json -d \
'{"type":1,"x":[1,2,3,4,5,6,7,8,9,10],"probs":[0.05,0.25,0.75,0.95]}' \
-H "Content-Type: application/json"

Note that the arguments of the json blob type probs and x match the parameter names of quantile.

Jeroen Ooms
  • 31,998
  • 35
  • 134
  • 207