nameutils - Identify given/family names and capitalize correctly
from nameutils import (
namecase, gnamecase, fnamecase, namecase_exception,
namesplit, nameparts, namesplit_exception, namejoin,
nametrim, normalize
)
# Case functions
full_name = namecase(full_name)
given_names = gnamecase(given_names) # i.e. Given name(s) only
family_name = fnamecase(family_name) # i.e. Family name only
family_name = fnamecase(family_name, given_names) # Individual exceptions
namecase_exception("Fitzell") # Add an exception for all members of a family
namecase_exception('DeVries', 'DiFrancesco') # Add more exceptions
namecase_exception("Marrier D'Unienville, Jean") # Add an individual exception
# Split functions
full_name = namesplit(full_name) # Format as "Family_name, Given_names"
(family_name, given_names) = nameparts(full_name) # Format as an array
namesplit_exception("Bryant Smith, Denise") # Multi-name family names
full_name = namejoin(family_name, given_names)
# Trim function
name = nametrim(name)
# Unicode normalization of internal data (default is NFC)
import unicodedata
normalize(lambda s: unicodedata.normalize('NFD', s)))
This module is useful when receiving a person's name that might be all uppercase, or in the wrong case, or it might have the given names and the family name combined in a single string (e.g., a single spreadsheet column), and you need to split the full name into its parts, and you want to set the capitalization correctly so as to show each person a little respect by taking the trouble to at least try to get their name right.
Getting the case right for people's names is difficult, and many software systems address this problem by not even trying, and using uppercase exclusively. It's ugly, but it's easy and consistent. We can do better. It can't be perfect, by default, but with ongoing adjustments to suit your evolving dataset, you can improve it to meet your needs.
People with complex grammatical aristocratic/topographic/patronymic family names often don't know how their own names should be capitalized. Or at least, they don't know how their own ancestors capitalized their name, or they know, but they disagree with it. Some people insist on having it their own way, and that's fine. This module, by default, prefers how their ancestors would have capitalized their names, but people can do whatever they want to their own names, and it's important to them, so this module supports general exceptions that apply to everyone with a particular family name, for when the default behaviour is definitely wrong, and it also supports exceptions that apply only to individuals who report that it is wrong for them.
Note: This module doesn't handle every name on Earth. Apart from Chinese, Japanese, and Korean family names, it only understands names written in Latin scripts, except perhaps by lucky accident. For example, names in Cyrillic work. It doesn't handle honorifics, titles, joined initials, or postnominals. It only handles names. But it does handle complex names coming from a variety of places (e.g., British Isles, Europe, Middle East, Africa, East Asia, Pacifika, Americas). By default, it doesn't correctly identify unhyphenated multi-name family names (like Spanish and Catalan names, unless the formal "y" or "i" is present). Such names need to be handled with split exceptions. It handles some mixed case names like McAdam, MacArthur, FitzSimmons, DeVito, VanZandt, etc., but there will be false negatives (and arguably false positives) which can be corrected with case exceptions. Over time, you will build up a set of case exceptions and split exceptions that meets the needs of your dataset.
Returns the supplied name with the capitalization fixed. See the EXAMPLES section below to see exactly what this means. This can be called in several ways:
For a full name (implicitly):
full_name = namecase("JOHN PETER SMITH")
full_name = namecase("SMITH, JOHN PETER")
For a full name (explicitly):
full_name = namecase("JOHN PETER SMITH", 'full')
full_name = namecase("SMITH, JOHN PETER", 'full')
Note that the full name can be supplied in the (ambiguous) natural order, with the given name(s) followed by the family name, or unambiguously, with the family name followed by a comma followed by the given name(s).
For a given name or names by itself (same as gnamecase(), see below):
given_names = namecase("JOHN PETER", 'given')
For a family name by itself (same as fnamecase(), see below):
family_name = namecase("SMITH", 'family')
For a family name by itself when you have (or might one day have) any case exceptions intended to only affect a single individual (same as fnamecase(), see below):
family_name = namecase("SMITH", 'family', "JOHN PETER")
Returns the supplied family name with the capitalization fixed. Same as: namecase(family_name, 'family'[, given_names]).
The given_names argument is technically optional, but it should be supplied, just in case you ever need case exceptions that only apply to an individual. This enables people with the same family name to have their use of that family name capitalized the way they want it to be. Once you have a need for such individual case exceptions, the given_names argument will become necessary everywhere, so it's best to supply it from the start.
Add one or more case exceptions. Whenever the above case functions subsequently capitalize the supplied name, the supplied capitalization will be returned, rather than the default behaviour.
There are two kinds of case exception. Some apply to everyone that shares a family name, and some apply to an individual.
Family-wide exceptions contain the family name, capitalized correctly:
namecase_exception("DiBona")
Individual case exceptions must be supplied as unambiguous full names in the form: "Family_name, Given_names", capitalized as specified by the named person:
namecase_exception("DiBona, John")
Returns 1 if the exception was successfully added. Returns 0 otherwise. The only reason for a failure is if the supplied exception is undefined or empty.
Returns the supplied full name converted to the unambiguous form: "Family_name, Given_names" with the capitalization fixed.
The full_name argument is expected to be in either form: "Given_names Family_name" or "Family_name, Given_names". Note that a string is returned with the family name followed by a comma and space followed by the given name(s).
Complex grammatical aristocratic/topographic/patronymic family names in Latin script are identified. See the EXAMPLES section below. But unhyphenated multi-name family names are not correctly identified by default. That requires split exceptions (see below). Spanish and Catalan multi-name family names are correctly identified when the two names are joined with "y" or "i", but when the joining word is not present, a split exception is needed.
With Chinese, Japanese, and Korean names, the family name appears first when written in their own scripts/characters. When romanized, Chinese and Korean family names might appear first or last. The same is true for Vietnamese names.
This module recognizes the 400 or so most common Chinese family names (97% of the population) in Chinese characters and in one romanized spelling, and additionally, the 100 most common Chinese family names (85% of the population) in pinyin and various other romanized spellings, as used in several countries. It also recognizes the 190 most common Korean family names (98% of the population) in Hangul, Hanja, and romanized. It also recognizes the 209 Vietnamese family names (100% of the population, apparently). There are too many Japanese family names (over 300,000) to maintain a list of them, so this module employs a statistical method to identify Japanese family names written in Kanji and Kana. I don't know what proportion of Japanese-named population it identifies.
Returns the supplied full name converted to a two-element array containing the family name and the given name(s), with the capitalization fixed.
The full_name argument is expected to be in either form: "Given_names Family_name" or "Family_name, Given_names".
This function converts the corresponding return value of namesplit() into a two-item array. See namesplit() above for more details. If the name contains a single "word", then it isn't splittable, and so a one-element array is returned. If the name is the empty string or undefined, then a zero-element array is returned
Chinese, Japanese, and Korean names in their own scripts/characters contain multiple words even though they don't contain spaces between them. If a full name is supplied, this function will return a two-element array.
Add one or more split exceptions. The exceptions must be supplied as full names in the unambiguous comma-separated form with the family name followed by a comma and space followed by the given name(s).
This is needed to support unhyphenated multi-name family names that aren't automatically identified, such as "Ah Mu, Corie", and even complex given names that would be misrecognized, such as "de Sousa, Fatima de Gois".
This is also needed to correct the situation when this module misidentifies the type of name, and splits it incorrectly. For example, a Japanese name with a family name consisting of two characters might be misidentified as a Chinese name with a family name consisting of one character.
Returns 1 if the exception was successfully added. Returns 0 otherwise. The only reason for a failure is if the supplied exception does not contain a comma.
Returns the full name composed of the supplied family name and given names.
Chinese, Japanese, and Korean names in their own scripts/characters are concatenated in Eastern name order (with the family name on the left). All names in Latin and other scripts are joined in Western name order (with the family name on the right), and with a space character added between the given names and the family name. Note that romanized Chinese, Japanese, and Korean names, and Vietnamese names, are always joined in Western name order.
Normalize this module's internal data using the supplied Unicode normalization function reference so as to match your application's choice of normalization. A likely choice would be NFD.
This is needed if the application's choice of Unicode normalization differs from whatever was used for the module's internal data in the module source code (i.e., NFC). A difference in normalization can lead to false negatives and incorrect results when matching names against internal data.
These examples show the default namecase() output for various forms of names. They also show which name forms are automatically recognized by namesplit(). Note that non-ASCII letters and punctuation in these examples have been replaced with the closest ASCII equivalents to avoid problems with some implementations of *roff. namesplit() also supports names in Chinese characters, Korean Hangul and Hanja, and Japanese Kanji and Kana, but they are not shown here for the same reason:
John Peter Smith William Maitland of Lethington Shaun McAdam Fergus MacDonald Lachlan Macquarie James FitzPatrick Patrick O'Brian Kelly St Clair David Le Page Pierre La Tour Rochelle Li Donni Giovanni Lo Giudice Estella d'Iapico-Bien Bruno dall'Agnese Bruno dell'Agnese Lorenzo de' Medici John de Groot Pierre de la Pierre Maria del Mar Maria dela Mar Maria dels Angels Giaccomo della Vella Giovanni delle Velle Maria dal Santos Marco dalla Vella Lorenza degli Castelli Maria di Francesco Giuseppe Tomasi di Lampedusa Pierre du Page Jorge da Silva Filipe do Santo Abilio dos Santos Adriana das Costas Oscar San Jose Catalina Santa Gutierrez Monica Santos Bernal Pablo Diego Ruiz y Picasso Carles Puigdemont i Casamajo Joao Duarte da Silva dos Santos da Costa de Sousa Joao Duarte da Silva Santos Costa e Sousa Hans von Pappenhim Thomas von der Trave Hans zu Pappenhim Hans von und zu Pappenhim Bram van Haag Jeroen der Haag Johanne ter Horst Sanne den Haag Laura van de Horst Eva van der Haag Willem van den Haag Mees van het Horst Henrik van Voorst tot Voorst Willem 'sGravesande Gemeente van 'sHertogenbosch Gemeente van 'tHoen Sigurd av Morgenstierne Maja von Munthe af Morgenstierne Lars Jonsson til Sudreim James DaSilva Jack DuBois Daniel LaForge Sally LeFevre Kristine VanZandt Patrick O Donoghue Micheal O hAodha Saoirse Ni Fhoghlua Michael Mac Donnchada Saoirse Nic Fhoghlua Michael Ua Donoghue Aisling Bean Ui Fhoghlua Saoirse Bean Mhic Fhoghlua Saoirse Ui Fhoghlua Saoirse Mhic Fhoghlua Rhys ap Dafydd Maredudd ab Owain Myfanwy ferch Maredudd Myfanwy verch Maredudd Camilla El Ali Mariam Al Musawi Bazif el-Bayeh Nariman al-Nassar Hizb ut-Tahrir Aziz ibn Hab Charbel bin Hab Angela bint Aziz Fatima binti Aziz Nadia binte Aziz David Ben Joseph # Incorrect when technically ambiguous ben Joseph, David # Correct when technically unambiguous David ben Joseph v'Rachel # Correct - this is really not ambiguous Leah bat Moshe Leah bat Moshe v'Rachel ha-Rav Devorah Rut bat Mordecai v' Tzipporah Leah mibeit Moshe v'Rachel ha-Levi Leah mimishpachat Moshe v'Rachel ha-Kohein Natalie Te Whare Ayize ka Nolwazi Oso'ese Ya'akov Y'honatan Sh'mu'el Onosa'i Tausa'afia Ka'ana'ana S'thembiso
It's impossible to actually do what this module attempts to do in a way that works correctly for everybody by default. There are too many people who want their names cased incorrectly (e.g., Da Vinci rather than da Vinci), and too many unhyphenated multi-name family names, and so many languages. This module handles complex grammatical aristocratic/topographic/patronymic (romanized) family names from various languages (e.g., French, Italian, Spanish, Catalan, Portuguese, English, Irish, Welsh, Scottish, German, Dutch, Swedish, Norwegian, Danish, Finnish, Zulu, Arabic, Hebrew), but there are many more languages that it doesn't know about. So, in order to keep all of your users happy, you will almost certainly need to build up your own list of case and split exceptions in a file or database, and have your application load them before processing any names. But if two people with exactly the same full name both insist on having their name capitalized differently to each other, that's not supported.
Different languages can have different case conventions for the same "word". For example, a Greek family name can start with el, but a Spanish family name can start with El. This module favours the most likely case (i.e., Spanish in this example). For other cases, this can be corrected with case exceptions.
Similarly, a Hebrew patronymic name can start with ben, but an Anglo given name can be Ben. That's fine if fnamecase() or gnamecase() are used, supplied with just the given name(s) or the family name, respectively, but namecase() supplied with an ambiguous full name will favour the Anglo interpretation. That is, unless the name contains other elements that make it obviously Hebrew, such as a matronymic component (e.g., v'Rachel), or a suffix such as ha-Rav. That will cause a technically ambiguous ben to be correctly identified as a patronymic prefix.
Similarly, when it comes to splitting/identifying the given names and the family name within a full name with namesplit() or nameparts(), the word ben is not interpreted as the start of a patronymic name (in the absence of other clues as indicated above), because Ben is more likely to be an Anglo middle name (although bat is always interpreted as the start of a patronymic name). Luckily, Hebrew names aren't used much outside of religious contexts, so this hopefully won't be much of a problem for this module. If it is, it can be corrected with split exceptions (or with more detailed Hebrew names).
Romanized Chinese, Korean and Vietnamese family names can appear at the start or at the end of a full name. This module detects them in either format. But there can be false positives when a Chinese or Korean family name is at the end of the full name, and the given name also looks like a Chinese or Korean family name. For example, namesplit() works better for Korean names where the family name appears at the start rather than at the end, because some Korean given names look like a family name. Other odd cases might arise due to never knowing which language a romanized name is from. But split exceptions should help when these cases are noticed.
Unicode strings are complicated. Some graphemes can occur in multiple ways. Any case and split exceptions are looked up via a hash key match. To increase the chance of matches succeeding when they should, you should probably normalize strings on input to your application using something like NFD (or maybe even Stringprep) before passing names to this module which assumes that any necessary preparation has already been done. If necessary, you can normalize this module's internal data (with normalize()) to match your application's choice of normalization.
The nameparts() function probably should have been called namesplit(), because it returns an array, and the namesplit() function probably should have been called something else, because it returns a string. But they are the names I'm used to, and I couldn't think of anything better, and now it's too late to change it.
Space characters are not preserved. Spaces at the start or end of a name are removed, as are spaces before commas, and before and after hyphens. There will always be a space after a comma. Any non-ASCII spaces are replaced with ASCII space. Let me know if that's a problem. It can probably be fixed, but I think it's a feature. If it's any consolation, non-ASCII apostrophe-like characters and hyphen-like characters are preserved. But if there is a case exception involving any apostrophe-like or hyphen-like characters, then they too are replaced by the actual character specified in the exception.
A (less comprehensive) version of this module (in another language) has been in use for over fifteen years at a small company with a dataset of about fifty thousand names. With that dataset, six generic case exceptions were needed, two individual case exceptions, and about nine hundred split exceptions.
It enabled the accurate identification of names in spreadsheets so as to check against ID number columns, and it made reports containing people's names much prettier than they would otherwise have been.
unicodedata, stringprep
20230709 raf <raf@raf.org>
Copyright (C) 2023 raf <raf@raf.org>
This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3 of the License, or (at your option) any later version.