Path: csiph.com!newsfeed.hal-mli.net!feeder3.hal-mli.net!news.stack.nl!newsfeed.xs4all.nl!newsfeed4a.news.xs4all.nl!xs4all!newsgate.cistron.nl!newsgate.news.xs4all.nl!post.news.xs4all.nl!not-for-mail
To: python-list@python.org
From: Peter Otten <__peter__@web.de>
Subject: Re: Python's re module and genealogy problem
Date: Fri, 13 Jun 2014 18:26:55 +0200
Organization: None
References: <bvr01iFu926U1@mid.individual.net> <c00ivgF5cjpU1@mid.individual.net>
Mime-Version: 1.0
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 7Bit
User-Agent: KNode/4.11.5
Precedence: list
Newsgroups: comp.lang.python
Message-ID: <mailman.11059.1402676838.18130.python-list@python.org>
Lines: 79
NNTP-Posting-Host: 2001:888:2000:d::a6
Xref: csiph.com comp.lang.python:73269

BrJohan wrote:

> On 11/06/2014 14:23, BrJohan wrote:
>> For some genealogical purposes I consider using Python's re module.
>>
>> Rather many names can be spelled in a number of similar ways, and in
>> order to match names even if they are spelled differently, I will build
>> regular expressions, each of which is supposed to match  a number of
>> similar names.
>>
>> I guess that there will be a few hundred such regular expressions
>> covering most popular names.
>>
>> Now, my problem: Is there a way to decide whether any two - or more - of
>> those regular expressions will match the same string?
>>
>> Or, stated a little differently:
>>
>> Can it, for a pair of regular expressions be decided whether at least
>> one string matching both of those regular expressions, can be
>> constructed?
>>
>> If it is possible to make such a decision, then how? Anyone aware of an
>> algorithm for this?
> 
> Thank you all for valuable input and interesting thoughts.
> 
> After having reconsidered my problem, it might be better to approach it
> a little differently.
> 
> Either to state the regexps simply like:
> "(Kristina)|(Christina)|(Cristine)|(Kristine)"
> instead of "((K|(Ch))ristina)|([CK]ristine)"
> 
> Or to put the namevariants in some sequence of sets having elements like:
> ("Kristina", "Christina", "Cristine", "Kristine")
> Matching is then just applying the 'in' operator.
> 
> I see two distinct advantages.
> 1. Readability and maintainability
> 2. Any namevariant occurring in just one regexp or set means no risk of
> erroneous matching.
> 
> Comments?

I like the simple variant

kristinas = ("Kristina", "Christina", "Cristine", "Kristine")

But instead of matching with "in" you could build a dict that maps the name 
variants to a normalised name

normalized_names = {
    "Kristina": "Kristina",
    "Christina": "Kristina",
    ...
    "John": "John",
    "Johann": "John",
    ...
}
def normalized(name):
    return normalized_names.get(name, name)

If you put persons in another dict or a database indexed by the normalised 
name 

lookup = {
    "Kristina": ["Kristina Smith", "Christina Miller"],
    ...
}

you can find all Kristinas with two look-ups:

>>> lookup[normalized("Kristine")]
['Kristina Smith', 'Christina Miller']

PS: A problem with this approach might be that (name in nameset_A) and (name 
in nameset_B) implies nameset_A == nameset_B