CODEWITHSUNDEEP
rhash
wdkrnls
wdkrnls

Reputation: 4692

Make a lookup table from a data.frame

I have a data.frame which has only one unique non-NA value in all columns but one, which only has NA. 

data <- data.frame(A = c("egg", "egg"), B = c(NA, "bacon"), C = c("ham", "ham"), D = c(NA, NA))

How can I use it to create a lookup table of the form below?

lookup <- make_lookup(key=unique_values(data), value=names(data))
lookup[["egg"]] # returns "A"
lookup[["bacon"]] # returns "B"
lookup[["ham"]] # returns "C"
lookup[["NA"]] # returns "D"

EDIT

Based on Frank's answer below, I'm trying to have my lookup table refer to multiple values.

keys <- lapply(data, function(x) if(is.factor(x)) levels(x) else "bacon")
vals <- names(data)

keys

$A
[1] "egg"

$B
[1] "bacon"

$C
[1] "ham"

$D
[1] "bacon"

vals

[1] "A" "B" "C" "D"

tapply(vals, keys, c)

Error in tapply(vals, keys, c) : arguments must have same length

Upvotes: 3

Views: 2189

Answers (1)

Frank
Frank

Reputation: 66819

Here is one way. The lookup is a vector:

keys <- sapply(data,function(x)if(is.factor(x))levels(x)else "NA")
vals <- names(data)

lookup <- setNames(vals,keys)

I've replace NA with "NA" since I couldn't figure out how to use the former.

The syntax lookup[["egg"]] works, but also lookup["egg"]. The reverse lookup is rlookup <- keys, accessible the same way: rlookup["A"].

For keys with multiple values. If the keys may map to a vector of values, use

lookup <- tapply(vals,keys,c)

Try this out with keys <- sapply(data,function(x)if(is.factor(x))levels(x)else "bacon") and vals as above, for example (as in the OP's comment, below). Now the lookup is a list and so can only be accessed with double brackets: lookup[["bacon"]]. The reverse lookup works as before.

For general column classes. If the columns of data are not all factors, the if/else conditions will need to be modified or generalized. Here is a version of @akrun's generalized solution from the comments:

keys <- sapply(data,function(x)c(unique(as.character(x)[!is.na(x)]),"NA")[1])

Upvotes: 5

Related Questions