bug with AND and OR operators in regex?

[von_olaf]

I am struggling to write the correct regex pattern to match the following condition

(contains the word other) OR (contains both us AND car)

This code works as expected:

    str_detect(c('us cars',
                 'u.s. cars',
                 'us and bikes',
                 'other'),
               regex('other|((?=.*us)(?=.*car))',
                     ignore_case = TRUE))
    [1]  TRUE FALSE FALSE  TRUE

However, if I try to include variations of us (united states) such as u.s. and u.s then the pattern does not work anymore.

    str_detect(c('us cars',
                 'u.s. cars',
                 'us and bikes',
                 'other'),
               regex('other|((?=.*us)(?=.*u.s.)(?=.*u.s)(?=.*car))',
                     ignore_case = TRUE))
    [1] FALSE FALSE FALSE  TRUE

What is the issue here? Why is the AND pattern working in the first case but not in the second?
Thanks!

[andresrcs]

The problem is with your use of the ".", this special character means any character on regex if you want a literal dot you have to escape it with \\.

library(stringr)
str_detect(c('us cars',
             'u.s. cars',
             'us and bikes',
             'other'),
           regex('other|((?=.*u\\.*s\\.*)(?=.*car))',
                 ignore_case = TRUE))
#> [1]  TRUE  TRUE FALSE  TRUE

^{Created on 2019-02-12 by the reprex package (v0.2.1)}

[von_olaf]

thanks but then I am confused. isnt . a character as well? it should match with the regex dot (any character) right?

[von_olaf]

for instance, here I am escaping and still does not work

> str_detect(c('us cars',
+              'u.s. cars',
+              'us and bikes',
+              'other'),
+            regex('other|((?=.*us)(?=.*u\\.s\\.)(?=.*u\\.s)(?=.*car).*)',
+                  ignore_case = TRUE))
[1] FALSE FALSE FALSE  TRUE

[andresrcs]

Nope, because you are just tellin any character but not how many times, .* this means any character 0 or more times.

[von_olaf]

this works.

str_detect(c('us cars',
             'u.s. cars',
             'us and bikes',
             'other'),
           regex('other|((?=.*u\\.s\\.)(?=.*car).*)',
                 ignore_case = TRUE))

It seems the issue is having more than one AND. Am I completely crazy here?!

[andresrcs]

The idea of a regular expression is to define a pattern not each variation of the text as you are trying to do.

[von_olaf]

absolutely. but on purely logical grounds, I dont get why combining four logical conditions with AND does not work while it works only with two. Perhaps there is a limit in the number of lookaheads that can be done?

[andresrcs]

You are getting the logic wrong, none of your words match the pattern you where specifying
((?=.*us)(?=.*u\\.s\\.)(?=.*u\\.s)(?=.*car))
this string would match that US U.S. U.S car do you see the reason? you are using AND not OR

[von_olaf]

damn you are right... time to go to bed I guess.... Thanks for helping out!!