[erlang-questions] Mnesia fragmented tables: mnesia:dirty_next and mnesia:dirty_first

[Evans, Matthew]

Thanks for the information.

I have another question about QLC. How is this constructed internally?

In my case the table is disc_only_copies, does QLC read X records into RAM? Or the whole table?

Thanks

Matt

-----Original Message-----
From: Igor Ribeiro Sucupira [mailto:igorrs@REDACTED] 
Sent: Wednesday, June 09, 2010 8:14 PM
To: Evans, Matthew
Cc: erlang-questions@REDACTED
Subject: Re: [erlang-questions] Mnesia fragmented tables: mnesia:dirty_next and mnesia:dirty_first

On Wed, Jun 9, 2010 at 7:59 PM, Evans, Matthew <mevans@REDACTED> wrote:
> Hi,
>
> I have a fragmented database where I want to do an mnesia:dirty_first(Tab), and mnesia:dirty_next(Tab,Key).
>
> These don't appear to be supported.

The mnesia_frag module can probably handle that, if you call
mnesia:activity/4 like this:
F = mnesia:activity(sync_dirty, fun mnesia:first/1, [my_table], mnesia_frag).
mnesia:activity(sync_dirty, fun mnesia:next/2, [my_table, F], mnesia_frag).
...

But maybe it will be more efficient to use a cursor, after creating it
inside of a dirty operation:
Cursor = mnesia:activity(sync_dirty, fun qlc:cursor/1, [qlc:q([R || R
<- mnesia:table(my_table)])], mnesia_frag).

However, if you need to be able to handle errors (e.g. an unavailable
fragment) smoothly (retrieve all the data that is available), you
should actually do something like what you are already doing. Being
that the case, I recommend that you read this post:
http://igorrs.blogspot.com/2010/05/mnesia-one-year-later-part-2.html

Good luck.
Igor.

> I have got around this by:
>
> 1) Using mnesia:table_info/2 to get the frag hash state.
>
> 2) Create a list of all my fragments (as atoms), in a specific order.
>
> The above data is saved in a state record.
>
> Then whenever I get a get_next I call mnesia_frag_hash:key_to_frag_number/2, and from the list of fragments I saved in 2, above, I can get the actual fragment to search in.
>
> I can then call mnesia:dirty_next(Fragment,Key). If I get '$end_of_table' I can move to the next fragment in the list of fragments and so on until the list of fragments is exhausted.
>
> Other than doing something similar with mnesia:dirty_slot/2 is there any other way to do a get_next traversal of a fragmented table?
>
> Thanks
>
> Matt


-- 
"The secret of joy in work is contained in one word - excellence. To
know how to do something well is to enjoy it." - Pearl S. Buck.