Re: [sqlite] Patch for consideration: auto_vacuum slack.

2017-02-14 Thread Chris Brody
On Mon, Feb 13, 2017 at 9:13 PM, Richard Hipp wrote: > [...] > This makes me want to ask: Is anybody still using auto_vacuum? And > if they are, should they be? I am thinking to change the commonly-used Cordova/PhoneGap sqlite plugin [1] to enable auto-vacuum by default for

Re: [sqlite] Patch for consideration: auto_vacuum slack.

2017-02-14 Thread Jens Alfke
> On Feb 13, 2017, at 12:13 PM, Richard Hipp wrote: > > Does anybody really care anymore that a > database file might have a few dozen pages on its freelist? Or if > they do care, does anybody lack the temp space sufficient to run a > real VACUUM? The issue of vacuuming has

Re: [sqlite] Patch for consideration: auto_vacuum slack.

2017-02-13 Thread Scott Hess
On Mon, Feb 13, 2017 at 12:13 PM, Richard Hipp wrote: > > Scott: The motivation for your patch seem to be to get auto-vacuum to > run a little faster. But if performance is your goal, why not just > turn auto-vacuum off? Or, failing that, set it to INCREMENTAL and > then run

Re: [sqlite] Patch for consideration: auto_vacuum slack.

2017-02-13 Thread Stephen Chrzanowski
I use it, not due to drive space, but because I sometimes check my DB into a source code repo. On Mon, Feb 13, 2017 at 3:13 PM, Richard Hipp wrote: > On 2/13/17, Scott Hess wrote: > > > > Below is a patch which implements [PRAGMA auto_vacuum_slack_pages =

Re: [sqlite] Patch for consideration: auto_vacuum slack.

2017-02-13 Thread Donald Shepherd
I use auto-vacuum in my application storage for work. This was introduced in about 2013. The motivation was more political than anything though, as convincing some as to the introduction requiring jumping through some pretty arbitrary hoops. Enabling auto-vacuum was one of those, to mitigate

Re: [sqlite] Patch for consideration: auto_vacuum slack.

2017-02-13 Thread Richard Hipp
On 2/13/17, Scott Hess wrote: > > Below is a patch which implements [PRAGMA auto_vacuum_slack_pages = N]. > This setting allows client code to signal that auto_vacuum can leave pages > on the freelist until releasing them would allow a db size change. This makes me want to ask:

[sqlite] Patch for consideration: auto_vacuum slack.

2017-02-13 Thread Scott Hess
A developer was asking me questions about auto_vacuum I/O characteristics, because they were worried about "churn", where a page is moved to fill a freelist gap, then soon enough a new page is allocated anyhow, so the move wasn't really necessary. This made me wonder if auto_vacuum recognized