On 16 Jun 2010, at 5:33am, Jay A. Kreibich wrote:
> On Tue, Jun 15, 2010 at 08:58:22PM -0700, Roger Binns scratched on the wall:
>> On 06/15/2010 07:59 PM, Simon Slavin wrote:
>>> The standard assumption about SQLite is that it's faster to do your INSERTs
>>> first, then create the indices.
>>
On 16 Jun 2010, at 5:04am, Eric Smith wrote:
> Simon Slavin wrote:
>
>> The standard assumption about SQLite is that it's faster to do your
>> INSERTs first, then create the indices. How much of a difference this
>> makes depends on a lot of things.
>
> On what things does it depend?
Not
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 06/15/2010 09:33 PM, Jay A. Kreibich wrote:
> Contiguous, yes, but the pages may not be in any logical order.
> The internal node pages will get shuffled as the tree is built,
> meaning you might still have a significant number of seeks.
Seek
On Tue, Jun 15, 2010 at 08:58:22PM -0700, Roger Binns scratched on the wall:
> -BEGIN PGP SIGNED MESSAGE-
> Hash: SHA1
>
> On 06/15/2010 07:59 PM, Simon Slavin wrote:
> > The standard assumption about SQLite is that it's faster to do your INSERTs
> > first, then create the indices.
>
> I
Simon Slavin wrote:
> The standard assumption about SQLite is that it's faster to do your
> INSERTs first, then create the indices. How much of a difference this
> makes depends on a lot of things.
On what things does it depend?
--
Eric A. Smith
Sendmail may be safely run set-user-id to
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 06/15/2010 07:59 PM, Simon Slavin wrote:
> The standard assumption about SQLite is that it's faster to do your INSERTs
> first, then create the indices.
If the index is created first then the data in the table and the pages
making up the index wil
On 16 Jun 2010, at 3:55am, Eric Smith wrote:
> Let's say my app has (only) inserts followed by (only) reads.
>
> The reads are best served by some indices. So I can create the indices
> before the INSERTs, or after them.
>
> In general, should I expect a run time perf difference between t
Let's say my app has (only) inserts followed by (only) reads.
The reads are best served by some indices. So I can create the indices
before the INSERTs, or after them.
In general, should I expect a run time perf difference between these two
options?
Eric
--
Eric A. Smith
Louis Paste
8 matches
Mail list logo