advance on main.R

[epclust.git] / code / draft_R_pkg / R / main.R

diff --git a/code/draft_R_pkg/R/main.R b/code/draft_R_pkg/R/main.R
index 4120b39..c3e4b54 100644 (file)
--- a/code/draft_R_pkg/R/main.R
+++ b/code/draft_R_pkg/R/main.R
@@ -1,26 +1,82 @@
-epclust = function(data=NULL, con=NULL, raw=FALSE, K, nbPerChunk, ...)
-{
+#TODO: setRefClass... to avoid copy data !!
+#http://stackoverflow.com/questions/2603184/r-pass-by-reference
 
+#fields: data (can be NULL or provided by user), coeffs (will be computed
+#con can be a character string naming a file; see readLines()
+#data can be in DB format, on one column : TODO: guess (from header, or col. length...)
 
-#TODO: just a wrapper which calls ppam.exe (system("...")) and reads output (binary) file to retrieve medoids + IDs
-       #on input: can be data or con; data handled by writing it to file (ascii or bin ?!),
-       #con handled 
 
+writeTmp(curves [uncompressed coeffs, limited number - nbSeriesPerChunk], last=FALSE) #if last=TRUE, close the conn
+readTmp(..., from index, n curves) #careful: connection must remain open
+#TODO: write read/write tmp reference ( on file in .tmp/ folder ... )
 
-       #options for tmp files: in RAM, on disk, on DB (can be distributed)
+#data:
+#stop("Unrecognizable 'data' argument (must be numeric, functional or connection)")
 
+#WER: "end" to apply stage 2 after stage 1 iterated, "mix" (or anything else...?!) to apply it after every stage 1
+epclust = function(data, K, nbPerChunk, WER="end", ncores=NULL, writeTmp=ref_writeTmp, readTmp=ref_readTmp) #where to put/retrieve intermediate results; if not provided, use file on disk
+{
 
 
-       if (!is.null(data))
+       #on input: can be data or con; data handled by writing it to file (ascii or bin ?!),
+#data: con or matrix or DB
+
+       #1) acquire data (process curves, get as coeffs)
+       if (is.numeric(data))
        {
                #full data matrix
-               
-       } else if (!is.null(con))
+               index = 1
+               n = nrow(data)
+               while (index < n)
+               {
+                       writeTmp( getCoeffs(data) )
+                       index = index + nbSeriesPerChunk
+               }
+       } else if (is.function(data))
+       {
+               #custom user function to retrieve next n curves, probably to read from DB
+               writeTmp( getCoeffs( data(nbPerChunk) ) )
+       } else
        {
                #incremental connection
                #read it one by one and get coeffs until nbSeriesPerChunk
                #then launch a clustering task............
-       } else
-               stop("at least 'data' or 'con' argument must be present")
+               #TODO: find a better way to parse than using a temp file
+               ascii_lines = readLines(data, nbSeriesPerChunk)
+               seriesChunkFile = ".tmp/seriesChunk"
+               writeLines(ascii_lines, seriesChunkFile)
+               writeTmp( getCoeffs( read.csv(seriesChunkFile) ) )
+       }
+
+       library(parallel)
+       ncores = ifelse(is.numeric(ncores), ncores, parallel::detectCores())
+       cl = parallel::makeCluster(ncores)
+115     parallel::clusterExport(cl=cl, varlist=c("X", "Y", "K", "p"), envir=environment())
+116     li = parallel::parLapply(cl, 1:B, getParamsAtIndex)
 
+       #2) process coeffs (by nbSeriesPerChunk) and cluster in parallel (just launch async task, wait for them to complete, and re-do if necessary)
+       repeat
+       {
+               completed = rep(FALSE, ............)
+               #while there is jobs to do (i.e. size of tmp "file" is greater than nbSeriesPerChunk),
+               #A) determine which tasks which processor will do (OK)
+               #B) send each (sets of) tasks in parallel
+               #C) flush tmp file (current parallel processes will write in it)
+               #always check "complete" flag (array, as I did in MPI) to know if "slaves" finished
+       }
+
+parallel::stopCluster(cl)
+
+       #3) readTmp last results, apply PAM on it, and return medoids + identifiers
+
+       #4) apply stage 2 (in parallel ? inside task 2) ?)
+       if (WER == "end")
+       {
+               #from center curves, apply stage 2...
+       }
+}
+
+getCoeffs = function(series)
+{
+       #... return wavelets coeffs : compute in parallel !
 }