Remove only initial NA values from data.table-CodePudding

In the below data.table, I want to remove all the NA values till the time at least two consecutive numeric values are encountered in column y. As per this condition, all rows from 1 to 9 should be removed from the table below.

Since I am cleaning a large number of groups, it will be great if the solution considers the performance part.

temp_dt = structure(list(group = c("B", "B", "B", "B", "B", "B", "B", "B", 
"B", "B", "B", "B", "B", "B", "B", "B", "B", "B", "B", "B", "B", 
"B", "B", "B", "B", "B", "B", "B", "B", "B"), x = c(940, 960, 
980, 1000, 1040, 1060, 1080, 1100, 1100, 1120, 1140, 1160, 1180, 
1190, 1200, 1200, 1220, 1240, 1260, 1280, 1300, 1300, 1320, 1340, 
1360, 1380, 1400, 1400, 1410, 1420), y = c(NA, NA, NA, NA, NA, 
NA, NA, 0.525, NA, 2.425, 2.425, NA, NA, NA, 2.425, NA, 2.425, 
1.975, 2.45, 2.45, 1.2, NA, NA, 2.275, 2.375, 2.75, 2.675, NA, 
2.75, 3.05)), row.names = c(NA, -30L), class = c("data.table", 
"data.frame"))

    group    x     y
 1:     B  940    NA
 2:     B  960    NA
 3:     B  980    NA
 4:     B 1000    NA
 5:     B 1040    NA
 6:     B 1060    NA
 7:     B 1080    NA
 8:     B 1100 0.525
 9:     B 1100    NA
10:     B 1120 2.425
11:     B 1140 2.425
12:     B 1160    NA
13:     B 1180    NA
14:     B 1190    NA
15:     B 1200 2.425
16:     B 1200    NA
17:     B 1220 2.425
18:     B 1240 1.975
19:     B 1260 2.450
20:     B 1280 2.450
21:     B 1300 1.200
22:     B 1300    NA
23:     B 1320    NA
24:     B 1340 2.275
25:     B 1360 2.375
26:     B 1380 2.750
27:     B 1400 2.675
28:     B 1400    NA
29:     B 1410 2.750
30:     B 1420 3.050
    group    x     y

I think the solution starts with making group of NA values but don't know how to proceed further -

temp_dt[, na_grp := rleid(is.na(y)), by = group]

Following code partially solves my problem, the only part not solved is of two consecutive values -

temp_dt[!is.na(y) & !na_grp %in% c(1L)]

CodePudding user response：

You could use shift to test two consecutive values, and cumsum to hold the selection once initial trigger condition is met:

temp_dt[,.SD[cumsum(!is.na(y)&!is.na(shift(y,-1)))>0],by=group]

    group    x     y
 1:     B 1120 2.425
 2:     B 1140 2.425
 3:     B 1160    NA
 4:     B 1180    NA
 5:     B 1190    NA
 6:     B 1200 2.425
 7:     B 1200    NA
 8:     B 1220 2.425
 9:     B 1240 1.975
10:     B 1260 2.450
11:     B 1280 2.450
12:     B 1300 1.200
13:     B 1300    NA
14:     B 1320    NA
15:     B 1340 2.275
16:     B 1360 2.375
17:     B 1380 2.750
18:     B 1400 2.675
19:     B 1400    NA
20:     B 1410 2.750
21:     B 1420 3.050