Re: [RFC 1/4] hashtable: introduce a small and naive hashtable

From: Sasha Levin
Date: Thu Aug 02 2012 - 13:54:20 EST


On 08/02/2012 07:44 PM, Josh Triplett wrote:
> On Thu, Aug 02, 2012 at 06:48:07PM +0200, Sasha Levin wrote:
>> On 08/02/2012 06:15 PM, Josh Triplett wrote:
>>> On Thu, Aug 02, 2012 at 03:04:19PM +0200, Sasha Levin wrote:
>>>> On 08/02/2012 01:23 PM, Sasha Levin wrote:
>>>>>> #define DEFINE_HASH_TABLE(name, length) struct hash_table name = { .count = length, .buckets = { [0 ... (length - 1)] = HLIST_HEAD_INIT } }
>>>>> The limitation of this approach is that the struct hash_table variable must be 'static', which is a bit limiting - see for example the use of hashtable in 'struct user_namespace'.
>>>>>
>>>>
>>>> What if we just use two possible decelerations? One of static structs and one for regular ones.
>>>>
>>>> struct hash_table {
>>>> size_t bits;
>>>> struct hlist_head buckets[];
>>>> };
>>>>
>>>> #define DEFINE_HASHTABLE(name, bits) \
>>>> union { \
>>>> struct hash_table name; \
>>>> struct { \
>>>> size_t bits; \
>>>
>>> This shouldn't use "bits", since it'll get expanded to the macro
>>> argument.
>>>
>>>> struct hlist_head buckets[1 << bits]; \
>>>> } __name; \
>>>
>>> __##name
>>>
>>>> }
>>>>
>>>> #define DEFINE_STATIC_HASHTABLE(name, bit) \
>>>> static struct hash_table name = { .bits = bit, \
>>>> .buckets = { [0 ... (bit - 1)] = HLIST_HEAD_INIT } }
>>>
>>> You probably wanted to change that to [0 ... ((1 << bit) - 1)] , to
>>> match DEFINE_HASHTABLE.
>>
>> I wrote it by hand and didn't compile test, will fix all of those.
>>
>>> Since your definition of DEFINE_HASHTABLE would also work fine when used
>>> statically, why not just always use that?
>>>
>>> #define DEFINE_STATIC_HASHTABLE(name, bits) static DEFINE_HASHTABLE(name, bits) = { .name.bits = bits }
>>
>> It will get defined fine, but it will be awkward to use. We'd need to pass anonymous union to all the functions that handle this hashtable, which isn't pretty.
>
> No, it'll still use the anonymous union, so you'll still have a thing of
> type "struct hash_table" with the given name, and you can use that name
> with the hash-table functions.

We can use 'struct hash_table' directly, but then the call will look awkward :)

Consider this case (I've placed arbitrary values into size and name:

/* I've "preprocessed" the DEFINE macro below */
union {
struct hash_table table;
struct {
size_t bits;
struct hlist_head buckets[32];
}
} my_hashtable;

void foo(struct hash_table *table)
{
/* Do something */
}

int main(void)
{
foo(my_hashtable); /* This is what the user expects to work, and won't work in this case */

foo(&my_hashtable.table); /* This is what he has to do, which means the user has to know about the internal structure of the union */
}
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/