Re-expanding a compressed dataframe to include zero values in missing rows - r

Question

Given a dataset in the following form:
> Test
Pos Watson Crick Total
1 39023 0 0 0
2 39024 0 0 0
3 39025 0 0 0
4 39026 2 1 3
5 39027 0 0 0
6 39028 0 4 4
7 39029 0 0 0
8 39030 0 1 1
9 39031 0 0 0
10 39032 0 0 0
11 39033 0 0 0
12 39034 1 0 1
13 39035 0 0 0
14 39036 0 0 0
15 39037 3 0 3
16 39038 2 0 2
17 39039 0 0 0
18 39040 0 1 1
19 39041 0 0 0
20 39042 0 0 0
21 39043 0 0 0
22 39044 0 0 0
23 39045 0 0 0
I can compress these data to remove zero rows with the following code:
a=subset(Test, Total!=0)
> a
Pos Watson Crick Total
4 39026 2 1 3
6 39028 0 4 4
8 39030 0 1 1
12 39034 1 0 1
15 39037 3 0 3
16 39038 2 0 2
18 39040 0 1 1
How would I code the reverse transformation? i.e. To convert dataframe a back into the original form of Test.
More specifically: without any access to the original data, how would I re-expand the data (to include all sequential "Pos" rows) for an arbitrary range of Pos?
Here, the ID column is irrelevant. In a real example, the ID numbers are just row numbers created by R. In a real example, the compressed dataset will have sequential ID numbers.

Answer 1

Here's another possibility, using base R. Unless you explicitly provide the initial and the final value of Pos, the first and the last index value in the restored dataframe will correspond to the values given in the "compressed" dataframe a:
restored <- data.frame(Pos=(a$Pos[1]:a$Pos[nrow(a)])) # change range if required
restored <- merge(restored,a, all=TRUE)
restored[is.na(restored)] <- 0
#> restored
# Pos Watson Crick Total
#1 39026 2 1 3
#2 39027 0 0 0
#3 39028 0 4 4
#4 39029 0 0 0
#5 39030 0 1 1
#6 39031 0 0 0
#7 39032 0 0 0
#8 39033 0 0 0
#9 39034 1 0 1
#10 39035 0 0 0
#11 39036 0 0 0
#12 39037 3 0 3
#13 39038 2 0 2
#14 39039 0 0 0
#15 39040 0 1 1
Possibly the last step can be combined with the merge function by using the na.action option correctly, but I didn't find out how.

Answer 2

You need to know at least the Pos values you want to fill in. Then, it's a combination of join and mutate operations in dplyr.
Test <- read.table(text = "
Pos Watson Crick Total
1 39023 0 0 0
2 39024 0 0 0
3 39025 0 0 0
4 39026 2 1 3
5 39027 0 0 0
6 39028 0 4 4
7 39029 0 0 0
8 39030 0 1 1
9 39031 0 0 0
10 39032 0 0 0
11 39033 0 0 0
12 39034 1 0 1
13 39035 0 0 0
14 39036 0 0 0
15 39037 3 0 3
16 39038 2 0 2
17 39039 0 0 0
18 39040 0 1 1
19 39041 0 0 0
20 39042 0 0 0
21 39043 0 0 0
22 39044 0 0 0")
library(dplyr)
Nonzero <- Test %>% filter(Total > 0)
All_Pos <- Test %>% select(Pos)
Reconstruct <-
All_Pos %>%
left_join(Nonzero) %>%
mutate_each(funs(ifelse(is.na(.), 0, .)), Watson, Crick, Total)
In my code, All_Pos contains all valid positions as a one-column data frame; the mutate_each() call converts NA values to zeros. If you only know the largest MaxPos, you can construct it using
All_Pos <- data.frame(seq_len(MaxPos))