Sqlite and performances in LiveCode

[Jérôme Rosat]

Thank you Pete  for that clarification.

Jérôme

Le 7 mars 2012 à 17:37, Pete a écrit :

> I emailed sqlitemanager support re the manage/sql tab performance
> difference. Mystery solved by the reply:
> 
> "In the Manage panel when you see SELECT * FROM table the real query
> executed is SELECT rowid FROM table and then a SELECT * FROM table WHERE
> rowid=N is execute for each visible row of the listbox. In the SQL panel
> query executed is just what you type"
> On Mar 6, 2012 5:47 PM, "Pete" <pete at mollysrevenge.com> wrote:
> 
>> I just noticed that the same vendor that makes SQLiteManager also makes an
>> SQLite server package (cubeSQL) that does implement MVCC.  But you're
>> right, it's definitely not there in standard SQLite. An SQLite server
>> sounds interesting though.
>> Pete
>> 
>> On Tue, Mar 6, 2012 at 11:35 AM, Bernard Devlin <bdrunrev at gmail.com>wrote:
>> 
>>> I notice from things discussed on the Valentina list that they
>>> recommend using API methods rather than SQL calls, because the former
>>> will be faster due to less overhead.  So these different ways of
>>> interacting with a sqlite database might well represent an application
>>> using the sqlite API.
>>> 
>>> Although when I use the Valentina VStudio SQL interface I will still
>>> see results being returned in 25 ms, compared to 500 ms going through
>>> sql yoga > revdb > v4rev sql > valentina.  This is doing a "select *"
>>> from a table with 22k rows (10mb in size).  That's not to say that I'm
>>> outraged at 500ms.  There are other steps one can take to make an
>>> interface responsive at these timescales.
>>> 
>>> The Valentina guys are pretty used to comparing Livecode with the
>>> other platforms for which they provide Valentina.  On 19/12/2010
>>> Ruslan had this to say about Livecode:
>>> 
>>> "REALbasic - as I remember -- was very fast, near to C code...so it
>>> gives e.g. 5% overhead to pure C engine... Revolution and Director
>>> give me x4-x10 overhead I think....Revolution because all params are
>>> packed to strings ... Director because quite complex SDK similar to MS
>>> COM ..."  This conclusion by Ruslan appears to be only in the
>>> discussion of calling the Valentina API, therefore one would expect
>>> that adding other layers (sql, revdb, etc.) one would see other layers
>>> adding their own torpor to the action.
>>> 
>>> Does using explict transation control make any difference with sqlite?
>>> It shouldn't matter in a case where the data is only being read. I
>>> know on MVCC databases this is still a factor if one has a cursor
>>> containing the read data, as such data might still be altered from the
>>> cursor and the currency control of the MVCC requires a handle on such
>>> data to be able to maintain consistency.  I can't see that it would
>>> apply to sqlite, and this site confirms that sqlite does not implement
>>> MVCC: http://www.sqliteconcepts.org/SI_index.html
>>> 
>>> Bernard
>>> 
>>> On Tue, Mar 6, 2012 at 5:15 PM, Pete <pete at mollysrevenge.com> wrote:
>>>> I got the same results on the movies database and even more pronounced
>>>> differences on a table in one of my own databases with ~48k records in
>>> it:
>>>> SQL tab 0.519secs, Manage tab 0.045 secs, more than a 10-fold
>>> difference!
>>>> Just a straight SELECT * in each case.
>>>> Pete
>>> 
>>> _______________________________________________
>>> use-livecode mailing list
>>> use-livecode at lists.runrev.com
>>> Please visit this url to subscribe, unsubscribe and manage your
>>> subscription preferences:
>>> http://lists.runrev.com/mailman/listinfo/use-livecode
>>> 
>>> 
>> 
>> 
>> --
>> Pete
>> Molly's Revenge <http://www.mollysrevenge.com>
>> 
>> 
>> 
> _______________________________________________
> use-livecode mailing list
> use-livecode at lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-livecode