Re: How to use associative arrays in Ada 2005?

[Simon Wright <simon@pushface.org>]

Matthew Heaney <mheaney@on2.com> writes:

> Simon Wright wrote:
>>
>> There's no reason I can see to use maps of maps *for this problem*. I
>> would use a hash function (Nation x Family_Name x Name) => Hash_Type
>> in a single map.
>
> Yes, that's right, you could do it either way.
>
> I prefer that OP's approach of using a map of maps of maps, since that
> allows you to perform queries like "give me all the entries for this
> nation", etc.

But that only works for the first map. You couldn't find all the
Smiths that way.

If that's what you're after, you need (in database terms) a second
independent index.

>> Matt, I see (A.18.4(5)) that Ada.Containers doesn't allow hash
>> collisions (which, if true, is demanding a lot of hash function
>> designers). 
>
> You're misreading that paragraph.  It does say that:
>
> "A map never contains two or more nodes with equivalent keys."
>
> but that's very different from saying "hash collisions aren't allowed."

I've just read (and re-read, and re-read) A.18.5(43) and I see what
you mean. Perhaps it's a bit late. I had read 'same hash =>
equivalent' whereas it's actually 'equivalent => same hash'.

> If you have as your key a record comprising the 3 strings (nation,
> family name, given name), then "=" would be adequate for
> Equivalent_Keys.

I would have a record with (nation, family_name, given_name, age) --
if you have filtered out a container with just the Smiths you'd want
to know the other facts for each of them, too.