From: Benjamin Auder <benjamin.auder@somewhere>
Date: Fri, 6 Jan 2017 11:33:55 +0000 (+0100)
Subject: progress on main.R
X-Git-Url: https://git.auder.net/%7B%7B%20asset%28%27mixstore/css/current/%3C?a=commitdiff_plain;h=3d06151562906a7ff2efafe18ba1e742a122ce72;p=epclust.git

progress on main.R
---

diff --git a/code/draft_R_pkg/R/main.R b/code/draft_R_pkg/R/main.R
index 3411720..a01385e 100644
--- a/code/draft_R_pkg/R/main.R
+++ b/code/draft_R_pkg/R/main.R
@@ -4,33 +4,50 @@
 #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...)
-epclust = function(data=NULL, con=NULL, raw=FALSE, K, nbPerChunk, ..., where_to_store_tmp_data, and how ?)
-#options for tmp files: in RAM, on disk, on DB (can be distributed)
+
+
+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 ... )
+
+epclust = function(data=NULL, K, nbPerChunk, ..., writeTmp=ref_writeTmp, readTmp=ref_readTmp) #where to put/retrieve intermediate results; if not provided, use file on disk
 {
 
 
 	#on input: can be data or con; data handled by writing it to file (ascii or bin ?!),
+#data: con or matrix or DB
 
-
-	if (!is.null(data))
+	#1) acquire data (process curves, get as coeffs)
+	if (is.numeric(data))
 	{
 		#full data matrix
 		index = 1
 		n = nrow(data)
 		while (index < n)
 		{
-			getCoeffs(data
+			writeTmp( getCoeffs(data) )
 			index = index + nbSeriesPerChunk
 		}
-	} else if (!is.null(con))
+	} 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............
-		readLines()
+		ascii_lines = readLines(data, nbSeriesPerChunk)
+		seriesChunkFile = ".tmp/seriesChunk" #TODO: find a better way
+		writeLines(ascii_lines, seriesChunkFile)
+		writeTmp( getCoeffs( read.csv(seriesChunkFile) ) )
 	} else
-		stop("at least 'data' or 'con' argument must be present")
+		stop("Unrecognizable 'data' argument (must be numeric, functional or connection)")
+
+	#2) process coeffs (by nbSeriesPerChunk) and cluster in parallel (just launch async task, wait for them to complete, and re-do if necessary)
+
 
+	#3) apply stage 2 (in parallel ? inside task 2) ?)
 }
 
 getCoeffs = function(series)