Break string into several columns using tidyr::extract regex

In the regex used, we are matchng one more more punctuation characters ([[:punct:]]+) i.e. @ followed by capturing the numeric part ((\\d+) - this will be our first column of interest), followed by one or more white-space (\\s+), followed by the second capture group (\\S+ - one or more non white-space character i.e. "ANO_CENSO" for the first row), followed by space (\\s+), then we capture the third group (([[:alum:]$]+) - i.e. one or more characters that include the alpha numeric along with $ so as to match $Char1), next we match one or more characters that are not a letter ([^A-Za-z]+- this should get rid of the space and *) and the last part we capture one or more characters that are not * (([^*]+).

sasdic %>% 
      extract(a, into=c('int_pos', 'var_name', 'x', 'label'),
   "[[:punct:]](\\d+)\\s+(\\S+)\\s+([[:alnum:]$]+)[^A-Za-z]+([^*]+)")

#  int_pos           var_name      x                 label
#1       1          ANO_CENSO      5          Ano do Censo
#2      71            TP_SEXO $Char1                  Sexo
#3      72        TP_COR_RACA $Char1              Cor/raça
#4      74 FK_COD_PAIS_ORIGEM      4 Código País de origem

This is another option, though it uses the data.table package instead of tidyr:

library(data.table)
setDT(sasdic)

# split label
sasdic[, c("V1","label") := tstrsplit(a, "/\\*|\\*/")]                   
# remove leading "@", split remaining parts
sasdic[, c("int_pos","var_name","x") := tstrsplit(gsub("^@","",V1)," +")]
# remove unneeded columns
sasdic[, c("a","V1") := NULL]                                            

sasdic

#                    label int_pos           var_name       x
# 1:          Ano do Censo       1          ANO_CENSO      5.
# 2:                  Sexo      71            TP_SEXO $Char1.
# 3:              Cor/raça      72        TP_COR_RACA $Char1.
# 4: Código País de origem      74 FK_COD_PAIS_ORIGEM      4.

This assumes that the "remaining parts" (aside from the label) are space-separated.

---

This could also be done in one block (which is what I would do):

sasdic[, c("a","label","int_pos","var_name","x") := {
  x   = tstrsplit(a, "/\\*|\\*/")
  x1s = tstrsplit(gsub("^@","",x[[1]])," +")
  c(list(NULL), x1s, x[2])
}]