填充矩阵的更有效方法

Question

I am trying to cbind a very large matrix with a data frame I have and I am running into memory issues due to the size of the matrix.

I have data:

set.seed(123)
df1 <- data.frame(replicate(5, sample(1:20, 10, rep=TRUE)))
colnames(df1) <- c("col1", "col2", "col3", "col4", "important_col")
df2 <- data.frame(replicate(20, sample(0:0, nrow(df1), rep=TRUE)))
colnames(df2) <- gsub("X", "", colnames(df2))
df_fin <- cbind(df1, df2)

The following Works and does as I want on a small sample but when applied to rows of hundreds of thousands and columns of 1000 + I have the memory issues.

vecp <- colnames(df2)

imp_col <- df1$important_col

matrix <-  matrix(vecp, byrow = TRUE,
                           nrow = length(imp_col),
                           ncol = length(vecp),
                           dimnames = list(1:length(imp_col), vecp))

d <- ifelse(matrix == imp_col, 1, 0)


df_fin <- cbind(df1, d)

Where I am trying to make the code more efficient (is where the I have memory issues) at line d <- ifelse(matrix == imp_col, 1, 0) .

Is there a way I can make the matrix a sparse matrix before I apply the ifesle statement.

I build a matrix like the following:

   col1 col2 col3 col4 important_col 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
1    11   14    3   11             1 0 0 0 0 0 0 0 0 0  0  0  0  0  0  0  0  0  0  0  0
2     1    1   19   15             4 0 0 0 0 0 0 0 0 0  0  0  0  0  0  0  0  0  0  0  0
3     3   17   10   10             6 0 0 0 0 0 0 0 0 0  0  0  0  0  0  0  0  0  0  0  0
4    13   10    8   17            10 0 0 0 0 0 0 0 0 0  0  0  0  0  0  0  0  0  0  0  0
5    18    5    3   18            19 0 0 0 0 0 0 0 0 0  0  0  0  0  0  0  0  0  0  0  0
6    11   10    9    5            17 0 0 0 0 0 0 0 0 0  0  0  0  0  0  0  0  0  0  0  0
7     5   11   18   16            17 0 0 0 0 0 0 0 0 0  0  0  0  0  0  0  0  0  0  0  0
8     5    8   13    8             6 0 0 0 0 0 0 0 0 0  0  0  0  0  0  0  0  0  0  0  0
9    10    1    7   16            12 0 0 0 0 0 0 0 0 0  0  0  0  0  0  0  0  0  0  0  0
10    4   17   17    3             4 0 0 0 0 0 0 0 0 0  0  0  0  0  0  0  0  0  0  0  0

The end product is like:

   col1 col2 col3 col4 important_col 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
1     6   20   18   20             3 0 0 1 0 0 0 0 0 0  0  0  0  0  0  0  0  0  0  0  0
2    16   10   14   19             9 0 0 0 0 0 0 0 0 1  0  0  0  0  0  0  0  0  0  0  0
3     9   14   13   14             9 0 0 0 0 0 0 0 0 1  0  0  0  0  0  0  0  0  0  0  0
4    18   12   20   16             8 0 0 0 0 0 0 0 1 0  0  0  0  0  0  0  0  0  0  0  0
5    19    3   14    1             4 0 0 0 1 0 0 0 0 0  0  0  0  0  0  0  0  0  0  0  0
6     1   18   15   10             3 0 0 1 0 0 0 0 0 0  0  0  0  0  0  0  0  0  0  0  0
7    11    5   11   16             5 0 0 0 0 1 0 0 0 0  0  0  0  0  0  0  0  0  0  0  0
8    18    1   12    5            10 0 0 0 0 0 0 0 0 0  1  0  0  0  0  0  0  0  0  0  0
9    12    7    6    7             6 0 0 0 0 0 1 0 0 0  0  0  0  0  0  0  0  0  0  0  0
10   10   20    3    5            18 0 0 0 0 0 0 0 0 0  0  0  0  0  0  0  0  0  1  0  0

Which I Will then make into a sparse matrix.

Answer 1

The problem is that d is the same size as your matrix, so if your matrix is huge then you'll have two of them. One posible option (although probably slower) is to iterate through the columns and change them one at a time, this only creates objects the same size as one column of your matrix. You could give this a try:

for (i in 1:ncol(matrix)) matrix[, i] <- matrix[, i] == imp_col

The expression returns a boolean but if your matrix is made of integers then they will be converted to 0 and 1.