datashield.login: Extend datashield.login from package opal

In IulianD/dsSandbox: Local DataSHIELD sandbox

Description

Creates the necessary software infrastructure for the usual opal and datashield functions to be able to run in local pseudo-sessions as well as in remote real sessions

Usage

datashield.login(which_connections = names(local_conns), ...)

Arguments

which_connections	optional, a vector containing the names of the local "sessions" to "connect" to. Normally the output of dssCreateFakeServers (or a subset of it)
...	optional, the parameters for opal::datashield.login (logins dataframe, etc) in case we want to connect to any remote, real opal servers

Details

This function creates the final connection object and assigns it in the global environment. Moreover:

• If necessary establishes connections to remote, real servers

• Creates variants of the generics datashield.assign , datashield.aggregate and datashield.symbols that work on the local pseudo sessions.

• Returns a vector that can be used in various dsBaseClient (and other) functions. Attention: these functions will work only if called with an explicit datasources parameter (ex: ds.var('some_vector', datasources = myopals) and not: ds.var('some_vector'))

Value

a vector containing the names of all the establised connections (real and fake)

See Also

dssCreateFakeServers

Examples

# Mixed opal connections, local(fake, in my session) + remote(real)
# Read a real connection dataframe fro a file
logindata <- read.delim('/path to your/logindata.txt')
# say it looks like this:
logindata
  server                     url       user      password    table
1  real1  https://remote.opal.node  some_user   some_pass   proj.TABLE
# create 2 local connections: 
x <- dssCreateFakeServers('all_conns',c('fake1', 'fake2'))
#login:
opals <- datashield.login(x,logindata)
# opals contains 3 connections, 2 locals, one remote. You can examine them:
opals
all_conns
# create a function called partial_data in the local session as well as on the remote node
# on the remote node this function must be published as an aggregate method:
partial_data <- function (what, start = NULL, end = NULL) {
 data(list = list(what), envir = environment())
 my.df <- get(what)
 if (is.null(start)) {
   start <- 1
 }
 if (is.null(end)) {
   end <- nrow(my.df)
 }
 my.df <- my.df[start:end, ]
 assign(what, my.df, pos = parent.frame())
 TRUE
}
the above loads a piece of an R dataset
# we can use it to load 3 chunks of the 'iris' dataset, each on a different node
datashield.aggregate(opals['fake1'], as.symbol('partial.data("iris", 1, 40)'))
datashield.aggregate(opals['fake2'], as.symbol('partial.data("iris", 41, 100)'))
datashield.aggregate(opals['real1'], as.symbol('partial.data("iris", 101, 150)'))
# run some ds... commands:
ds.summary('iris', datasources = opals)
ds.mean('iris', datasources = opals) # always specify the datasources explicitly
# where's my local data?
ls(envir = all_conns$locals$fake1$envir)
ls(envir = all_conns$locals$fake2$envir)

datashield.logout(opals)

IulianD/dsSandbox documentation built on Dec. 24, 2019, 9:44 a.m.