dhis2-devs team mailing list archive

[Bob Jolliffe <bobjolliffe@xxxxxxxxx>]

2010/3/4 Lars Helge Øverland <larshelge@xxxxxxxxx>:
>
>
> 2010/3/4 Bob Jolliffe <bobjolliffe@xxxxxxxxx>
>>
>> I agree depending on the uniqueness of names is not a good idea.  I
>> saw John said that facilities in India do have codes - the 16 digit
>> ones which are built hierarchicly (is that a word?).
>
> OK I wasn't aware of this code.. Certainly if we could generate a string
> based on the orgunit code plus its ancestors it will be unique. How does one
> generate a fixed-length alphanumeric string based on this sequence of names
> btw?

I'm thinking that we just generate the local part.  The facility codes
already exist.  So in Db you would have fields for patient id and
issuing-facility. I think it is common to store the issuing authority
with an id anyway.  In most cases the issuing-facility might be null
because it could be assumed.  If a patient is moved into the db from
somewhere else, or if patients are combined from different facilities
for one of many reasons, then the facility code would need to be
populated to avoid clashes.  This bears some resemblance I think to
the uuid disambiguation referred to by Saptarshi.

Regards
Bob

>>
>> Is it possible to for a patient id to continue this hierarchy so that
>> a patient might have say a 6-8 digit local identifier - preferably
>> Base30 as I see Saptarshi has just chimed in.  But his
>> "fully-qualified" id would be the 16 digit one + the local part.  I
>> can see that generation and allocation of these might be problematic
>> and the internal uuid might be a good (if expensive) failsafe.  How
>> does openmrs deal with this?  Saptarshi, is at as you have suggested?
>> I would be a bit concerned that management of a pool of ids strikes me
>> as something which could easily fall apart.  Isn't it better to
>> generate them on demand from some random source and test for
>> uniqueness before inserting into the database?  If its not unique then
>> it simply tries another till its happy?
>>
>> Bob
>>
>>
>> 2010/3/4 Lars Helge Øverland <larshelge@xxxxxxxxx>:
>> >
>> >
>> > Been chatting a bit with John and he expressed concern about the
>> > orgunit-randomnumber apprach. In India there are multiple installations
>> > and
>> > one cannot know for sure that an orgunit name will be unique. How do we
>> > deal
>> > with this?
>> > Using a globally unique identifier could be a solution, but the standard
>> > Java implementation (UUID) uses 32 characters and is a bit long. Is
>> > implementing our own, shorter one an option?
>> >
>> > Lars
>> >
>> >
>
>