R: Coding looping constructs - r

Question

I am looking for some help on achieving below looping requirement through a simplified indexing or apply constructs in R. Doing it by 'for' loops seem computationally complex and inefficient. Hence, I am looking for any help to achieve it in an efficient manner;
The data table reference is as below;
The sequence I am trying to get is all positive and negative number index (row and column) sequences per row as below;
For row 1: 1-4-5, 1-4-8, 1-4-11;
Column 'Occurances' specifies the potential number of sequences per row.
Finally, I am trying to get a data frame similar to below (shown only for first and second rows) with all occurrences with each index on a column;
Any help is highly appreciated. Thank you very much

Answer 1

There's a lot of way you can do this. To do it in an efficient manner you should probably use base R. The more rows and columns you have to check, the more you will need to be careful with how you code this.
Here are two examples of how you could run it, see which works best for you.
library(purrr)
library(dplyr)
# create table to test code on, n1 x n2 dataframe with a random sample of -1, 0, 1
n1 <- 10
n2 <- 10
to_test <- map(1:n1, ~sample(c(-1, 0, 1), size = c(n2), replace = T)) %>%
`names<-`(seq_along(.)) %>%
bind_cols()
# Split table into a list of rows
to_test_row_list <- split(to_test, 1:nrow(to_test))
# For each item in the list
sub_tables <- mapply(FUN = function(list_in, row_in){
# create a dataframe with the ron number in the first row
crossing(row = row_in,
# cross join the indexes of the columns with are less than and
# more than zero for the other two cols
crossing(data.frame(gt = which(list_in > 0)),
data.frame(lt = which(list_in < 0))))},
# Inputs for the mapply function FUN, the list of rows and the number for each row
list_in = to_test_row_list,
row_in = names(to_test_row_list),
# Do not simply dataframes into lists
SIMPLIFY = F)
# Turn list of tables into one long table
res1 <- bind_rows(sub_tables)
res1
# The same code in one pipe
res2 <- to_test %>%
split(seq_along(.)) %>%
map2(.x = .,
.y = names(.),
~crossing (data.frame(gt = which(.x > 0)),
data.frame(lt = which(.x < 0))) %>%
mutate(row = .y) %>% select(row, everything())) %>%
bind_rows()
res2

Answer 2

This works in Base-R, the task is accomplished almost entirely in the first line of code. The rest is just cleaning the output to make it exactly as asked for. Without proper example data (you can use dput(...) to share) there will certainly be issues with using this code exactly as presented with your data.
new_data <- do.call(rbind,apply(mydata,1, function(x) merge(x[x > 0], x[x < 0]) ))
new_data$from <- sub("X(\\d).*","\\1",row.names(new_data))
new_data <- new_data[,c(3,1,2)]
rownames(new_data) <- c()
sample data:
mydata <- data.frame("1"=c(0,0,0,-45,57,0,0,51,0,0,45,0),"3"=c(4,4,0,5,654,34,-6,65,-37,4,56,56))
mydata <- t(mydata)
output:
> new_data
from x y
1 1 57 -45
2 1 51 -45
3 1 45 -45
4 3 4 -6
5 3 4 -6
6 3 5 -6
7 3 654 -6
8 3 34 -6
9 3 65 -6
10 3 4 -6
11 3 56 -6
12 3 56 -6
13 3 4 -37
14 3 4 -37
15 3 5 -37
16 3 654 -37
17 3 34 -37
18 3 65 -37
19 3 4 -37
20 3 56 -37
21 3 56 -37

Answer 3

Here's a tidyverse approach if you want to keep things nested neatly:
library(tidyverse)
df <- tibble::tribble(
~`1`, ~`2`, ~`3`, ~`4`, ~`5`, ~`6`, ~`7`, ~`8`, ~`9`, ~`10`, ~`11`, ~`12`,
0, 0, 0L, -45.2, 57, 0, 0, 82.7, 0, 0, 58.7, 0,
48.8, 65, 0L, 35.5, 50.8, 42.2, -89.6, 52.8, -45.8, 26.4, 51.1, 85.7,
63.1, 83.3, 0L, 21.5, 60, 0, 0, 69, 0, -84.3, 61, 0
)
df %>%
rownames_to_column(var = "row_idx") %>%
pivot_longer(cols = -row_idx, names_to = "col_idx") %>%
group_by(row_idx) %>%
nest() %>%
mutate(
df_of_pairs = map(data, ~ expand.grid(which(.$value < 0), which(.$value > 0))),
combos = map_int(df_of_pairs, nrow)
)
#> # A tibble: 3 x 4
#> # Groups: row_idx [3]
#> row_idx data df_of_pairs combos
#> <chr> <list> <list> <int>
#> 1 1 <tibble [12 x 2]> <df[,2] [3 x 2]> 3
#> 2 2 <tibble [12 x 2]> <df[,2] [18 x 2]> 18
#> 3 3 <tibble [12 x 2]> <df[,2] [6 x 2]> 6
Created on 2020-05-13 by the reprex package (v0.3.0)
Then if you want to get the list of pairs, simply add %>% unnest(df_of_pairs) to the end of the pipeline:
df %>%
rownames_to_column(var = "row_idx") %>%
pivot_longer(cols = -row_idx, names_to = "col_idx") %>%
group_by(row_idx) %>%
nest() %>%
mutate(
df_of_pairs = map(data, ~ expand.grid(which(.$value < 0), which(.$value > 0))),
combos = map_int(df_of_pairs, nrow)
) %>%
unnest(df_of_pairs)
# A tibble: 27 x 5
# Groups: row_idx [3]
row_idx data Var1 Var2 combos
<chr> <list> <int> <int> <int>
1 1 <tibble [12 x 2]> 4 5 3
2 1 <tibble [12 x 2]> 4 8 3
3 1 <tibble [12 x 2]> 4 11 3
4 2 <tibble [12 x 2]> 7 1 18
5 2 <tibble [12 x 2]> 9 1 18
6 2 <tibble [12 x 2]> 7 2 18
7 2 <tibble [12 x 2]> 9 2 18
8 2 <tibble [12 x 2]> 7 4 18
9 2 <tibble [12 x 2]> 9 4 18
10 2 <tibble [12 x 2]> 7 5 18
# ... with 17 more rows