textstat_readability() returns nothing for an empty document

[kbenoit]

The question has arisen in #1969 of how to handle NA documents. What about empty string documents? Right now, textstat_readability() returns nothing if the document was "". This is not true for other textstat returns. We should make them consistent.

library("quanteda")
## Package version: 2.1.0

txt <- c(d1 = "The cat in the hat", d2 = "", d3 = "Once upon a time.")
corp <- corpus(txt)
toks <- tokens(corp)
dfmat <- dfm(toks)

textstat_readability(txt)
##   document  Flesch
## 1       d1 117.160
## 2       d3  97.025
textstat_readability(corp)
##   document  Flesch
## 1       d1 117.160
## 2       d3  97.025

textstat_dist(dfmat) %>%
  as.data.frame()
##   document1 document2 euclidean
## 1        d3        d1  3.464102
## 2        d1        d3  3.464102

textstat_entropy(dfmat)
##   document  entropy
## 1       d1 1.921928
## 2       d2 0.000000
## 3       d3 2.321928

textstat_lexdiv(toks)
##   document TTR
## 1       d1 0.8
## 2       d2 NaN
## 3       d3 1.0
textstat_lexdiv(dfmat)
##   document TTR
## 1       d1 0.8
## 2       d2 NaN
## 3       d3 1.0

Suggestion

Return NA for textstat_readability() if the document is zero.