Re: [sqlite] SQLite with branching

2019-11-10 Thread Robert M. Münch
On 4 Nov 2019, at 22:25, Jens Alfke wrote: >> On Nov 4, 2019, at 4:57 AM, Simon Slavin wrote: >> >> That's one of the reasons that the source code for SQLite is public: so that >> people can add the features they want. > > Totally agree. However, when you go off the mainline of SQLite you lose

Re: [sqlite] SQLite with branching

2019-11-08 Thread Bernardo Ramos
I included WAL mode and mmap on the LiteTree simple benchmark. It turns out that WAL mode is as fast as LiteTree on Linux (with a hard disk) for writes and a little slower on reads. On MacBook Pro (with SSD) LiteTree is faster on both writing and reading. SQLite's mmap make it slightly

Re: [sqlite] SQLite with branching

2019-11-07 Thread Jens Alfke
> On Nov 7, 2019, at 9:02 AM, Bernardo Ramos wrote: > > If you are interested in just the performance without the branching feature, > there are at least 3 options: > > 1. SQLigthning: I was thinking in updating it to the last version of SQLite That would be awesome! I have looked at it a

Re: [sqlite] SQLite with branching

2019-11-07 Thread Bernardo Ramos
Hi! I am the creator of LiteTree (also LiteReplica, LiteSync and 3 new products that will be released soon). When I was planning to add branching I discovered many ways to implement it. I selected the one that satisfied performance over disk usage. It can also be implemented the other way

Re: [sqlite] SQLite with branching

2019-11-05 Thread Jens Alfke
> On Nov 5, 2019, at 1:27 AM, Dominique Devienne wrote: > > AFAIK, that was one of the goals of SQLite4 [1], to change the backend to LSM. LMDB (LiteTree's back-end) doesn't use LSM; it's a B-tree manager. The speedup appears to come from a combination of techniques like eliminating caching

Re: [sqlite] SQLite with branching

2019-11-05 Thread Dominique Devienne
On Tue, Nov 5, 2019 at 10:01 AM Wout Mertens wrote: > On Mon, Nov 4, 2019 at 10:26 PM Jens Alfke wrote: > > > I don't have a practical use for the branching features, though they're > cool, but I'm salivating at the thought of a 2x speedup. > > With all the work that's put into eking out small

Re: [sqlite] SQLite with branching

2019-11-05 Thread Wout Mertens
On Mon, Nov 4, 2019 at 10:26 PM Jens Alfke wrote: > I don't have a practical use for the branching features, though they're cool, > but I'm salivating at the thought of a 2x speedup. > With all the work that's put into eking out small performance increases in > SQLite, I'd imagine the devs

Re: [sqlite] SQLite with branching

2019-11-04 Thread Jens Alfke
> On Nov 4, 2019, at 4:57 AM, Simon Slavin wrote: > > That's one of the reasons that the source code for SQLite is public: so that > people can add the features they want. Totally agree. However, when you go off the mainline of SQLite you lose some things, like easy updating to new SQLite

Re: [sqlite] SQLite with branching

2019-11-04 Thread Simon Slavin
The post you quoted points to exactly that: a version of SQLite that handles branches. Check it out. That's one of the reasons that the source code for SQLite is public: so that people can add the features they want. ___ sqlite-users mailing list

Re: [sqlite] SQLite with branching

2019-11-04 Thread Robert M. Münch
@SQLite Guys: Do you have something like branching on your roadmap? I really like this feature and see a lot of use-cases beside the blockchain topic. And, of course if this works with your encryption extension that would be awesome. * simple versioning of a database: Useful when you want to

[sqlite] SQLite with branching

2018-08-29 Thread Simon Slavin
I have no connection with the following project. Described poorly on the web site so here's my own description: This is an extension of SQLite which allows branched versions, each new branch creating one dataset which existed before the new branch and a