'update'

[epclust.git] / data / preprocessing / convert_32kEDF.R

diff --git a/data/preprocessing/convert_32kEDF.R b/data/preprocessing/convert_32kEDF.R
new file mode 100644 (file)
index 0000000..2e6798a
--- /dev/null
+++ b/data/preprocessing/convert_32kEDF.R
@@ -0,0 +1,24 @@
+convert_32kEDF = function(orig_csv, nb_series_per_chunk)
+{
+       datetimes = #...TODO: all 3 years? year-by-year is better
+       orig_con = file(orig_csv, open="r") #2009, 2010 or 2011
+       ignored = readLines(orig_con, 1) #skip header
+       serie_length = length(datetimes) #around 365*24*2 = 17520
+       sep = if (year==2009) "," else if (year==2010) ";" else ";"
+
+scan(orig_con, character(), sep=",", nlines=1, quiet=TRUE)
+       library(sqldf, quietly=TRUE)
+       ids = read.csv.sql(file_csv, header = TRUE, sep = ","
+               sql = "select * from file_csv group by FK_CCU_ID")
+       index = 0
+       repeat
+       {
+               if (index+1 >= length(ids))
+                       break
+               request = "select CPP_DATE_PUISSANCE,CPP_PUISSANCE_BRUTE where FK_CCU_ID in ("
+               for (id in ids[index + seq_len(nb_series_per_chunk)])
+                       request = paste(request, id, ",", sep="")
+               request = paste(request, ") order by FK_CCU_ID,CPP_DATE_PUISSANCE", sep="")
+               series_chunk = read.csv.sql(file_csv, header = TRUE, sep = ",", sql = request)
+               
+               index = index + 17520