| 1 | #TODO: setRefClass... to avoid copy data !! |
| 2 | #http://stackoverflow.com/questions/2603184/r-pass-by-reference |
| 3 | |
| 4 | #fields: data (can be NULL or provided by user), coeffs (will be computed |
| 5 | #con can be a character string naming a file; see readLines() |
| 6 | #data can be in DB format, on one column : TODO: guess (from header, or col. length...) |
| 7 | epclust = function(data=NULL, con=NULL, raw=FALSE, K, nbPerChunk, ..., where_to_store_tmp_data, and how ?) |
| 8 | #options for tmp files: in RAM, on disk, on DB (can be distributed) |
| 9 | { |
| 10 | |
| 11 | |
| 12 | #on input: can be data or con; data handled by writing it to file (ascii or bin ?!), |
| 13 | |
| 14 | |
| 15 | if (!is.null(data)) |
| 16 | { |
| 17 | #full data matrix |
| 18 | index = 1 |
| 19 | n = nrow(data) |
| 20 | while (index < n) |
| 21 | { |
| 22 | getCoeffs(data |
| 23 | index = index + nbSeriesPerChunk |
| 24 | } |
| 25 | } else if (!is.null(con)) |
| 26 | { |
| 27 | #incremental connection |
| 28 | #read it one by one and get coeffs until nbSeriesPerChunk |
| 29 | #then launch a clustering task............ |
| 30 | readLines() |
| 31 | } else |
| 32 | stop("at least 'data' or 'con' argument must be present") |
| 33 | |
| 34 | } |
| 35 | |
| 36 | getCoeffs = function(series) |
| 37 | { |
| 38 | #... return wavelets coeffs : compute in parallel ! |
| 39 | } |