Unconsistent behavior in `word` function

[pgrandinetti]

I don't get if the following behavior of the word function is desired, or is a bug

> words = "this is a very long sentence to test"
> word(words, start=-4, end=-1)  # last 4 words, OK
[1] "long sentence to test"
> word(words, start=-8, end=-1)  # OK
[1] "this is a very long sentence to test"
> word(words, start=-9, end=-1)  # OK, imho
[1] NA
> word(words, start=-10, end=-1)  # what is this?!
[1] "is a very long sentence to test" "a very long sentence to test"    "very long sentence to test"     
[4] "long sentence to test"           "sentence to test"                "to test"                        
[7] "test"    

[hadley]

Could you please rework your reproducible example to use the reprex package ? That makes it easier to see both the input and the output, formatted in such a way that I can easily re-run in a local session.

[pgrandinetti]

I guess we just need to clarify what is the behavior of the word function when the input argument start is negative and is greater than the sentence length.

From your doc:

start: integer vector giving position of first word to extract. Defaults to first word. If negative, counts backwards from last character.

But the behavior is not always clear. Sometimes I get a list as result.

If you copy this and run reprex() should give what you asked for (assuming that I understood what you meant!)

library(stringr)
library(reprex)
sentence = "this is a very long sentence to test"
word(sentence, start=-8, end=-1)  # OK
word(sentence, start=-9, end=-1)  # OK, imho because 9 > length
word(sentence, start=-10, end=-1)  # what is this, a vector?!

[hadley]

Please include the output of using reprex in the issue so I can see both the input and the output.

[pgrandinetti]

library(stringr)
library(reprex)
sentence = "this is a very long sentence to test"
word(sentence, start=-8, end=-1)  # OK
#> [1] "this is a very long sentence to test"
word(sentence, start=-9, end=-1)  # OK, imho because 9 > length
#> [1] NA
word(sentence, start=-10, end=-1)  # what is this, a vector?!
#> [1] "is a very long sentence to test" "a very long sentence to test"   
#> [3] "very long sentence to test"      "long sentence to test"          
#> [5] "sentence to test"                "to test"                        
#> [7] "test"

[pdelboca]

I've been testing this bug and the problem is that the formula it uses to convert negative values into actual positions doesn't work when the negative number used in since parameter is bigger (in absolute value) than the len of words in the sentence. (It still returns a negative start)

start[neg_start] <- start[neg_start] + len[neg_start] + 1L

As start is then used for indexing a matrix this causes the behavior reported by @pgrandinetti.

One way to solve the problem is, after converting the negative values, replace all the start values lower than 1 to 1. (Currently replaces only indexes past end with NA)

This will fix the issue and give consistent api with str_sub that returns all the string when abs(-since) > len(string).

If this solution is ok I could implement it and make a PR.

[hadley]

Sounds good to me!