phonenumbers

Merge to upstream commit d56bf9da349b; relevant code changes:

• Support Russian extension character "доб" as a valid one while parsing the numbers.

Just metadata changes.

Merge to upstream commit 7e4e754bda09; relevant code changes:

• Improve parsing logic to be smarter about national-prefix detection & stripping based on possible-lengths (IS_POSSIBLE_LOCAL_ONLY and INVALID_LENGTH). Enables e.g. adding Iran short-codes starting with "096" without the need to hack IR's national prefix parsing config.

Merge to upstream commit fb9aa53ecfa3; relevant code changes:

• Documentation fix for number_type

Merge to upstream commit 1ad92eb35a44; relevant code changes:

• New method supported_calling_codes() API to return all the calling codes that the library considers valid, both for geographical and non-geographical entities.
• Added is_sms_service_for_region(numobj, region_dialing_from) API in short number info library. An SMS service is where the primary or only intended usage is to receive and/or send text messages (SMSs). This includes MMS as MMS numbers downgrade to SMS if the other party isn't MMS-capable. The is_sms_service metadata is also serialized for the first time.
• Documentation update for private variables _VALID_PUNCTUATION and _SINGLE_INTERNATIONAL_PREFIX, also renaming the latter from _UNIQUE_INTERNATIONAL_PREFIX.

Merge to upstream commit 242a186f1fbf; relevant code changes:

• Removing leading_zero_possible from the metadata and all the places it is referenced in the build and prod code. Will be removed from the metadata proto itself in a subsequent release. This should not affect users of the library - the only place it was used was format_in_original_format, and only initially to try and avoid modifying the input number by removing/adding digits inadvertently. Now this is checked at the end of the method anyway. However slight formatting differences with this method on invalid numbers starting with 0s may be noticed in some countries.
• Updated the documentation for the is_number_geographical API.
• Small comment improvements for parse() method to point users at keep_raw_input parameter.
• Added a new enum for CountryCodeSource called UNSPECIFIED. This is used as a default value, and will be returned if someone calls parse(keep_raw_input=False) and then .country_code_sourc on the result. If users want an actual value for this then they should call parse(keep_raw_input=True) instead; the values that were previously returned after calling this method will not change.
• Deletion of the possible number pattern in the phonemetadata.proto file and all generated code. This has not been used for a long time.

Merge to upstream commit 52cff9d8837f; relevant code changes:

• Add can_be_internationally_dialled public API; This was already in JS.

Merge to upstream commit 9923d9211432; relevant code changes:

• Small fix for possible out-of-bounds exception on RFC3966 input where no phone context was actually provided.

Merge to upstream commit 02dbc0921cf5; relevant code changes:

• Changing is_possible_with_reason to return the enums INVALID_LENGTH and IS_POSSIBLE_LOCAL_ONLY, where these apply. is_possible_number continues to consider IS_POSSIBLE or IS_POSSIBLE_LOCAL_ONLY numbers as possible to dial. Announcement: https://groups.google.com/d/msg/libphonenumber-discuss/sPhYzdzFCmg/6tYsS1f6DgAJ
• Doc updates for region_code_for_number to clarify that it does not work for short-codes or invalid numbers.

Merge to upstream commit 77affd08f65f; relevant code changes:

• Doc changes to update references to ISO country codes to CLDR region codes, which are what we actually use. Notice in some of the mapper files the country code is still the ISO one, because it's used as part of a description of a language only (e.g. zh-TW vs zh-CN). Edited the language comments there since they can be two or three letters (c.f. fil for filipino). Also edited the comment for leadingDigits at the territory level in the phonemetadata.proto file.
• New API methods: is_possible_number_for_type and is_possible_number_for_type_with_reason, along with supported_types_for_region and supported_types_for_non_geo_entity. These allow you to query which types (e.g. Mobile) exist for a particular region, and work out if a number is possible for that type (this is a simple length check) rather than for the region as a whole.

Merge to upstream commit 46a38545ba01; relevant code changes:

• Doc fix for geocoder.py to explain the cases where an empty string might be returned.

Metadata changes only.

Merge to upstream commit 5507fdbf623f; relevant code changes:

• Making national_significant_number more robust in the face of malicious input. This now ignores the number_of_leading_zeros if they are less than zero.

Merge to upstream commit ad297a10ba19; relevant code changes:

• Added two new enum values to ValidationResult - IS_POSSIBLE_LOCAL_ONLY and INVALID_LENGTH. Added more documentation to the existing values; see the docstrings for when these are going to be used. Note that the API for is_possible_number_with_reason has not yet been changed to return these values. IS_POSSIBLE_LOCAL_ONLY will be returned for some values which currently return IS_POSSIBLE, and INVALID_LENGTH will be returned for some values which currently return TOO_LONG.
• Fix for is_number_match to ignore the number_of_leading_zeros field when comparing numbers unless italian_leading_zero is True, and to consider default values to match the same value when explicitly set for these two fields. This fix shouldn't be needed for anyone correctly creating phone numbers using "parse" as recommended.

Merge to upstream commit 3b16d6b06497; relevant code changes:

• Exported normalize_diallable_chars_only. This method is already public in the C++ implementation. It has also now been added to the Javascript implementation.

Merge to upstream commit c210dbca4e93; relevant code changes:

• Introduced new is_carrier_specific_for_region API in shortnumberinfo. This determines whether the provided short code is carrier-specific or not when dialed from the given region.

Merge to upstream commit 1eb06f31e1dd; relevant code changes:

• Removing the ability for .._for_region methods in shortnumberinfo.py to work on strings, as well of phone number objects. These have been marked deprecated for months. Any users of these methods should call phonenumbers.parse first to create a PhoneNumber object, and pass this in.
• Support semicolon as extension character while parsing phone numbers. This is not applicable when you are trying to find the phone numbers.

Merge to upstream commit e905483f87cf; relevant code changes:

• Removing all references to possible_number_pattern other than in the metadata itself.

Merge to upstream commit fedbc7020703; code changes:

• GitHub project changes:
  • Changed tag to vX.Y.Z from release-X.Y.Z; this may affect ports and derived projects.
• Metadata structure changes in XML file:
  • Mobile and Fixed-Line blocks are no longer assumed to inherit missing data from the GeneralDesc, but are treated like every other phone number type. This means that for the non-geographical country codes, like +800, the example number has been moved from generalDesc to the relevant number types, and the code in getExampleNumberForNonGeoEntity has been changed to look at these sub-types for an example number. This also means that the "NA" and "-1" blocks present in the metadata to indicate that no mobile or fixed-line numbers appear for the entity have been removed. There should no longer be an exampleNumber at the generalDesc level, but it should be present at every PhoneNumberDesc with data.
• Code changes:
  • Using new possibleLengthInfo to decide whether a short number is the right length or not. This could result in more specific results; whereas before, a number from length 3 to length 6 may have been deemed possible, now we may exclude a number of length 5.
  • Add hash (#) as a diallable character. Numbers with # in them will no longer have formatting applied in format_n_original_format, and normalize_diallable_chars_only now retains the # symbol.
  • example_number_for_non_geo_entity has been changed to look at the specific number types, not just the generalDesc, for the example numbers; this is a necessary change after the metadata structure change detailed above.

Merge to upstream commit ad0ce0c94501; code changes:

• Fixed phonemetadata.py not to merge from a NumberFormat's unset bool national_prefix_optional_when_formatting.

Merge to upstream commit 2d0d216f6032; code changes:

• Stop setting empty preferred_domestic_carrier_code, and if we are passed such a number then treat the empty field as if unset.

Merge to upstream commit 8c37310deb49; code changes:

• Switching the internal implementation of is_possible_number and related functions to use the new possibleLengths metadata. This affects a lot of countries, making is_possible_number more restrictive as more precise data is available. It also affects parsing ambiguous and invalid numbers, as we decide whether to strip a possible national prefix (1) or country code based on the length of the number and whether it is possible before or after this.
• Formatting, naming and comment tweaks to follow style guide
• Removal of unneeded _can_be_geocoded method in the timezone.py file, using phonenumberutil.py instead

Merge to upstream commit 1ec4d341c3cd; no code changes that affect the Python version so this is just a version bump to stay in sync with upstream.

Merge to upstream commit 7cc500f588db; code changes:

• phonemetadata.py has two more fields to represent possible lengths of phone numbers. Changed buildmetadatafromxml.py to alter the way that metadata about possible-lengths information is consumed when constructing metadata to populate these. Discussion list email

Merge to upstream commit ddf60b1c175e; code changes:

• Made is_number_geographical() public and added is_number_type_geographical(), and changed the geocoder to use this when checking whether to give a detailed answer or country-level only.