R typecast NA vs string - r

Question

I had a problem with some of my code and I fixed it but don't fully understand why the error was an error
The code looked like:
for(i in 1:3){df = rbind.fill(z, data.frame(id=i,
data=if(is.null(x$results[[i]]$synopsis$data))
{NA}else{x$results[[i]]$synopsis$data}))}
The issue I had was, if the first data value was indeed null, I would get NA but then for the second and third I would either get another NA or if there was data I wouldn't get it, I would get 1.
If the first value was data, then I would get the data and for the other two I would either get NA or the correct data.
I'm not a computer scientist but a dev who sits near me (but doesnt know R) suggested it was something to do with different typecasts of NA and a string. To solve the issue I changed the NA to "0" (I suppose "NA" would work too).
I'd just like a more thorough explanation into what was happening. My layman's understanding is if NA was the first result, then every result is in that "format" where something is either NA or not and not is handled as 1 which is kinda of like a Boolean response?
Example:
my.list <- list(list(),structure(
list(
experience = structure(
list(
start = "Hi"
),.Names = c("start")),
`_meta` = structure(
list(weight = 1L, `_sources` = list(structure(
list(`_origin` = "a"), .Names = "_origin"
))),.Names = c("weight", "_sources"))),.Names = c("experience", "_meta")))
my.list[[1]]$experience$start
NULL
my.list[[2]]$experience$start
[1] "Hi"
df <- NULL
for(i in 1:2){df = rbind.fill(df, data.frame(id=i,
data=if(is.null(my.list[[i]]$experience$start))
{NA}else{my.list[[i]]$experience$start}))}
Then
df2 <- NULL
for(i in 1:2){df2 = rbind.fill(df2, data.frame(id=i,
data=if(is.null(my.list[[i]]$experience$start))
{"NA"}else{my.list[[i]]$experience$start}))}
Results:
df: df2:
id data id data
1 NA 1 NA
2 1 2 HI

Answer 1

Olivia, thanks for clarifications.
You are nearly there. As you loop, indeed the first iteration will determine the class of the column data of your output data.frame df.
In scenario 1, you can have a better idea by going through the loop step by step:
df <- NULL
i=1
df = rbind.fill(df, data.frame(id=i,
data=if(is.null(my.list[[i]]]$experience$start)) {NA}
else{my.list[[i]]$experience$start}))
df
id data
1 1 NA
Then, have a look at class of df$data
class(df$data)
[1] "logical"
Which is derived from: mode(NA) (logical).
As an alternative way, when you store data related in your set of experiments in a list, you should try to use a "R-ish" way to manipulate this list.
For instance, you can try:
sapply(my.list, FUN=function(element)element$experience$start)
[[1]]
NULL
[[2]]
[1] "Hi"
Which highlights that you tries to gather together sets of incompatibles contents: simplification can't go simpler than this list -- if you unlist you would dismiss this meaningful NULL