Parsing of valid FR cell phone numbers fails in 0.6.47

[nicolasrouanne]

Some valid FR cell phone numbers are now considered invalid with release 0.6.47, while they were valid with 0.6.46

For instance

require 'faker'

phone_number =  Faker::Base.numerify('+33 6 ## ## ## ##') 
parsed_phone_number = Phonelib.parse(phone_number)
parsed_phone_number.valid?

Example numbers considered as invalid in 0.6.47 although they are valid in 0.6.46 (and they are indeed valid): +33696339246, +33692746428, +33639538676, +33639891873, +33692018258, etc.

[daddyz]

@nicolasrouanne Hi, I checked these numbers in the original libphonenumber here and number is possible but not valid there. You can open an issue for them

[ahubertaircall]

Hello, We are facing the same issue on our end, FYI I filled a report ticket on libphonenumber side: https://issuetracker.google.com/issues/179166703.

[nicolasrouanne]

Just to sum it up after libphonenumber answered the ticket:

• what seemed to be an issue in phonelib is not, phonelib is barely using libphonenumber for validation
• libphonenumber seems to be correctly validating French phone numbers; when they say it's invalid... it's that it is

  for instance+33696339246 which seems to have a valid french format, as in "+ 33 6 # ## ## ## ##", is actually invalid as it overlaps with Reunion number

I was generating phones using this pattern Faker::Base.numerify('+33 6 ## ## ## ##'), which actually generates "possible" formats (i.e. with the correct number pattern), but not always "valid" phone numbers. Phone number validation is more complex than just a regex.

I then switched to Faker::PhoneNumber.cell_phone, which seems promising generating valid phone numbers. However, there seems to be an issue in faker's fr locale for valid french phone numbers ATM: https://github.com/faker-ruby/faker/issues/1650, https://github.com/faker-ruby/faker/issues/1957. May be fixed by https://github.com/faker-ruby/faker/pull/1968, but it's not merged at the time of writing.

I'm closing this issue, as it's not a bug, rather a failure on my side to generate random valid test data.