Re: [PATCH] Sqlite DB Pool

Hi Alex,

why should it be necessary to define lock_timeout? I ask because we 
don't allow it for any DBs from our dbpool why should we make exception 
for sqlite?

Thanks,
Alex

Alejandro Guerrieri schrieb:
> Alex? Did you see this? I've already commited the updates on sqlbox and 
> this would definitely improve sqlite2/3 usability.
> 
> Regards,
> 
> Alejandro
> 
> El 13/11/2008, a las 01:20 a.m., Alejandro Guerrieri escribió:
> 
>> Did anyone reviewed this again? I think it's now fixed.
>>
>> I'm about to commit a big update on sqlbox and it would be great to be 
>> able to have sqlite2 and sqlite3 behaving properly.
>>
>> Regards,
>>
>> Alejandro Guerrieri
>>
>> El 08/11/2008, a las 12:17 p.m., Alejandro Guerrieri escribió:
>>
>>> Ups, > 0 would be better right?
>>>
>>> Changed the datatype to int because that's what the 
>>> sqlite_busy_timeout() expects.
>>>
>>> [Too little coffee that night] ;)
>>>
>>> Fixed on the code plese re-download.
>>>
>>> One last note: the value must be properly initialized, otherwise the 
>>> struct default could be anything. I've checked the Kannel code and 
>>> since there's no dlr_sqlite.c nor any other use for this (yet), no 
>>> further patching is necessary.
>>>
>>> Regards,
>>>
>>> Alejandro Guerrieri
>>>
>>> El 07/11/2008, a las 01:57 p.m., Alexander Malysh escribió:
>>>
>>>> Hi,
>>>>
>>>> hmm, are you sure ? ;)
>>>>
>>>> +    if (conf->lock_timeout != NULL) {
>>>>
>>>> but lock_timeout defined as long...
>>>>
>>>> Thanks,
>>>> Alex
>>>>
>>>> Alejandro Guerrieri schrieb:
>>>>> Hi,
>>>>> I've made this small patch that fixes a couple of things on the 
>>>>> Sqlite implementation:
>>>>> * Sqlite2's [sqlite-connection] group was missing from cfg.def, 
>>>>> though dbpool_sqlite.c and all the hooks were in place. This patch 
>>>>> adds the missing section.
>>>>> * This patch adds a "lock-timeout" configuration option to both 
>>>>> sqlite connection groups, where a value in milliseconds can be 
>>>>> specified. That value controls the time sqlite waits before 
>>>>> throwing a lock ("busy") error. This could be an issue because 
>>>>> sqlite locks at the database level (yuk!), and this can surely be a 
>>>>> problem under moderate load (I wouldn't recommend sqlite for heavy 
>>>>> traffic at all). If not set the default behaviour is maintained, so 
>>>>> nothing's broken ;)
>>>>> Please check the post here:
>>>>> http://www.blogalex.com/archives/50
>>>>> Direct download:
>>>>> http://www.blogalex.com/wp-content/uploads/2008/11/kannel-sqlite-pool-patches.diff BTW, 
>>>>> my long-term goal is to have on Sqlbox the same DB capabilities as 
>>>>> Kannel have (Sqlbox only supports MySql and Postgres and _maybe_ MS 
>>>>> SQL). I'm making good progress with sqlite so far, and this patches 
>>>>> aim to complete the Kannel support to make it practical to use the 
>>>>> DB pool with Sqlbox.
>>>>> MS SQL by Ct-Lib is supposed to be working, though is not supported 
>>>>> at Kannel afaik (I'll try to port the code to Kannel so it can be 
>>>>> linked from gwlib as the rest does when I have some spare time).
>>>>> Regards,
>>>>> Alejandro Guerrieri
>>>>
>>>>
>>>
>>
> 
> 
>